Skip to content

yash-meshram/predicting-red-hat-business-value

Repository files navigation

Predicting Red Hat Business Value

Red Hat has a large amount of data about the behavior of individuals over time. They wanted to use this data to predict which individual they should approach.

Table of Contents

Overview

  • Predicted the potential business value of a person based on its characteristics and activities.
  • Done Data Cleaning, Feature Engineering, and Feature Selection
  • Build the Machine Learning model (Random Forest Classifier) for predicting the business value

Data Description

Data Source: Kaggle

The dataset contains 3 files:

File Name Contains Description of file
act_train.zip act_train.csv Contains all the activities that each person had performed, characteristic of these activities, and the outcome (whether or not the company approach them or not)
act_test.zip act_test.csv Contains all the activities that each person had performed and characteristics of these activities
people.zip people.csv Contains all the unique people who have performed activities and their characteristics

Approach

Data Cleaning

  • Covert the string elements to int or float elements
  • Change the formate of dates

Feature Engineering

In act_train.csv, we have a column named "date" which gives the date on which the person does an activity. There is also a column labeled "date" in the people.csv file which gives you the date at which people characteristics is been noted. When we merge these two files we have columns named "date_x" (representing date column from act_train.csv file) and "date_y" (representing date column from people.csv file). Both of these features can be useful for our prediction. Thus we will introduce a new feature "date" which will be the difference of date_x and date_y.

date = date_x - date_y

Feature Selection

Lets have a look in training datset. Here we can observe that almost 92% of the elements in columns (from char_1 to char_9) are non-null. Thus we can eliminate these features.

Image

Traning Model

Build RandomForestClassifier model to predict the outcome

Evaluation

The model had been evaluated using AUC Metric. Got an accuracy of 99%

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published