clothing_recommender_project.py

# -*- coding: utf-8 -*-
"""Clothing_Recommender Project .ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1nw0ewNdkx8o3WULAp2ynhHpbq1kVq7YZ

Clean the data and use input
"""

## Import and Organize Data ##

import pandas as pd 
import numpy as np 
import matplotlib.pyplot as plt 
from sklearn.model_selection import train_test_split

#read clean file (downloaded from Task 1)
df=pd.read_csv('CleanedData.csv', sep=',') 

#Pivot table (clothingID, age, rating) - Nan is replaced with 0 
train = df.pivot_table(index='Age', columns='ClothingID', values='Rating')

#sort train data
train = train.sort_values('Age', ascending=True)

###Create a greeting 

print("Welcome, let us recommend a product for you")

#Take user input 

Name =input('Please enter your name: ')
Age = int(input('Please enter your age: '))
CID_user = int(input("Enter Clothing ID: ")) #90
while CID_user not in train.columns:
  print('Invalid: No data for ID')
  CID_user = int(input("Enter valid Clothing ID: "))
rating_user = float(input("Enter Rating for Clothing ID: ")) #4

##use this later (if user has more than one rating to enter)
#entries = int(input("How many ratings will you enter? "))
#for x in range(entries):

#create array with user data 
userArray = pd.DataFrame().reindex_like(train)
userArray.dropna(thresh=1,inplace=True)
userArray.loc[Age,CID_user] = rating_user #enter user data 

from sklearn.metrics.pairwise import nan_euclidean_distances

#find euclidean distance between all rows of train and first row of test  *ignores nan
distance = np.zeros((0,2)) #create empty array 
for index, row in train.iterrows():  #iterate through each row of train 
  result = float(nan_euclidean_distances([userArray.loc[Age]], [train.loc[index]])) #compute the euclidean distance between two rows, *confirmed it works thru excel
  result_array = [index, result] #place age and distance into an array 
  distance = np.append(distance,[result_array],axis= 0) 

#convert array to a dataframe
dfDistance = pd.DataFrame({'Age': distance[:, 0], 'E-Distance': distance[:, 1]})
dfDistance.head()

k= 5
#sort by distance, reset the index 
dfDistance = dfDistance.sort_values('E-Distance', ascending=True).head(20)
dfDistance = dfDistance.reset_index(drop=True) 
dfDistance.drop(dfDistance[dfDistance.index > k-1].index, inplace=True)
dfDistance.head()

#NOTE: for calculating the predicted rating, could use an IDW Interpolation function shown here https://stackoverflow.com/questions/3104781/inverse-distance-weighted-idw-interpolation-with-python
#just using mean of each to test a solution, will come back and try more complex/accurate functions later 

#assume k of 5####
k_array = pd.DataFrame().reindex_like(train)
meanArray = pd.DataFrame()

for x in dfDistance['Age']:
  k_array = k_array.append([train.loc[x]]) #make array of the k closest ages

meanArray = meanArray.append(k_array.mean(),ignore_index = True).transpose()
meanArray.dropna(axis=0,inplace=True)
meanArray.columns = ["Mean"]
meanArray = meanArray[meanArray.Mean == 5]

recommend = list(meanArray.index.values)
print("recommended ClothingID's are: ")
print(recommend)

#feedback, clothingID (choose top 5), department 
#reverse lookup clothingID for department 
# feedback (choose first 3)