/Churn-Prediction-Ensemble-Techniques

Customer Churn classification using various ensemble techniques.

Primary LanguageJupyter NotebookMIT LicenseMIT

Churn Prediction using Ensemble Techniques

Business Objective

In our case study, we will be working on a churn dataset. Churned Customers are those who have decided to end their relationship with their existing company.

XYZ is a service-providing company that provides customers with a one-year subscription plan for their product. The company wants to know if the customers will renew the subscription for the coming year or not.

Data Description

This data provides information about a video streaming service company, where they want to predict if the customer will churn or not. The CSV consists of around 2000 rows and 16 columns.

Tech Stack

  • Language: Python
  • Libraries: NumPy, Pandas, Matplotlib, scikit-learn, pickle, imbalanced-learn (imblearn), LIME

Approach

  1. Importing the required libraries and reading the dataset.
  2. Feature Engineering
    • Dropping unwanted columns.
  3. Model Building
    • Performing train-test split
    • Random Forest Model
    • AdaBoost Model
    • Gradient Boosting Model
  4. Model Validation (Predictions)
    • Recall score
    • Precision score
    • F1-score
    • ROC and AUC
  5. Feature Importance
    • Create a function to find important features.
    • Plot the features.
  6. LIME Implementation
    • Define a function for implementing the LIME technique over the dataset.

Modular Code Overview

  1. input: Contains all the data for analysis, including a CSV file.
  2. src: This is the most important folder of the project. It contains all the modularized code for all the above steps in a modularized manner. This folder consists of:
    • Engine.py
    • ML_Pipeline

The ML_Pipeline is a folder that contains all the functions put into different Python files, which are appropriately named. These Python functions are then called inside the Engine.py file.

  1. output: The output folder contains three subfolders:

    • LIME_reports: Contains the LIME reports generated for all three algorithms.
    • Models: Contains the models generated for all three algorithms.
    • ROC_curves: Contains the ROC curves generated for all three algorithms.

  2. lib: This is a reference folder. It contains the original IPython notebook that we saw in the videos.