conv1d-generic-forest-cover.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:25.168504Z",
     "start_time": "2020-09-19T11:51:25.086472Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 1;\n",
       "                var nbb_unformatted_code = \"%load_ext nb_black\\n%load_ext autoreload\\n\\n%autoreload 2\";\n",
       "                var nbb_formatted_code = \"%load_ext nb_black\\n%load_ext autoreload\\n\\n%autoreload 2\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%load_ext nb_black\n",
    "%load_ext autoreload\n",
    "\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:26.307219Z",
     "start_time": "2020-09-19T11:51:25.169593Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 2;\n",
       "                var nbb_unformatted_code = \"import os\\nfrom pathlib import Path\\n\\nfrom requests import get\\nimport pandas as pd\\nimport numpy as np\\n\\nfrom sklearn.metrics import accuracy_score, balanced_accuracy_score\\nfrom sklearn.preprocessing import LabelEncoder\\nfrom tensorflow.keras.utils import plot_model\\nfrom tensorflow.keras.callbacks import EarlyStopping\\n\\nimport logging\\n\\nlogging.basicConfig(level=logging.WARN)\";\n",
       "                var nbb_formatted_code = \"import os\\nfrom pathlib import Path\\n\\nfrom requests import get\\nimport pandas as pd\\nimport numpy as np\\n\\nfrom sklearn.metrics import accuracy_score, balanced_accuracy_score\\nfrom sklearn.preprocessing import LabelEncoder\\nfrom tensorflow.keras.utils import plot_model\\nfrom tensorflow.keras.callbacks import EarlyStopping\\n\\nimport logging\\n\\nlogging.basicConfig(level=logging.WARN)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import os\n",
    "from pathlib import Path\n",
    "\n",
    "from requests import get\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.metrics import accuracy_score, balanced_accuracy_score\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from tensorflow.keras.utils import plot_model\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "\n",
    "import logging\n",
    "\n",
    "logging.basicConfig(level=logging.WARN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:26.479022Z",
     "start_time": "2020-09-19T11:51:26.308582Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 3;\n",
       "                var nbb_unformatted_code = \"from xplainet.input_utils import preproc_dataset\\nfrom xplainet.model import build_model\\nfrom xplainet.random_utils import setup_seed, SEED\\nfrom sklearn.model_selection import StratifiedShuffleSplit\\n\\nimport matplotlib.pyplot as plt\\nfrom matplotlib.pyplot import imshow\\n\\n%matplotlib inline\\n\\nimport shutil\\nimport gzip\";\n",
       "                var nbb_formatted_code = \"from xplainet.input_utils import preproc_dataset\\nfrom xplainet.model import build_model\\nfrom xplainet.random_utils import setup_seed, SEED\\nfrom sklearn.model_selection import StratifiedShuffleSplit\\n\\nimport matplotlib.pyplot as plt\\nfrom matplotlib.pyplot import imshow\\n\\n%matplotlib inline\\n\\nimport shutil\\nimport gzip\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from xplainet.input_utils import preproc_dataset\n",
    "from xplainet.model import build_model\n",
    "from xplainet.random_utils import setup_seed, SEED\n",
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.pyplot import imshow\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "import shutil\n",
    "import gzip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:26.502031Z",
     "start_time": "2020-09-19T11:51:26.480141Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 4;\n",
       "                var nbb_unformatted_code = \"setup_seed()\";\n",
       "                var nbb_formatted_code = \"setup_seed()\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "setup_seed()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:26.533255Z",
     "start_time": "2020-09-19T11:51:26.502919Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 5;\n",
       "                var nbb_unformatted_code = \"def download(url, out, force=False, verify=True):\\n    out.parent.mkdir(parents=True, exist_ok=True)\\n    if force and out.exists():\\n        print(f\\\"Removing file at {str(out)}\\\")\\n        out.unlink()\\n\\n    if out.exists():\\n        print(\\\"File already exists.\\\")\\n        return\\n    print(f\\\"Downloading {url} at {str(out)} ...\\\")\\n    # open in binary mode\\n    with out.open(mode=\\\"wb\\\") as file:\\n        # get request\\n        response = get(url, verify=verify)\\n        for chunk in response.iter_content(100000):\\n            # write to file\\n            file.write(chunk)\";\n",
       "                var nbb_formatted_code = \"def download(url, out, force=False, verify=True):\\n    out.parent.mkdir(parents=True, exist_ok=True)\\n    if force and out.exists():\\n        print(f\\\"Removing file at {str(out)}\\\")\\n        out.unlink()\\n\\n    if out.exists():\\n        print(\\\"File already exists.\\\")\\n        return\\n    print(f\\\"Downloading {url} at {str(out)} ...\\\")\\n    # open in binary mode\\n    with out.open(mode=\\\"wb\\\") as file:\\n        # get request\\n        response = get(url, verify=verify)\\n        for chunk in response.iter_content(100000):\\n            # write to file\\n            file.write(chunk)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def download(url, out, force=False, verify=True):\n",
    "    out.parent.mkdir(parents=True, exist_ok=True)\n",
    "    if force and out.exists():\n",
    "        print(f\"Removing file at {str(out)}\")\n",
    "        out.unlink()\n",
    "\n",
    "    if out.exists():\n",
    "        print(\"File already exists.\")\n",
    "        return\n",
    "    print(f\"Downloading {url} at {str(out)} ...\")\n",
    "    # open in binary mode\n",
    "    with out.open(mode=\"wb\") as file:\n",
    "        # get request\n",
    "        response = get(url, verify=verify)\n",
    "        for chunk in response.iter_content(100000):\n",
    "            # write to file\n",
    "            file.write(chunk)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:26.575840Z",
     "start_time": "2020-09-19T11:51:26.534379Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 6;\n",
       "                var nbb_unformatted_code = \"def plot_history(history):\\n    loss_list = [s for s in history.history.keys() if \\\"loss\\\" in s and \\\"val\\\" not in s]\\n    val_loss_list = [s for s in history.history.keys() if \\\"loss\\\" in s and \\\"val\\\" in s]\\n    acc_list = [s for s in history.history.keys() if \\\"AUC\\\" in s and \\\"val\\\" not in s]\\n    val_acc_list = [s for s in history.history.keys() if \\\"AUC\\\" in s and \\\"val\\\" in s]\\n\\n    if len(loss_list) == 0:\\n        print(\\\"Loss is missing in history\\\")\\n        return\\n\\n    ## As loss always exists\\n    epochs = range(1, len(history.history[loss_list[0]]) + 1)\\n\\n    ## Loss\\n    plt.figure(1)\\n    for l in loss_list:\\n        plt.plot(\\n            epochs,\\n            history.history[l],\\n            \\\"b\\\",\\n            label=\\\"Training loss (\\\"\\n            + str(str(format(history.history[l][-1], \\\".5f\\\")) + \\\")\\\"),\\n        )\\n    for l in val_loss_list:\\n        plt.plot(\\n            epochs,\\n            history.history[l],\\n            \\\"g\\\",\\n            label=\\\"Validation loss (\\\"\\n            + str(str(format(history.history[l][-1], \\\".5f\\\")) + \\\")\\\"),\\n        )\\n\\n    plt.title(\\\"Loss\\\")\\n    plt.xlabel(\\\"Epochs\\\")\\n    plt.ylabel(\\\"Loss\\\")\\n    plt.legend()\\n\\n    plt.show()\";\n",
       "                var nbb_formatted_code = \"def plot_history(history):\\n    loss_list = [s for s in history.history.keys() if \\\"loss\\\" in s and \\\"val\\\" not in s]\\n    val_loss_list = [s for s in history.history.keys() if \\\"loss\\\" in s and \\\"val\\\" in s]\\n    acc_list = [s for s in history.history.keys() if \\\"AUC\\\" in s and \\\"val\\\" not in s]\\n    val_acc_list = [s for s in history.history.keys() if \\\"AUC\\\" in s and \\\"val\\\" in s]\\n\\n    if len(loss_list) == 0:\\n        print(\\\"Loss is missing in history\\\")\\n        return\\n\\n    ## As loss always exists\\n    epochs = range(1, len(history.history[loss_list[0]]) + 1)\\n\\n    ## Loss\\n    plt.figure(1)\\n    for l in loss_list:\\n        plt.plot(\\n            epochs,\\n            history.history[l],\\n            \\\"b\\\",\\n            label=\\\"Training loss (\\\"\\n            + str(str(format(history.history[l][-1], \\\".5f\\\")) + \\\")\\\"),\\n        )\\n    for l in val_loss_list:\\n        plt.plot(\\n            epochs,\\n            history.history[l],\\n            \\\"g\\\",\\n            label=\\\"Validation loss (\\\"\\n            + str(str(format(history.history[l][-1], \\\".5f\\\")) + \\\")\\\"),\\n        )\\n\\n    plt.title(\\\"Loss\\\")\\n    plt.xlabel(\\\"Epochs\\\")\\n    plt.ylabel(\\\"Loss\\\")\\n    plt.legend()\\n\\n    plt.show()\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_history(history):\n",
    "    loss_list = [s for s in history.history.keys() if \"loss\" in s and \"val\" not in s]\n",
    "    val_loss_list = [s for s in history.history.keys() if \"loss\" in s and \"val\" in s]\n",
    "    acc_list = [s for s in history.history.keys() if \"AUC\" in s and \"val\" not in s]\n",
    "    val_acc_list = [s for s in history.history.keys() if \"AUC\" in s and \"val\" in s]\n",
    "\n",
    "    if len(loss_list) == 0:\n",
    "        print(\"Loss is missing in history\")\n",
    "        return\n",
    "\n",
    "    ## As loss always exists\n",
    "    epochs = range(1, len(history.history[loss_list[0]]) + 1)\n",
    "\n",
    "    ## Loss\n",
    "    plt.figure(1)\n",
    "    for l in loss_list:\n",
    "        plt.plot(\n",
    "            epochs,\n",
    "            history.history[l],\n",
    "            \"b\",\n",
    "            label=\"Training loss (\"\n",
    "            + str(str(format(history.history[l][-1], \".5f\")) + \")\"),\n",
    "        )\n",
    "    for l in val_loss_list:\n",
    "        plt.plot(\n",
    "            epochs,\n",
    "            history.history[l],\n",
    "            \"g\",\n",
    "            label=\"Validation loss (\"\n",
    "            + str(str(format(history.history[l][-1], \".5f\")) + \")\"),\n",
    "        )\n",
    "\n",
    "    plt.title(\"Loss\")\n",
    "    plt.xlabel(\"Epochs\")\n",
    "    plt.ylabel(\"Loss\")\n",
    "    plt.legend()\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Forest covertype : loading data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:28.403571Z",
     "start_time": "2020-09-19T11:51:26.576844Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File already exists.\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 7;\n",
       "                var nbb_unformatted_code = \"url = (\\n    \\\"https://archive.ics.uci.edu/ml/machine-learning-databases/covtype/covtype.data.gz\\\"\\n)\\ndataset_name = \\\"forest-cover-type\\\"\\nout_zip = Path(os.getcwd() + \\\"/data/forest-cover-type.zip\\\")\\nout = Path(os.getcwd() + \\\"/data/\\\" + dataset_name + \\\".csv\\\")\\n\\ndownload(url, out_zip)\\nwith gzip.open(out_zip, \\\"rb\\\") as f_in:\\n    with open(out, \\\"wb\\\") as f_out:\\n        shutil.copyfileobj(f_in, f_out)\\n\\ntarget = \\\"Covertype\\\"\\nbool_columns = [\\n    \\\"Wilderness_Area1\\\",\\n    \\\"Wilderness_Area2\\\",\\n    \\\"Wilderness_Area3\\\",\\n    \\\"Wilderness_Area4\\\",\\n    \\\"Soil_Type1\\\",\\n    \\\"Soil_Type2\\\",\\n    \\\"Soil_Type3\\\",\\n    \\\"Soil_Type4\\\",\\n    \\\"Soil_Type5\\\",\\n    \\\"Soil_Type6\\\",\\n    \\\"Soil_Type7\\\",\\n    \\\"Soil_Type8\\\",\\n    \\\"Soil_Type9\\\",\\n    \\\"Soil_Type10\\\",\\n    \\\"Soil_Type11\\\",\\n    \\\"Soil_Type12\\\",\\n    \\\"Soil_Type13\\\",\\n    \\\"Soil_Type14\\\",\\n    \\\"Soil_Type15\\\",\\n    \\\"Soil_Type16\\\",\\n    \\\"Soil_Type17\\\",\\n    \\\"Soil_Type18\\\",\\n    \\\"Soil_Type19\\\",\\n    \\\"Soil_Type20\\\",\\n    \\\"Soil_Type21\\\",\\n    \\\"Soil_Type22\\\",\\n    \\\"Soil_Type23\\\",\\n    \\\"Soil_Type24\\\",\\n    \\\"Soil_Type25\\\",\\n    \\\"Soil_Type26\\\",\\n    \\\"Soil_Type27\\\",\\n    \\\"Soil_Type28\\\",\\n    \\\"Soil_Type29\\\",\\n    \\\"Soil_Type30\\\",\\n    \\\"Soil_Type31\\\",\\n    \\\"Soil_Type32\\\",\\n    \\\"Soil_Type33\\\",\\n    \\\"Soil_Type34\\\",\\n    \\\"Soil_Type35\\\",\\n    \\\"Soil_Type36\\\",\\n    \\\"Soil_Type37\\\",\\n    \\\"Soil_Type38\\\",\\n    \\\"Soil_Type39\\\",\\n    \\\"Soil_Type40\\\",\\n]\\n\\nint_columns = [\\n    \\\"Elevation\\\",\\n    \\\"Aspect\\\",\\n    \\\"Slope\\\",\\n    \\\"Horizontal_Distance_To_Hydrology\\\",\\n    \\\"Vertical_Distance_To_Hydrology\\\",\\n    \\\"Horizontal_Distance_To_Roadways\\\",\\n    \\\"Hillshade_9am\\\",\\n    \\\"Hillshade_Noon\\\",\\n    \\\"Hillshade_3pm\\\",\\n    \\\"Horizontal_Distance_To_Fire_Points\\\",\\n]\\n\\nfeature_columns = int_columns + bool_columns + [target]\\nto_remove = []\\ntrain = pd.read_csv(out, sep=\\\",\\\", low_memory=False, header=None, names=feature_columns)\";\n",
       "                var nbb_formatted_code = \"url = (\\n    \\\"https://archive.ics.uci.edu/ml/machine-learning-databases/covtype/covtype.data.gz\\\"\\n)\\ndataset_name = \\\"forest-cover-type\\\"\\nout_zip = Path(os.getcwd() + \\\"/data/forest-cover-type.zip\\\")\\nout = Path(os.getcwd() + \\\"/data/\\\" + dataset_name + \\\".csv\\\")\\n\\ndownload(url, out_zip)\\nwith gzip.open(out_zip, \\\"rb\\\") as f_in:\\n    with open(out, \\\"wb\\\") as f_out:\\n        shutil.copyfileobj(f_in, f_out)\\n\\ntarget = \\\"Covertype\\\"\\nbool_columns = [\\n    \\\"Wilderness_Area1\\\",\\n    \\\"Wilderness_Area2\\\",\\n    \\\"Wilderness_Area3\\\",\\n    \\\"Wilderness_Area4\\\",\\n    \\\"Soil_Type1\\\",\\n    \\\"Soil_Type2\\\",\\n    \\\"Soil_Type3\\\",\\n    \\\"Soil_Type4\\\",\\n    \\\"Soil_Type5\\\",\\n    \\\"Soil_Type6\\\",\\n    \\\"Soil_Type7\\\",\\n    \\\"Soil_Type8\\\",\\n    \\\"Soil_Type9\\\",\\n    \\\"Soil_Type10\\\",\\n    \\\"Soil_Type11\\\",\\n    \\\"Soil_Type12\\\",\\n    \\\"Soil_Type13\\\",\\n    \\\"Soil_Type14\\\",\\n    \\\"Soil_Type15\\\",\\n    \\\"Soil_Type16\\\",\\n    \\\"Soil_Type17\\\",\\n    \\\"Soil_Type18\\\",\\n    \\\"Soil_Type19\\\",\\n    \\\"Soil_Type20\\\",\\n    \\\"Soil_Type21\\\",\\n    \\\"Soil_Type22\\\",\\n    \\\"Soil_Type23\\\",\\n    \\\"Soil_Type24\\\",\\n    \\\"Soil_Type25\\\",\\n    \\\"Soil_Type26\\\",\\n    \\\"Soil_Type27\\\",\\n    \\\"Soil_Type28\\\",\\n    \\\"Soil_Type29\\\",\\n    \\\"Soil_Type30\\\",\\n    \\\"Soil_Type31\\\",\\n    \\\"Soil_Type32\\\",\\n    \\\"Soil_Type33\\\",\\n    \\\"Soil_Type34\\\",\\n    \\\"Soil_Type35\\\",\\n    \\\"Soil_Type36\\\",\\n    \\\"Soil_Type37\\\",\\n    \\\"Soil_Type38\\\",\\n    \\\"Soil_Type39\\\",\\n    \\\"Soil_Type40\\\",\\n]\\n\\nint_columns = [\\n    \\\"Elevation\\\",\\n    \\\"Aspect\\\",\\n    \\\"Slope\\\",\\n    \\\"Horizontal_Distance_To_Hydrology\\\",\\n    \\\"Vertical_Distance_To_Hydrology\\\",\\n    \\\"Horizontal_Distance_To_Roadways\\\",\\n    \\\"Hillshade_9am\\\",\\n    \\\"Hillshade_Noon\\\",\\n    \\\"Hillshade_3pm\\\",\\n    \\\"Horizontal_Distance_To_Fire_Points\\\",\\n]\\n\\nfeature_columns = int_columns + bool_columns + [target]\\nto_remove = []\\ntrain = pd.read_csv(out, sep=\\\",\\\", low_memory=False, header=None, names=feature_columns)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "url = (\n",
    "    \"https://archive.ics.uci.edu/ml/machine-learning-databases/covtype/covtype.data.gz\"\n",
    ")\n",
    "dataset_name = \"forest-cover-type\"\n",
    "out_zip = Path(os.getcwd() + \"/data/forest-cover-type.zip\")\n",
    "out = Path(os.getcwd() + \"/data/\" + dataset_name + \".csv\")\n",
    "\n",
    "download(url, out_zip)\n",
    "with gzip.open(out_zip, \"rb\") as f_in:\n",
    "    with open(out, \"wb\") as f_out:\n",
    "        shutil.copyfileobj(f_in, f_out)\n",
    "\n",
    "target = \"Covertype\"\n",
    "bool_columns = [\n",
    "    \"Wilderness_Area1\",\n",
    "    \"Wilderness_Area2\",\n",
    "    \"Wilderness_Area3\",\n",
    "    \"Wilderness_Area4\",\n",
    "    \"Soil_Type1\",\n",
    "    \"Soil_Type2\",\n",
    "    \"Soil_Type3\",\n",
    "    \"Soil_Type4\",\n",
    "    \"Soil_Type5\",\n",
    "    \"Soil_Type6\",\n",
    "    \"Soil_Type7\",\n",
    "    \"Soil_Type8\",\n",
    "    \"Soil_Type9\",\n",
    "    \"Soil_Type10\",\n",
    "    \"Soil_Type11\",\n",
    "    \"Soil_Type12\",\n",
    "    \"Soil_Type13\",\n",
    "    \"Soil_Type14\",\n",
    "    \"Soil_Type15\",\n",
    "    \"Soil_Type16\",\n",
    "    \"Soil_Type17\",\n",
    "    \"Soil_Type18\",\n",
    "    \"Soil_Type19\",\n",
    "    \"Soil_Type20\",\n",
    "    \"Soil_Type21\",\n",
    "    \"Soil_Type22\",\n",
    "    \"Soil_Type23\",\n",
    "    \"Soil_Type24\",\n",
    "    \"Soil_Type25\",\n",
    "    \"Soil_Type26\",\n",
    "    \"Soil_Type27\",\n",
    "    \"Soil_Type28\",\n",
    "    \"Soil_Type29\",\n",
    "    \"Soil_Type30\",\n",
    "    \"Soil_Type31\",\n",
    "    \"Soil_Type32\",\n",
    "    \"Soil_Type33\",\n",
    "    \"Soil_Type34\",\n",
    "    \"Soil_Type35\",\n",
    "    \"Soil_Type36\",\n",
    "    \"Soil_Type37\",\n",
    "    \"Soil_Type38\",\n",
    "    \"Soil_Type39\",\n",
    "    \"Soil_Type40\",\n",
    "]\n",
    "\n",
    "int_columns = [\n",
    "    \"Elevation\",\n",
    "    \"Aspect\",\n",
    "    \"Slope\",\n",
    "    \"Horizontal_Distance_To_Hydrology\",\n",
    "    \"Vertical_Distance_To_Hydrology\",\n",
    "    \"Horizontal_Distance_To_Roadways\",\n",
    "    \"Hillshade_9am\",\n",
    "    \"Hillshade_Noon\",\n",
    "    \"Hillshade_3pm\",\n",
    "    \"Horizontal_Distance_To_Fire_Points\",\n",
    "]\n",
    "\n",
    "feature_columns = int_columns + bool_columns + [target]\n",
    "to_remove = []\n",
    "train = pd.read_csv(out, sep=\",\", low_memory=False, header=None, names=feature_columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:28.429159Z",
     "start_time": "2020-09-19T11:51:28.405316Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(581012, 55)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 8;\n",
       "                var nbb_unformatted_code = \"train.shape\";\n",
       "                var nbb_formatted_code = \"train.shape\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:28.455105Z",
     "start_time": "2020-09-19T11:51:28.430707Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Elevation', 'Aspect', 'Slope', 'Horizontal_Distance_To_Hydrology',\n",
       "       'Vertical_Distance_To_Hydrology', 'Horizontal_Distance_To_Roadways',\n",
       "       'Hillshade_9am', 'Hillshade_Noon', 'Hillshade_3pm',\n",
       "       'Horizontal_Distance_To_Fire_Points', 'Wilderness_Area1',\n",
       "       'Wilderness_Area2', 'Wilderness_Area3', 'Wilderness_Area4',\n",
       "       'Soil_Type1', 'Soil_Type2', 'Soil_Type3', 'Soil_Type4', 'Soil_Type5',\n",
       "       'Soil_Type6', 'Soil_Type7', 'Soil_Type8', 'Soil_Type9', 'Soil_Type10',\n",
       "       'Soil_Type11', 'Soil_Type12', 'Soil_Type13', 'Soil_Type14',\n",
       "       'Soil_Type15', 'Soil_Type16', 'Soil_Type17', 'Soil_Type18',\n",
       "       'Soil_Type19', 'Soil_Type20', 'Soil_Type21', 'Soil_Type22',\n",
       "       'Soil_Type23', 'Soil_Type24', 'Soil_Type25', 'Soil_Type26',\n",
       "       'Soil_Type27', 'Soil_Type28', 'Soil_Type29', 'Soil_Type30',\n",
       "       'Soil_Type31', 'Soil_Type32', 'Soil_Type33', 'Soil_Type34',\n",
       "       'Soil_Type35', 'Soil_Type36', 'Soil_Type37', 'Soil_Type38',\n",
       "       'Soil_Type39', 'Soil_Type40', 'Covertype'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 9;\n",
       "                var nbb_unformatted_code = \"train.columns\";\n",
       "                var nbb_formatted_code = \"train.columns\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:28.786530Z",
     "start_time": "2020-09-19T11:51:28.456068Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Building tailored column\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/ipykernel_launcher.py:14: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \n",
      "/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/ipykernel_launcher.py:15: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  from ipykernel import kernelapp as app\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 10;\n",
       "                var nbb_unformatted_code = \"if \\\"Set\\\" not in train.columns:\\n    print(\\\"Building tailored column\\\")\\n    train_valid_index, test_index = next(\\n        StratifiedShuffleSplit(n_splits=1, test_size=0.1, random_state=SEED).split(\\n            range(train[target].shape[0]), train[target].values\\n        )\\n    )\\n    train_index, valid_index = next(\\n        StratifiedShuffleSplit(n_splits=1, test_size=0.1, random_state=SEED).split(\\n            train_valid_index, train[target].values[train_valid_index]\\n        )\\n    )\\n    train[\\\"Set\\\"] = \\\"train\\\"\\n    train[\\\"Set\\\"][valid_index] = \\\"valid\\\"\\n    train[\\\"Set\\\"][test_index] = \\\"test\\\"\\n    # train.to_csv((out.parent / \\\"train_bench.csv\\\").as_posix(), index=False)\";\n",
       "                var nbb_formatted_code = \"if \\\"Set\\\" not in train.columns:\\n    print(\\\"Building tailored column\\\")\\n    train_valid_index, test_index = next(\\n        StratifiedShuffleSplit(n_splits=1, test_size=0.1, random_state=SEED).split(\\n            range(train[target].shape[0]), train[target].values\\n        )\\n    )\\n    train_index, valid_index = next(\\n        StratifiedShuffleSplit(n_splits=1, test_size=0.1, random_state=SEED).split(\\n            train_valid_index, train[target].values[train_valid_index]\\n        )\\n    )\\n    train[\\\"Set\\\"] = \\\"train\\\"\\n    train[\\\"Set\\\"][valid_index] = \\\"valid\\\"\\n    train[\\\"Set\\\"][test_index] = \\\"test\\\"\\n    # train.to_csv((out.parent / \\\"train_bench.csv\\\").as_posix(), index=False)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if \"Set\" not in train.columns:\n",
    "    print(\"Building tailored column\")\n",
    "    train_valid_index, test_index = next(\n",
    "        StratifiedShuffleSplit(n_splits=1, test_size=0.1, random_state=SEED).split(\n",
    "            range(train[target].shape[0]), train[target].values\n",
    "        )\n",
    "    )\n",
    "    train_index, valid_index = next(\n",
    "        StratifiedShuffleSplit(n_splits=1, test_size=0.1, random_state=SEED).split(\n",
    "            train_valid_index, train[target].values[train_valid_index]\n",
    "        )\n",
    "    )\n",
    "    train[\"Set\"] = \"train\"\n",
    "    train[\"Set\"][valid_index] = \"valid\"\n",
    "    train[\"Set\"][test_index] = \"test\"\n",
    "    # train.to_csv((out.parent / \"train_bench.csv\").as_posix(), index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:51:28.954932Z",
     "start_time": "2020-09-19T11:51:28.787616Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 11;\n",
       "                var nbb_unformatted_code = \"train_indices = train[train.Set == \\\"train\\\"].index\\nvalid_indices = train[train.Set == \\\"valid\\\"].index\\ntest_indices = train[train.Set == \\\"test\\\"].index\";\n",
       "                var nbb_formatted_code = \"train_indices = train[train.Set == \\\"train\\\"].index\\nvalid_indices = train[train.Set == \\\"valid\\\"].index\\ntest_indices = train[train.Set == \\\"test\\\"].index\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_indices = train[train.Set == \"train\"].index\n",
    "valid_indices = train[train.Set == \"valid\"].index\n",
    "test_indices = train[train.Set == \"test\"].index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:09.621678Z",
     "start_time": "2020-09-19T11:51:28.955904Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'constant_cols': [],\n",
       " 'bool_cols': ['Soil_Type1',\n",
       "  'Soil_Type40',\n",
       "  'Soil_Type12',\n",
       "  'Soil_Type31',\n",
       "  'Soil_Type20',\n",
       "  'Soil_Type16',\n",
       "  'Soil_Type13',\n",
       "  'Soil_Type25',\n",
       "  'Soil_Type15',\n",
       "  'Wilderness_Area1',\n",
       "  'Soil_Type22',\n",
       "  'Soil_Type26',\n",
       "  'Soil_Type10',\n",
       "  'Soil_Type4',\n",
       "  'Soil_Type36',\n",
       "  'Soil_Type14',\n",
       "  'Soil_Type2',\n",
       "  'Soil_Type27',\n",
       "  'Soil_Type33',\n",
       "  'Soil_Type30',\n",
       "  'Soil_Type35',\n",
       "  'Soil_Type24',\n",
       "  'Soil_Type11',\n",
       "  'Soil_Type6',\n",
       "  'Soil_Type18',\n",
       "  'Soil_Type7',\n",
       "  'Soil_Type17',\n",
       "  'Soil_Type3',\n",
       "  'Wilderness_Area4',\n",
       "  'Wilderness_Area2',\n",
       "  'Soil_Type8',\n",
       "  'Soil_Type39',\n",
       "  'Soil_Type37',\n",
       "  'Soil_Type23',\n",
       "  'Soil_Type38',\n",
       "  'Soil_Type29',\n",
       "  'Wilderness_Area3',\n",
       "  'Soil_Type34',\n",
       "  'Soil_Type28',\n",
       "  'Soil_Type32',\n",
       "  'Soil_Type19',\n",
       "  'Soil_Type21',\n",
       "  'Soil_Type5',\n",
       "  'Soil_Type9'],\n",
       " 'num_cols': [],\n",
       " 'cat_cols': ['Hillshade_9am',\n",
       "  'Horizontal_Distance_To_Roadways',\n",
       "  'Aspect',\n",
       "  'Horizontal_Distance_To_Fire_Points',\n",
       "  'Slope',\n",
       "  'Hillshade_3pm',\n",
       "  'Horizontal_Distance_To_Hydrology',\n",
       "  'Hillshade_Noon',\n",
       "  'Vertical_Distance_To_Hydrology',\n",
       "  'Elevation'],\n",
       " 'num_encoder': [],\n",
       " 'bool_encoder': [SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder()],\n",
       " 'max_nb': 5793,\n",
       " 'cat_encoder': [SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder()]}"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 12;\n",
       "                var nbb_unformatted_code = \"input_train, params = preproc_dataset(train.loc[train_indices], target, [\\\"Set\\\"])\\nparams\";\n",
       "                var nbb_formatted_code = \"input_train, params = preproc_dataset(train.loc[train_indices], target, [\\\"Set\\\"])\\nparams\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "input_train, params = preproc_dataset(train.loc[train_indices], target, [\"Set\"])\n",
    "params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.405467Z",
     "start_time": "2020-09-19T11:52:09.622588Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 13;\n",
       "                var nbb_unformatted_code = \"input_valid, _ = preproc_dataset(train.loc[valid_indices], target, [\\\"Set\\\"], params)\\ninput_test, _ = preproc_dataset(train.loc[test_indices], target, [\\\"Set\\\"], params)\";\n",
       "                var nbb_formatted_code = \"input_valid, _ = preproc_dataset(train.loc[valid_indices], target, [\\\"Set\\\"], params)\\ninput_test, _ = preproc_dataset(train.loc[test_indices], target, [\\\"Set\\\"], params)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "input_valid, _ = preproc_dataset(train.loc[valid_indices], target, [\"Set\"], params)\n",
    "input_test, _ = preproc_dataset(train.loc[test_indices], target, [\"Set\"], params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.428045Z",
     "start_time": "2020-09-19T11:52:14.406692Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 14;\n",
       "                var nbb_unformatted_code = \"target_encoder = LabelEncoder()\";\n",
       "                var nbb_formatted_code = \"target_encoder = LabelEncoder()\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "target_encoder = LabelEncoder()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.466979Z",
     "start_time": "2020-09-19T11:52:14.428939Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 15;\n",
       "                var nbb_unformatted_code = \"train[target] = target_encoder.fit_transform(train[target].values.reshape(-1))\\ny_train = train[target].values[train_indices]\\ny_valid = train[target].values[valid_indices]\\ny_test = train[target].values[test_indices]\";\n",
       "                var nbb_formatted_code = \"train[target] = target_encoder.fit_transform(train[target].values.reshape(-1))\\ny_train = train[target].values[train_indices]\\ny_valid = train[target].values[valid_indices]\\ny_test = train[target].values[test_indices]\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train[target] = target_encoder.fit_transform(train[target].values.reshape(-1))\n",
    "y_train = train[target].values[train_indices]\n",
    "y_valid = train[target].values[valid_indices]\n",
    "y_test = train[target].values[test_indices]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.501553Z",
     "start_time": "2020-09-19T11:52:14.467843Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'constant_cols': [],\n",
       " 'bool_cols': ['Soil_Type1',\n",
       "  'Soil_Type40',\n",
       "  'Soil_Type12',\n",
       "  'Soil_Type31',\n",
       "  'Soil_Type20',\n",
       "  'Soil_Type16',\n",
       "  'Soil_Type13',\n",
       "  'Soil_Type25',\n",
       "  'Soil_Type15',\n",
       "  'Wilderness_Area1',\n",
       "  'Soil_Type22',\n",
       "  'Soil_Type26',\n",
       "  'Soil_Type10',\n",
       "  'Soil_Type4',\n",
       "  'Soil_Type36',\n",
       "  'Soil_Type14',\n",
       "  'Soil_Type2',\n",
       "  'Soil_Type27',\n",
       "  'Soil_Type33',\n",
       "  'Soil_Type30',\n",
       "  'Soil_Type35',\n",
       "  'Soil_Type24',\n",
       "  'Soil_Type11',\n",
       "  'Soil_Type6',\n",
       "  'Soil_Type18',\n",
       "  'Soil_Type7',\n",
       "  'Soil_Type17',\n",
       "  'Soil_Type3',\n",
       "  'Wilderness_Area4',\n",
       "  'Wilderness_Area2',\n",
       "  'Soil_Type8',\n",
       "  'Soil_Type39',\n",
       "  'Soil_Type37',\n",
       "  'Soil_Type23',\n",
       "  'Soil_Type38',\n",
       "  'Soil_Type29',\n",
       "  'Wilderness_Area3',\n",
       "  'Soil_Type34',\n",
       "  'Soil_Type28',\n",
       "  'Soil_Type32',\n",
       "  'Soil_Type19',\n",
       "  'Soil_Type21',\n",
       "  'Soil_Type5',\n",
       "  'Soil_Type9'],\n",
       " 'num_cols': [],\n",
       " 'cat_cols': ['Hillshade_9am',\n",
       "  'Horizontal_Distance_To_Roadways',\n",
       "  'Aspect',\n",
       "  'Horizontal_Distance_To_Fire_Points',\n",
       "  'Slope',\n",
       "  'Hillshade_3pm',\n",
       "  'Horizontal_Distance_To_Hydrology',\n",
       "  'Hillshade_Noon',\n",
       "  'Vertical_Distance_To_Hydrology',\n",
       "  'Elevation'],\n",
       " 'num_encoder': [],\n",
       " 'bool_encoder': [SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder()],\n",
       " 'max_nb': 5793,\n",
       " 'cat_encoder': [SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder(),\n",
       "  SafeLabelEncoder()]}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 16;\n",
       "                var nbb_unformatted_code = \"params\";\n",
       "                var nbb_formatted_code = \"params\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.622332Z",
     "start_time": "2020-09-19T11:52:14.502650Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow_addons/options.py:47: RuntimeWarning: Traceback (most recent call last):\n",
      "  File \"/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow_addons/activations/mish.py\", line 44, in mish\n",
      "    return _mish_custom_op(x)\n",
      "  File \"/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow_addons/activations/mish.py\", line 59, in _mish_custom_op\n",
      "    return _activation_so.ops.addons_mish(x)\n",
      "  File \"/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow_addons/utils/resource_loader.py\", line 64, in ops\n",
      "    self._ops = tf.load_op_library(get_path_to_datafile(self.relative_path))\n",
      "  File \"/work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow/python/framework/load_library.py\", line 58, in load_op_library\n",
      "    lib_handle = py_tf.TF_LoadLibrary(library_filename)\n",
      "tensorflow.python.framework.errors_impl.NotFoundError: /work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow_addons/custom_ops/activations/_activation_ops.so: undefined symbol: _ZN10tensorflow20OpKernelConstruction14MatchSignatureEN4absl14lts_2020_02_254SpanIKNS_8DataTypeEEES6_\n",
      "\n",
      "\n",
      "The mish C++/CUDA custom op could not be loaded.\n",
      "For this reason, Addons will fallback to an implementation written\n",
      "in Python with public TensorFlow ops. There worst you might experience with\n",
      "this is a moderate slowdown on GPU. There can be multiple\n",
      "reason for this loading error, one of them may be an ABI incompatibility between\n",
      "the TensorFlow installed on your system and the TensorFlow used to compile\n",
      "TensorFlow Addons' custom ops. The stacktrace generated when loading the\n",
      "shared object file was displayed above.\n",
      "\n",
      "If you want this warning to disappear, either make sure the TensorFlow installed\n",
      "is compatible with this version of Addons, or tell TensorFlow Addons to\n",
      "prefer using Python implementations and not custom C++/CUDA ones. You can do that\n",
      "by changing the TF_ADDONS_PY_OPS flag\n",
      "either with the environment variable:\n",
      "```bash\n",
      "TF_ADDONS_PY_OPS=1 python my_script.py\n",
      "```\n",
      "or in your code, after your imports:\n",
      "```python\n",
      "import tensorflow_addons as tfa\n",
      "import ...\n",
      "import ...\n",
      "\n",
      "tfa.options.TF_ADDONS_PY_OPS = True\n",
      "```\n",
      "\n",
      "  warnings.warn(warning_msg, RuntimeWarning)\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 17;\n",
       "                var nbb_unformatted_code = \"model = build_model(\\n    params,\\n    lconv_dim=[4],\\n    lconv_num_dim=[8],\\n    emb_size=16,\\n    activation_num_first_layer=None,  # \\\"tanh\\\",\\n    output_activation=\\\"softmax\\\",\\n    output_dim=np.unique(y_train).shape[0],\\n)\";\n",
       "                var nbb_formatted_code = \"model = build_model(\\n    params,\\n    lconv_dim=[4],\\n    lconv_num_dim=[8],\\n    emb_size=16,\\n    activation_num_first_layer=None,  # \\\"tanh\\\",\\n    output_activation=\\\"softmax\\\",\\n    output_dim=np.unique(y_train).shape[0],\\n)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = build_model(\n",
    "    params,\n",
    "    lconv_dim=[4],\n",
    "    lconv_num_dim=[8],\n",
    "    emb_size=16,\n",
    "    activation_num_first_layer=None,  # \"tanh\",\n",
    "    output_activation=\"softmax\",\n",
    "    output_dim=np.unique(y_train).shape[0],\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.646069Z",
     "start_time": "2020-09-19T11:52:14.623308Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TensorShape([None, 84])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 18;\n",
       "                var nbb_unformatted_code = \"model.get_layer(\\\"output\\\")._build_input_shape\";\n",
       "                var nbb_formatted_code = \"model.get_layer(\\\"output\\\")._build_input_shape\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model.get_layer(\"output\")._build_input_shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.673862Z",
     "start_time": "2020-09-19T11:52:14.647011Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"explainable_model\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_cat (InputLayer)          [(None, 10)]         0                                            \n",
      "__________________________________________________________________________________________________\n",
      "large_emb (Embedding)           (None, 10, 16)       92704       input_cat[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "block_cat_0_conv (LocallyConnec (None, 10, 4)        640         large_emb[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "block_cat_0_activation (Activat (None, 10, 4)        0           block_cat_0_conv[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "input_bool (InputLayer)         [(None, 44)]         0                                            \n",
      "__________________________________________________________________________________________________\n",
      "reshape_cat_output (Reshape)    (None, 40)           0           block_cat_0_activation[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "concatenate (Concatenate)       (None, 84)           0           input_bool[0][0]                 \n",
      "                                                                 reshape_cat_output[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "output (Dense)                  (None, 7)            595         concatenate[0][0]                \n",
      "==================================================================================================\n",
      "Total params: 93,939\n",
      "Trainable params: 93,939\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 19;\n",
       "                var nbb_unformatted_code = \"model.summary()\";\n",
       "                var nbb_formatted_code = \"model.summary()\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.696306Z",
     "start_time": "2020-09-19T11:52:14.675121Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 20;\n",
       "                var nbb_unformatted_code = \"#!pip install pydot graphviz\";\n",
       "                var nbb_formatted_code = \"#!pip install pydot graphviz\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#!pip install pydot graphviz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.855994Z",
     "start_time": "2020-09-19T11:52:14.697325Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 21;\n",
       "                var nbb_unformatted_code = \"plot_model(\\n    model,\\n    # to_file=\\\"model.png\\\",\\n    show_shapes=True,\\n    show_layer_names=True,\\n    rankdir=\\\"TB\\\",\\n    expand_nested=False,\\n    dpi=96,\\n)\";\n",
       "                var nbb_formatted_code = \"plot_model(\\n    model,\\n    # to_file=\\\"model.png\\\",\\n    show_shapes=True,\\n    show_layer_names=True,\\n    rankdir=\\\"TB\\\",\\n    expand_nested=False,\\n    dpi=96,\\n)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_model(\n",
    "    model,\n",
    "    # to_file=\"model.png\",\n",
    "    show_shapes=True,\n",
    "    show_layer_names=True,\n",
    "    rankdir=\"TB\",\n",
    "    expand_nested=False,\n",
    "    dpi=96,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.885214Z",
     "start_time": "2020-09-19T11:52:14.856992Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(475918,)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 22;\n",
       "                var nbb_unformatted_code = \"y_train.shape\";\n",
       "                var nbb_formatted_code = \"y_train.shape\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.916457Z",
     "start_time": "2020-09-19T11:52:14.887658Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 23;\n",
       "                var nbb_unformatted_code = \"counts = np.unique(y_train, return_counts=True)[1]\\ncounts = counts.sum() / counts\";\n",
       "                var nbb_formatted_code = \"counts = np.unique(y_train, return_counts=True)[1]\\ncounts = counts.sum() / counts\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "counts = np.unique(y_train, return_counts=True)[1]\n",
    "counts = counts.sum() / counts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:52:14.942136Z",
     "start_time": "2020-09-19T11:52:14.917821Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 2.741573680966398,\n",
       " 1: 2.0547184637037934,\n",
       " 2: 16.045785569790965,\n",
       " 3: 213.41614349775784,\n",
       " 4: 62.07356201904265,\n",
       " 5: 33.20667038794306,\n",
       " 6: 28.304864993457834}"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 24;\n",
       "                var nbb_unformatted_code = \"class_weight = {}\\nfor i, count in enumerate(counts):\\n    class_weight[i] = count\\nclass_weight\";\n",
       "                var nbb_formatted_code = \"class_weight = {}\\nfor i, count in enumerate(counts):\\n    class_weight[i] = count\\nclass_weight\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "class_weight = {}\n",
    "for i, count in enumerate(counts):\n",
    "    class_weight[i] = count\n",
    "class_weight"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:48:26.264023Z",
     "start_time": "2020-09-19T11:48:26.236235Z"
    }
   },
   "source": [
    "class_weight = {}\n",
    "for i, count in enumerate(counts):\n",
    "    class_weight[i] = 1\n",
    "class_weight"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:20.505480Z",
     "start_time": "2020-09-19T11:52:14.943115Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/2000\n",
      "WARNING:tensorflow:From /work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow/python/ops/resource_variable_ops.py:1817: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "If using Keras pass *_constraint arguments to layers.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /work/.cache/poetry/xplainet-KS9Ipbc1-py3.7/lib/python3.7/site-packages/tensorflow/python/ops/resource_variable_ops.py:1817: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "If using Keras pass *_constraint arguments to layers.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "465/465 - 2s - loss: 13.3051 - val_loss: 1.8080\n",
      "Epoch 2/2000\n",
      "465/465 - 2s - loss: 11.3008 - val_loss: 1.3217\n",
      "Epoch 3/2000\n",
      "465/465 - 2s - loss: 8.4089 - val_loss: 1.1830\n",
      "Epoch 4/2000\n",
      "465/465 - 2s - loss: 7.1590 - val_loss: 1.1203\n",
      "Epoch 5/2000\n",
      "465/465 - 2s - loss: 6.5057 - val_loss: 1.0665\n",
      "Epoch 6/2000\n",
      "465/465 - 2s - loss: 6.0290 - val_loss: 1.0261\n",
      "Epoch 7/2000\n",
      "465/465 - 2s - loss: 5.6266 - val_loss: 0.9850\n",
      "Epoch 8/2000\n",
      "465/465 - 2s - loss: 5.2952 - val_loss: 0.9501\n",
      "Epoch 9/2000\n",
      "465/465 - 2s - loss: 5.0494 - val_loss: 0.9268\n",
      "Epoch 10/2000\n",
      "465/465 - 2s - loss: 4.8690 - val_loss: 0.9139\n",
      "Epoch 11/2000\n",
      "465/465 - 2s - loss: 4.7298 - val_loss: 0.8973\n",
      "Epoch 12/2000\n",
      "465/465 - 2s - loss: 4.6176 - val_loss: 0.8877\n",
      "Epoch 13/2000\n",
      "465/465 - 2s - loss: 4.5234 - val_loss: 0.8833\n",
      "Epoch 14/2000\n",
      "465/465 - 2s - loss: 4.4425 - val_loss: 0.8769\n",
      "Epoch 15/2000\n",
      "465/465 - 2s - loss: 4.3736 - val_loss: 0.8712\n",
      "Epoch 16/2000\n",
      "465/465 - 2s - loss: 4.3145 - val_loss: 0.8645\n",
      "Epoch 17/2000\n",
      "465/465 - 2s - loss: 4.2633 - val_loss: 0.8553\n",
      "Epoch 18/2000\n",
      "465/465 - 2s - loss: 4.2178 - val_loss: 0.8535\n",
      "Epoch 19/2000\n",
      "465/465 - 2s - loss: 4.1775 - val_loss: 0.8533\n",
      "Epoch 20/2000\n",
      "465/465 - 2s - loss: 4.1400 - val_loss: 0.8465\n",
      "Epoch 21/2000\n",
      "465/465 - 2s - loss: 4.1061 - val_loss: 0.8484\n",
      "Epoch 22/2000\n",
      "465/465 - 2s - loss: 4.0755 - val_loss: 0.8474\n",
      "Epoch 23/2000\n",
      "465/465 - 2s - loss: 4.0472 - val_loss: 0.8492\n",
      "Epoch 24/2000\n",
      "465/465 - 2s - loss: 4.0208 - val_loss: 0.8346\n",
      "Epoch 25/2000\n",
      "465/465 - 2s - loss: 3.9969 - val_loss: 0.8418\n",
      "Epoch 26/2000\n",
      "465/465 - 2s - loss: 3.9726 - val_loss: 0.8359\n",
      "Epoch 27/2000\n",
      "465/465 - 2s - loss: 3.9502 - val_loss: 0.8444\n",
      "Epoch 28/2000\n",
      "465/465 - 2s - loss: 3.9289 - val_loss: 0.8396\n",
      "Epoch 29/2000\n",
      "465/465 - 2s - loss: 3.9092 - val_loss: 0.8343\n",
      "Epoch 30/2000\n",
      "465/465 - 2s - loss: 3.8896 - val_loss: 0.8298\n",
      "Epoch 31/2000\n",
      "465/465 - 2s - loss: 3.8709 - val_loss: 0.8336\n",
      "Epoch 32/2000\n",
      "465/465 - 2s - loss: 3.8528 - val_loss: 0.8354\n",
      "Epoch 33/2000\n",
      "465/465 - 2s - loss: 3.8357 - val_loss: 0.8303\n",
      "Epoch 34/2000\n",
      "465/465 - 2s - loss: 3.8195 - val_loss: 0.8276\n",
      "Epoch 35/2000\n",
      "465/465 - 2s - loss: 3.8032 - val_loss: 0.8296\n",
      "Epoch 36/2000\n",
      "465/465 - 2s - loss: 3.7880 - val_loss: 0.8258\n",
      "Epoch 37/2000\n",
      "465/465 - 2s - loss: 3.7725 - val_loss: 0.8315\n",
      "Epoch 38/2000\n",
      "465/465 - 2s - loss: 3.7589 - val_loss: 0.8301\n",
      "Epoch 39/2000\n",
      "465/465 - 2s - loss: 3.7453 - val_loss: 0.8362\n",
      "Epoch 40/2000\n",
      "465/465 - 2s - loss: 3.7313 - val_loss: 0.8219\n",
      "Epoch 41/2000\n",
      "465/465 - 2s - loss: 3.7173 - val_loss: 0.8293\n",
      "Epoch 42/2000\n",
      "465/465 - 2s - loss: 3.7043 - val_loss: 0.8250\n",
      "Epoch 43/2000\n",
      "465/465 - 2s - loss: 3.6936 - val_loss: 0.8254\n",
      "Epoch 44/2000\n",
      "465/465 - 2s - loss: 3.6802 - val_loss: 0.8341\n",
      "Epoch 45/2000\n",
      "465/465 - 2s - loss: 3.6692 - val_loss: 0.8317\n",
      "Epoch 46/2000\n",
      "465/465 - 2s - loss: 3.6579 - val_loss: 0.8383\n",
      "Epoch 47/2000\n",
      "465/465 - 2s - loss: 3.6464 - val_loss: 0.8288\n",
      "Epoch 48/2000\n",
      "465/465 - 2s - loss: 3.6363 - val_loss: 0.8294\n",
      "Epoch 49/2000\n",
      "465/465 - 2s - loss: 3.6258 - val_loss: 0.8244\n",
      "Epoch 50/2000\n",
      "465/465 - 2s - loss: 3.6160 - val_loss: 0.8188\n",
      "Epoch 51/2000\n",
      "465/465 - 2s - loss: 3.6066 - val_loss: 0.8259\n",
      "Epoch 52/2000\n",
      "465/465 - 2s - loss: 3.5976 - val_loss: 0.8255\n",
      "Epoch 53/2000\n",
      "465/465 - 2s - loss: 3.5885 - val_loss: 0.8262\n",
      "Epoch 54/2000\n",
      "465/465 - 2s - loss: 3.5794 - val_loss: 0.8202\n",
      "Epoch 55/2000\n",
      "465/465 - 2s - loss: 3.5713 - val_loss: 0.8179\n",
      "Epoch 56/2000\n",
      "465/465 - 2s - loss: 3.5624 - val_loss: 0.8226\n",
      "Epoch 57/2000\n",
      "465/465 - 2s - loss: 3.5543 - val_loss: 0.8207\n",
      "Epoch 58/2000\n",
      "465/465 - 2s - loss: 3.5467 - val_loss: 0.8250\n",
      "Epoch 59/2000\n",
      "465/465 - 2s - loss: 3.5397 - val_loss: 0.8298\n",
      "Epoch 60/2000\n",
      "465/465 - 2s - loss: 3.5320 - val_loss: 0.8183\n",
      "Epoch 61/2000\n",
      "465/465 - 2s - loss: 3.5245 - val_loss: 0.8245\n",
      "Epoch 62/2000\n",
      "465/465 - 2s - loss: 3.5180 - val_loss: 0.8234\n",
      "Epoch 63/2000\n",
      "465/465 - 2s - loss: 3.5120 - val_loss: 0.8292\n",
      "Epoch 64/2000\n",
      "465/465 - 2s - loss: 3.5045 - val_loss: 0.8201\n",
      "Epoch 65/2000\n",
      "465/465 - 2s - loss: 3.4993 - val_loss: 0.8250\n",
      "Epoch 66/2000\n",
      "465/465 - 2s - loss: 3.4926 - val_loss: 0.8339\n",
      "Epoch 67/2000\n",
      "465/465 - 2s - loss: 3.4877 - val_loss: 0.8278\n",
      "Epoch 68/2000\n",
      "465/465 - 2s - loss: 3.4811 - val_loss: 0.8318\n",
      "Epoch 69/2000\n",
      "465/465 - 2s - loss: 3.4758 - val_loss: 0.8169\n",
      "Epoch 70/2000\n",
      "465/465 - 2s - loss: 3.4703 - val_loss: 0.8272\n",
      "Epoch 71/2000\n",
      "465/465 - 2s - loss: 3.4652 - val_loss: 0.8319\n",
      "Epoch 72/2000\n",
      "465/465 - 2s - loss: 3.4600 - val_loss: 0.8297\n",
      "Epoch 73/2000\n",
      "465/465 - 2s - loss: 3.4560 - val_loss: 0.8254\n",
      "Epoch 74/2000\n",
      "465/465 - 2s - loss: 3.4499 - val_loss: 0.8239\n",
      "Epoch 75/2000\n",
      "465/465 - 2s - loss: 3.4466 - val_loss: 0.8222\n",
      "Epoch 76/2000\n",
      "465/465 - 2s - loss: 3.4419 - val_loss: 0.8288\n",
      "Epoch 77/2000\n",
      "465/465 - 2s - loss: 3.4367 - val_loss: 0.8275\n",
      "Epoch 78/2000\n",
      "465/465 - 2s - loss: 3.4324 - val_loss: 0.8306\n",
      "Epoch 79/2000\n",
      "465/465 - 2s - loss: 3.4288 - val_loss: 0.8338\n",
      "Epoch 80/2000\n",
      "465/465 - 2s - loss: 3.4258 - val_loss: 0.8288\n",
      "Epoch 81/2000\n",
      "465/465 - 2s - loss: 3.4212 - val_loss: 0.8287\n",
      "Epoch 82/2000\n",
      "465/465 - 2s - loss: 3.4172 - val_loss: 0.8229\n",
      "Epoch 83/2000\n",
      "465/465 - 2s - loss: 3.4140 - val_loss: 0.8284\n",
      "Epoch 84/2000\n",
      "465/465 - 2s - loss: 3.4104 - val_loss: 0.8232\n",
      "Epoch 85/2000\n",
      "465/465 - 2s - loss: 3.4067 - val_loss: 0.8319\n",
      "Epoch 86/2000\n",
      "465/465 - 2s - loss: 3.4040 - val_loss: 0.8269\n",
      "Epoch 87/2000\n",
      "465/465 - 2s - loss: 3.4003 - val_loss: 0.8278\n",
      "Epoch 88/2000\n",
      "465/465 - 2s - loss: 3.3970 - val_loss: 0.8159\n",
      "Epoch 89/2000\n",
      "465/465 - 2s - loss: 3.3941 - val_loss: 0.8269\n",
      "Epoch 90/2000\n",
      "465/465 - 2s - loss: 3.3905 - val_loss: 0.8264\n",
      "Epoch 91/2000\n",
      "465/465 - 2s - loss: 3.3875 - val_loss: 0.8328\n",
      "Epoch 92/2000\n",
      "465/465 - 2s - loss: 3.3852 - val_loss: 0.8265\n",
      "Epoch 93/2000\n",
      "465/465 - 2s - loss: 3.3827 - val_loss: 0.8283\n",
      "Epoch 94/2000\n",
      "465/465 - 2s - loss: 3.3786 - val_loss: 0.8314\n",
      "Epoch 95/2000\n",
      "465/465 - 2s - loss: 3.3774 - val_loss: 0.8320\n",
      "Epoch 96/2000\n",
      "465/465 - 2s - loss: 3.3739 - val_loss: 0.8235\n",
      "Epoch 97/2000\n",
      "465/465 - 2s - loss: 3.3718 - val_loss: 0.8267\n",
      "Epoch 98/2000\n",
      "465/465 - 2s - loss: 3.3682 - val_loss: 0.8308\n",
      "Epoch 99/2000\n",
      "465/465 - 2s - loss: 3.3663 - val_loss: 0.8295\n",
      "Epoch 100/2000\n",
      "465/465 - 2s - loss: 3.3638 - val_loss: 0.8246\n",
      "Epoch 101/2000\n",
      "465/465 - 2s - loss: 3.3621 - val_loss: 0.8270\n",
      "Epoch 102/2000\n",
      "465/465 - 2s - loss: 3.3589 - val_loss: 0.8304\n",
      "Epoch 103/2000\n",
      "465/465 - 2s - loss: 3.3572 - val_loss: 0.8322\n",
      "Epoch 104/2000\n",
      "465/465 - 2s - loss: 3.3543 - val_loss: 0.8265\n",
      "Epoch 105/2000\n",
      "465/465 - 2s - loss: 3.3534 - val_loss: 0.8375\n",
      "Epoch 106/2000\n",
      "465/465 - 2s - loss: 3.3508 - val_loss: 0.8307\n",
      "Epoch 107/2000\n",
      "465/465 - 2s - loss: 3.3488 - val_loss: 0.8269\n",
      "Epoch 108/2000\n",
      "465/465 - 2s - loss: 3.3465 - val_loss: 0.8278\n",
      "Epoch 00108: early stopping\n",
      "CPU times: user 13min 40s, sys: 53.3 s, total: 14min 33s\n",
      "Wall time: 3min 5s\n"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 25;\n",
       "                var nbb_unformatted_code = \"%%time\\nhistory = model.fit(\\n    input_train,\\n    y_train.reshape(-1, 1),\\n    epochs=2000,\\n    batch_size=1024,\\n    validation_data=(input_valid, y_valid.reshape(-1, 1),),\\n    verbose=2,\\n    callbacks=[EarlyStopping(monitor=\\\"val_loss\\\", patience=20, verbose=1)],\\n    class_weight=class_weight\\n)\";\n",
       "                var nbb_formatted_code = \"%%time\\nhistory = model.fit(\\n    input_train,\\n    y_train.reshape(-1, 1),\\n    epochs=2000,\\n    batch_size=1024,\\n    validation_data=(input_valid, y_valid.reshape(-1, 1),),\\n    verbose=2,\\n    callbacks=[EarlyStopping(monitor=\\\"val_loss\\\", patience=20, verbose=1)],\\n    class_weight=class_weight\\n)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%time\n",
    "history = model.fit(\n",
    "    input_train,\n",
    "    y_train.reshape(-1, 1),\n",
    "    epochs=2000,\n",
    "    batch_size=1024,\n",
    "    validation_data=(input_valid, y_valid.reshape(-1, 1),),\n",
    "    verbose=2,\n",
    "    callbacks=[EarlyStopping(monitor=\"val_loss\", patience=20, verbose=1)],\n",
    "    class_weight=class_weight\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:20.629881Z",
     "start_time": "2020-09-19T11:55:20.506458Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 26;\n",
       "                var nbb_unformatted_code = \"plot_history(history)\";\n",
       "                var nbb_formatted_code = \"plot_history(history)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_history(history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.115023Z",
     "start_time": "2020-09-19T11:55:20.630852Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6461525366019748"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 27;\n",
       "                var nbb_unformatted_code = \"model_auc = accuracy_score(\\n    y_true=y_valid, y_pred=np.argmax(model.predict(input_valid), axis=1),\\n)\\nmodel_auc\";\n",
       "                var nbb_formatted_code = \"model_auc = accuracy_score(\\n    y_true=y_valid, y_pred=np.argmax(model.predict(input_valid), axis=1),\\n)\\nmodel_auc\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_auc = accuracy_score(\n",
    "    y_true=y_valid, y_pred=np.argmax(model.predict(input_valid), axis=1),\n",
    ")\n",
    "model_auc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.638951Z",
     "start_time": "2020-09-19T11:55:21.115976Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6885810266195612"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 28;\n",
       "                var nbb_unformatted_code = \"model_auc = balanced_accuracy_score(\\n    y_true=y_test, y_pred=np.argmax(model.predict(input_test), axis=1),\\n)\\nmodel_auc\";\n",
       "                var nbb_formatted_code = \"model_auc = balanced_accuracy_score(\\n    y_true=y_test, y_pred=np.argmax(model.predict(input_test), axis=1),\\n)\\nmodel_auc\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_auc = balanced_accuracy_score(\n",
    "    y_true=y_test, y_pred=np.argmax(model.predict(input_test), axis=1),\n",
    ")\n",
    "model_auc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.661912Z",
     "start_time": "2020-09-19T11:55:21.639916Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 29;\n",
       "                var nbb_unformatted_code = \"from xplainet.model import predict, encode\";\n",
       "                var nbb_formatted_code = \"from xplainet.model import predict, encode\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from xplainet.model import predict, encode"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.683942Z",
     "start_time": "2020-09-19T11:55:21.662857Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 30;\n",
       "                var nbb_unformatted_code = \"# This needs to be fixed to work for multiclass\";\n",
       "                var nbb_formatted_code = \"# This needs to be fixed to work for multiclass\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# This needs to be fixed to work for multiclass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.892248Z",
     "start_time": "2020-09-19T11:55:21.684946Z"
    }
   },
   "outputs": [
    {
     "ename": "ValueError",
     "evalue": "cannot reshape array of size 308 into shape (44,1)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-31-eecb7f9a6e50>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mprobs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mexplanations\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpredict\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput_test\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/work/xplainet/src/xplainet/model.py\u001b[0m in \u001b[0;36mpredict\u001b[0;34m(model, input_model)\u001b[0m\n\u001b[1;32m    215\u001b[0m         weights.append(\n\u001b[1;32m    216\u001b[0m             log_reg_weights[consumed : consumed + nb_weights].reshape(\n\u001b[0;32m--> 217\u001b[0;31m                 \u001b[0mnb_features\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnb_channel\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    218\u001b[0m             )\n\u001b[1;32m    219\u001b[0m         )\n",
      "\u001b[0;31mValueError\u001b[0m: cannot reshape array of size 308 into shape (44,1)"
     ]
    },
    {
     "data": {
      "application/javascript": [
       "\n",
       "            setTimeout(function() {\n",
       "                var nbb_cell_id = 31;\n",
       "                var nbb_unformatted_code = \"probs, explanations = predict(model, input_test)\";\n",
       "                var nbb_formatted_code = \"probs, explanations = predict(model, input_test)\";\n",
       "                var nbb_cells = Jupyter.notebook.get_cells();\n",
       "                for (var i = 0; i < nbb_cells.length; ++i) {\n",
       "                    if (nbb_cells[i].input_prompt_number == nbb_cell_id) {\n",
       "                        if (nbb_cells[i].get_text() == nbb_unformatted_code) {\n",
       "                             nbb_cells[i].set_text(nbb_formatted_code);\n",
       "                        }\n",
       "                        break;\n",
       "                    }\n",
       "                }\n",
       "            }, 500);\n",
       "            "
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "probs, explanations = predict(model, input_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.893333Z",
     "start_time": "2020-09-19T11:51:25.138Z"
    }
   },
   "outputs": [],
   "source": [
    "probs, encoded_output = encode(model, input_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.893940Z",
     "start_time": "2020-09-19T11:51:25.139Z"
    }
   },
   "outputs": [],
   "source": [
    "y_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.894543Z",
     "start_time": "2020-09-19T11:51:25.140Z"
    }
   },
   "outputs": [],
   "source": [
    "encoded_output.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.895279Z",
     "start_time": "2020-09-19T11:51:25.141Z"
    }
   },
   "outputs": [],
   "source": [
    "explanations.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.895852Z",
     "start_time": "2020-09-19T11:51:25.142Z"
    }
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# plt.rcdefaults()\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "def explain_plot(importances, columns):\n",
    "    selection = np.argsort(-np.absolute(importances))[:10]\n",
    "    # indexes = np.argsort(importances)\n",
    "    performance = importances[selection]\n",
    "    #     print(performance.shape)512, 256, 128, 64, 32, 1\n",
    "    y_pos = np.arange(performance.shape[0])\n",
    "\n",
    "    plt.barh(y_pos, performance, align=\"center\", alpha=0.5)\n",
    "    plt.yticks(y_pos, columns[selection])\n",
    "    # plt.xlabel('Usage')\n",
    "    plt.title(\"Feature importance\")\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.896488Z",
     "start_time": "2020-09-19T11:51:25.143Z"
    }
   },
   "outputs": [],
   "source": [
    "all_cols = np.array(params[\"bool_cols\"] + params[\"num_cols\"] + params[\"cat_cols\"])\n",
    "all_cols"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-18T18:48:54.151671Z",
     "start_time": "2020-09-18T18:48:54.127488Z"
    }
   },
   "source": [
    "## Explain global"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.897123Z",
     "start_time": "2020-09-19T11:51:25.144Z"
    }
   },
   "outputs": [],
   "source": [
    "probs_train, explanations_train = predict(model, input_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.897717Z",
     "start_time": "2020-09-19T11:51:25.145Z"
    }
   },
   "outputs": [],
   "source": [
    "global_explain = np.sum(explanations_train, axis=0)\n",
    "global_explain = global_explain / np.abs(global_explain).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.898366Z",
     "start_time": "2020-09-19T11:51:25.146Z"
    }
   },
   "outputs": [],
   "source": [
    "global_explain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.899103Z",
     "start_time": "2020-09-19T11:51:25.147Z"
    }
   },
   "outputs": [],
   "source": [
    "explain_plot(global_explain, all_cols)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Explain local"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.899640Z",
     "start_time": "2020-09-19T11:51:25.148Z"
    }
   },
   "outputs": [],
   "source": [
    "for i in range(20):\n",
    "    explain_plot(explanations[i], all_cols)\n",
    "    print(probs[i].item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.900374Z",
     "start_time": "2020-09-19T11:51:25.149Z"
    }
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.900956Z",
     "start_time": "2020-09-19T11:51:25.150Z"
    }
   },
   "outputs": [],
   "source": [
    "np.unique(y_test, return_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.901608Z",
     "start_time": "2020-09-19T11:51:25.151Z"
    }
   },
   "outputs": [],
   "source": [
    "y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.902285Z",
     "start_time": "2020-09-19T11:51:25.153Z"
    }
   },
   "outputs": [],
   "source": [
    "target_encoder.classes_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-09-19T11:55:21.902961Z",
     "start_time": "2020-09-19T11:51:25.153Z"
    }
   },
   "outputs": [],
   "source": [
    "confusion_matrix(\n",
    "    y_true=y_test,\n",
    "    y_pred=model.predict(input_test).reshape(-1) >= 0.5,\n",
    "    # labels=target_encoder.classes_,\n",
    ")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}