Churn Management Model for Video Streaming Service

Introduction

In this project, a machine learning model to predict customer churn for a video streaming service is buit. Subscription services often face the challenge of retaining customers, and predicting churn is crucial for targeted interventions. This repository contains a complete data science workflow to tackle this problem, using a unique dataset provided for the challenge.

Project Structure

The project is organized into the following phases:

1. Data Loading: Load the train and test datasets.
2. Exploratory Data Analysis (EDA): Analyze the data to understand distributions, correlations, and basic statistics.
3. Data Cleaning: Handle missing values and prepare the data for modeling.
4. Feature Selection: Select important features based on correlation and feature importance from a RandomForest model.
5. Model Building: Build a RandomForestClassifier to predict churn.
6. Model Evaluation: Evaluate the model using cross-validation.
7. Prediction: Make predictions on the test dataset.

Datasets

• train.csv: Training dataset with 243,787 subscriptions and the target variable Churn.
• test.csv: Test dataset with 104,480 subscriptions for which predictions are to be made.
• data_descriptions.csv: Description of the dataset features.

Requirements

• Python 3.6+
• Libraries: pandas, numpy, seaborn, matplotlib, scikit-learn, shap

Installation

1. Clone the repository:

   git clone https://github.com/your-username/churn-management-model.git
   cd churn-management-model

2. Install the required libraries:

   pip install pandas numpy seaborn matplotlib scikit-learn shap

Results

The model's performance is evaluated using ROC AUC. Mean CV ROC AUC: 0.7