Churn Prediction Project Report and Documentation

Check out the Complete report here

1. Data Preprocessing

Summary Statistics for Continuous Features

• Calculate statistics for numerical features.
• Check the distribution of continuous variables.

Number of Unique Values in Categorical Columns

• Explored the unique values in each categorical feature.
• Analyzed the cardinality of categorical variables.

Missing Values Check

• Verify if there are any missing values in the dataset.
• Ensure data completeness.

2. Exploratory Data Analysis (EDA)

• Analyze age vs. churn rate.
• Explore the distribution of subscription length.
• Investigate the distribution of monthly bills.
• Examine the distribution of total data usage.
• Visualize churn rate by location.

3. Feature Engineering

• Perform one-hot encoding for Gender.
• Created the Billing_Per_Usage feature, which is the ratio of Monthly_Bill to Total_Usage_GB.
• Generated the Tenure feature, representing the remaining subscription length.
• Encoded the Location feature using TargetEncoder.
• Scale numerical features using StandardScaler.
• Split the data into training and testing sets.

4. Model Building and Optimization

• Choose the Logistic Regression model for binary classification.
• Utilized Randomized Search CV for hyperparameter optimization.
• Defined hyperparameter ranges for logistic regression.

Hyperparameter Tuning (RandomizedSearchCV)

- Penalty: ['l1', 'l2', 'elasticnet', 'none']
- C: Uniform distribution between 0 and 4
- Solver: ['newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga']
- Maximum Iterations: [100, 200, 300, 400, 500]

5. Model Deployment

• Create a Flask API for model deployment.
• Expose a /predict endpoint that accepts JSON requests with customer features.
• Return JSON responses with predicted churn status.

Folder Structure

• app.py: Flask application.
• churn.ipynb: Jupyter notebook for the project.
• config/: Directory containing model, encoder, and scaler files.
• data/: Directory with the dataset file.
• test.py: Python script for testing the deployed model.
• transform.py: Feature engineering script.

6. Usage

To run the project:

1. Start the Flask server with python app.py.
2. Use test.py to send POST requests to the /predict endpoint to make predictions.

Example Input JSON:

{
    "Age": [32],
    "Gender": ["Male"],
    "Location": ["Houston"],
    "Subscription_Length_Months": [12],
    "Monthly_Bill": [48.76],
    "Total_Usage_GB": [172]
}

Example Output JSON :

{'predictions': [1]}