Customer Churn Prediction Analysis

This project uses the Telco dataset

Steps

1. Data Ingestion
2. Data Storage
3. Data Processing
4. Data Transformation
5. Exploratory Data Analysis(EDA)
6. Machine Learning Model
7. Visualization and Reporting
8. Deployment

Telco Churn Analysis

Overview

Telco Churn Analysis is a data engineering project that aims to analyze customer churn in the telecommunications sector. Using clustering, classification, and visualization techniques, this project provides insights into customer behavior and identifies factors contributing to churn. It enables businesses to devise strategies for customer retention and improve overall profitability.

Features

• Data Preprocessing: Handling missing values, scaling numerical features, and encoding categorical variables.
• Exploratory Data Analysis (EDA): Understanding key trends and correlations in the dataset.
• Customer Segmentation: Clustering customers into distinct groups based on their behavior.
• Churn Prediction: Using machine learning models to predict the likelihood of customer churn.
• Visualization: Graphical representations like PCA-based scatter plots and the Elbow Method for cluster analysis.

Key Techniques Used

• Clustering: K-Means algorithm to segment customers.
• Dimensionality Reduction: PCA (Principal Component Analysis) for visualizing clusters in 2D.
• Classification Metrics: Log Loss and other evaluation metrics to measure model performance.

Tools and Libraries

• Python: Programming language.
• Pandas: Data manipulation and analysis.
• Seaborn and Matplotlib: Data visualization.
• Scikit-learn: Machine learning algorithms and preprocessing.

Workflow

1. Data Preprocessing:

   • Cleaned the data by handling missing values.
   • Scaled numerical features using StandardScaler.
   • Encoded categorical variables (if present).

2. Exploratory Data Analysis (EDA):

   • Visualized trends in tenure, monthly charges, and total charges.
   • Analyzed churn distribution across various features.

3. Customer Segmentation:

   • Determined the optimal number of clusters using the Elbow Method.
   • Used K-Means to group customers.
   • Visualized clusters in 2D using PCA.

4. Churn Prediction:

   • Built machine learning models to predict churn probability.
   • Evaluated models using Log Loss.

5. Visualization:

   • Scatter plots for PCA components to visualize clusters.
   • Line plot for the Elbow Method to select the optimal number of clusters.

Results and Insights

• Identified distinct customer segments based on tenure and total charges.
• Highlighted clusters with higher churn rates, enabling targeted retention strategies.
• Achieved a log loss of 0.5116, indicating a reasonably good prediction model.

How to Use

1. Clone the repository:

   git clone https://github.com/yourusername/telco-churn-analysis.git

2. Navigate to the project directory:

   cd telco-churn-analysis

3. Install dependencies:

   pip install -r requirements.txt

4. Run the Jupyter Notebook:

   jupyter notebook Telco_Customer_Churn.ipynb

Dataset

The dataset used in this project is publicly available and contains information about customers in the telecommunications industry, including:

• Tenure: Duration of the customer’s relationship with the company.
• Monthly Charges: The amount billed monthly.
• Total Charges: The total amount billed to date.
• Churn: Indicates whether the customer has left the company.

Visualizations

• Elbow Method: Used to determine the optimal number of clusters for customer segmentation.
• PCA Scatter Plot: Visualized customer clusters in two dimensions.
• Churn Distribution: Highlighted churn rates across different customer segments.

Future Work

• Incorporate additional features for a deeper understanding of customer behavior.
• Experiment with advanced clustering techniques like DBSCAN or hierarchical clustering.
• Improve prediction accuracy with hyperparameter tuning and ensemble models.

Contributing

Contributions are welcome! Feel free to open issues or submit pull requests to enhance the project.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• Dataset: Provided by [Dataset Source/Platform].
• Libraries: Thanks to the open-source libraries used in this project.

---

Start exploring customer churn with Telco Churn Analysis!