Coursera - Data Science Challenge - Customer Churn Prediction

flowchart TB
    %% Data Sources
    subgraph "Data Sources"
        train["train.csv"]:::dataFile
        test["test.csv"]:::dataFile
        desc["data_descriptions.csv"]:::dataFile
    end

    %% Jupyter Notebook Environment
    jnb["Jupyter Notebook Environment (ChurnPrediction.ipynb)"]:::notebook

    %% Processing Modules inside the Notebook Pipeline
    preproc["Data Cleaning & Preprocessing\n(Handles missing values, categorical conversion, One-Hot Encoding via ColumnTransformer)"]:::process
    feat["Feature Engineering\n(Generates features from subscription details and usage patterns)"]:::process
    model["Model Training & Evaluation\n(Logistic Regression with hyperparameters, ROC AUC evaluation)"]:::model

    %% Output and Business Outcome Components
    output["Prediction Output\n(prediction_submission.csv)"]:::dataFile
    business["Business Insights\n(Churn Intervention Strategies)"]:::business
    doc["Documentation\n(README.md)"]:::doc

    %% Data Flow
    train --> jnb
    test  --> jnb
    desc  --> jnb
    jnb --> preproc
    preproc --> feat
    feat --> model
    model --> output
    output --> business
    output --> doc

    %% Click Events
    click train "https://github.com/geethasagarb/coursera-churn-prediction-challenge/blob/main/train.csv"
    click test "https://github.com/geethasagarb/coursera-churn-prediction-challenge/blob/main/test.csv"
    click desc "https://github.com/geethasagarb/coursera-churn-prediction-challenge/blob/main/data_descriptions.csv"
    click jnb "https://github.com/geethasagarb/coursera-churn-prediction-challenge/blob/main/ChurnPrediction.ipynb"
    click output "https://github.com/geethasagarb/coursera-churn-prediction-challenge/blob/main/prediction_submission.csv"
    click doc "https://github.com/geethasagarb/coursera-churn-prediction-challenge/blob/main/README.md"

    %% Styles
    classDef dataFile fill:#ffefba,stroke:#d35400,stroke-width:2px;
    classDef notebook fill:#d6eaf8,stroke:#2e86c1,stroke-width:2px;
    classDef process fill:#d5f5e3,stroke:#27ae60,stroke-width:2px;
    classDef model fill:#fdebd0,stroke:#e67e22,stroke-width:2px;
    classDef business fill:#e8f8f5,stroke:#16a085,stroke-width:2px;
    classDef doc fill:#fadbd8,stroke:#c0392b,stroke-width:2px;

Loading

Project Overview

This project focuses on predicting customer churn for a video streaming service using machine learning. I built a Logistic Regression model with feature engineering and pipeline-based preprocessing, achieving a ROC AUC score of 75.1%. The goal is to identify high-risk churn customers early, enabling the business to take proactive retention measures and reduce revenue loss.

---

Dataset Details

• Training Data: 243,787 records
• Test Data: 104,480 records
• Features: Subscription details, payment methods, content preferences, usage behavior, and demographics
• Target Variable: Churn (Binary: 1 = Churn, 0 = Retained)

---

Approach & Methodology

1. Data Cleaning & Preprocessing

• Handled missing values and ensured data consistency.
• Converted binary categorical variables (Yes/No) into numeric form.
• Applied One-Hot Encoding to categorical features using ColumnTransformer.

2. Feature Engineering

• Engineered relevant features from subscription type, user activity, and preferences.
• Created new insights from existing attributes to improve model performance.

3. Model Development

• Used Logistic Regression within a pipeline to streamline preprocessing and training.
• Optimized hyperparameters (max_iter = 1000, C=1.0, solver=lbfgs).
• Evaluated model performance using ROC AUC score.

4. Results & Business Impact

• Achieved a 75.1% ROC AUC score.
• Provided insights to help the business implement targeted retention strategies and reduce customer churn.

---

How the Business Problem is Solved with Logistic Regression

The core business problem was to identify customers at high risk of churning from the video streaming service. By using a Logistic Regression model with feature engineering and preprocessing, we achieved the following key outcomes:

Early Churn Prediction:

• The model predicts churn likelihood for each customer, allowing proactive retention strategies.
• With a 75.1% ROC AUC score, it effectively differentiates between customers likely to churn vs. stay.

Data-Driven Decision Making:

• The model provides insights into which customer segments (e.g., payment methods, subscription types, content preferences) are more likely to churn.
• This helps in tailoring marketing efforts and offering personalized retention incentives.

Cost Reduction & Revenue Growth:

• Preemptive retention strategies (e.g., discounts, content recommendations) can be targeted at high-risk users.
• Preventing churn directly translates to higher customer lifetime value (CLV) and reduced acquisition costs.

Scalability & Automation:

• The pipeline-based model can easily be updated with new data, making it scalable for future use.
• It can be deployed as an API or integrated into a CRM system for real-time churn prediction.

Conclusion

This solution aligns with the business objective by enabling strategic interventions before customers churn, leading to higher retention, improved profitability, and sustained user engagement.

---

Tools & Technologies Used

• Programming Language: Python
• Libraries: Pandas, NumPy, Scikit-learn
• Machine Learning: Logistic Regression, OneHotEncoder, ColumnTransformer
• Development Environment: Jupyter Notebook

---

Link to Coursera Data Science Challenge

For more details and to view the challenge notebook, visit the following link:
Coursera Data Science Challenge - Churn Prediction

---

How to Run the Project

1. Clone the repository:

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

2. Install dependencies:

   pip install -r requirements.txt

3. Run the Jupyter Notebook to train the model and generate predictions.

---