Customer Churn Prediction

A Project as an ML Intern given by Innobytes Services

A Classification model that predicts whether a customer will churn or not.

Open Notebook 🔥

       

Predicting Customer Churn using Machine Learning

This project builds a machine learning model capable of predicting whether a customer will churn (leave) based on their characteristics. The data comes from a Telco customer churn dataset.

1. Problem Definition

Understanding how customer attributes contribute the probability of their churn, insights & patterns etc.

How accurately can we predict customer churn based on customer attributes like usage behavior, contract type, and services used?

2. Data

The dataset is sourced from Telco's customer churn records and includes the following features:

• Customer demographics (gender, senior citizen status)
• Services used (internet, phone, etc.)
• Contract types (month-to-month, one year, two years)
• Billing information (monthly and total charges)

I have used the specific dataset provided in the Internship. For more details explore below sources.

Official Sources : Kaggle & IBM

3. Evaluation

We evaluate the model using classification metrics such as accuracy, F1-score, precision, recall, and ROC-AUC. The primary goal is to minimize misclassification errors, particularly false negatives.

Also our prime focus was to maximize the tradeoff between Recall & AUC Score.

4. Features

The dataset contains customer data with various service attributes. Key features include contract type, paperless billing, payment method, monthly charges, and tenure.

After model driven EDA, we understand the following features are crucial to predict Customer Churn!

5. Key Achievements

1. Data Preprocessing Pipeline:

   • Developed a robust preprocessing pipeline for cleaning and encoding features.
   • Efficiently handled missing data and categorical variables.

2. Machine Learning Model Development:

   • Built multiple classifiers: Logistic Regression, Decision Trees, Random Forest, XGBoost, LightGBM, and CatBoost.
   • Applied cross-validation for model evaluation.

3. Hyperparameter Tuning:

   • Used Bayesian optimization with BayesSearchCV to improve model performance by fine-tuning hyperparameters.

Final Result

The final model achieved a recall of 90% & AUC score of 75% , successfully predicting customer churn with good accuracy.

Confusion Matrix & ROC Curve for predictions!

For more details, refer to the Jupyter Notebook file.

---

Run the model locally!

1. pip install -r requirements_to_run.txt # install requirements!

2. streamlit run Streamlit_front/main_streamlit.py # From root directory

3. Open the server in your browser: http://127.0.0.1:8501/

4. Get predictions for you customer either with the form or upload a CSV file.

Other images :

Input data format	Uploading CSV File

---

Author: Dipesh Rathore 😎

Complete Analysis : Notebook

---