Skip to content

Latest commit

 

History

History
83 lines (54 loc) · 3.75 KB

README.md

File metadata and controls

83 lines (54 loc) · 3.75 KB

DEV_Projects

Telco Customer Churn Prediction

Overview

This project aims to predict customer churn for a telecommunications company using machine learning techniques. Churn, also known as customer attrition, refers to the phenomenon where customers stop doing business with a company. By accurately predicting which customers are likely to churn, the company can proactively implement retention strategies to reduce churn rates and improve profitability.

Dataset

The project uses the "Telco Customer Churn" dataset available on Kaggle: https://www.kaggle.com/datasets/blastchar/telco-customer-churn

The dataset contains information about customers of a telecom company, including demographic data, services subscribed to, account information, and whether or not the customer churned.

Project Structure

  • churn_prediction.py: The main Python script containing the code for data preprocessing, model training, hyperparameter tuning, evaluation, and feature importance analysis.
  • WA_Fn-UseC_-Telco-Customer-Churn.csv: The original dataset file.
  • requirements.txt: Lists the required Python libraries for running the code.
  • Initial_Model_roc_curve.png: ROC curve for the initial logistic regression model.
  • Best_Model_roc_curve.png: ROC curve for the logistic regression model with tuned hyperparameters.
  • feature_importance.png: Feature importance plot for the best model.

Approach

The project follows these main steps:

  1. Data Loading and Preprocessing:

    • Load the dataset into a Pandas DataFrame.
    • Handle missing values (in TotalCharges) by imputing with the mean.
    • Convert relevant features (e.g., SeniorCitizen) to categorical data types.
    • Perform one-hot encoding of categorical features.
    • Split the data into training and testing sets.

  2. Model Training and Evaluation (Logistic Regression):

    • Train an initial logistic regression model on the training set.
    • Evaluate the model on the test set using precision, recall, F1 score, accuracy, and AUC-ROC.
    • Visualize the ROC curve to assess the model's ability to discriminate between classes.

  3. Hyperparameter Tuning (Logistic Regression):

    • Use GridSearchCV to find the optimal hyperparameters for the logistic regression model, including C, penalty, solver, and class_weight.

  4. Retraining and Final Evaluation (Logistic Regression):

    • Retrain the logistic regression model using the best hyperparameters found in the previous step.
    • Evaluate the model's performance on the test set.
    • Analyze feature importance to identify the most significant predictors of churn.

Results

The best logistic regression model achieved the following performance on the test set:

Metric Score
Precision 0.62
Recall 0.51
F1 Score 0.56
Accuracy 0.79
AUC-ROC 0.83

The specific results may vary slightly depending on the random state of the data split.

ROC Curves

Initial Model Best Model (Tuned)
Initial Model ROC Curve Best Model ROC Curve

Feature Importance

Feature Importance Plot

Limitations and Future Directions

The model's recall could be improved, as missing out on potential churners can be costly for the company. We could try incorporating more features, experimenting with different algorithms (e.g., XGBoost), or employing techniques like SMOTE to address class imbalance.

How to Run

  1. Clone the Repository:

    git clone [https://github.com/](https://github.com/)<DevasivaBA>/<DEV_Projects>.git