CS4089 ProJect

Churn Prediction project

Professor : Dr. Schwartzman

---

Content

1. Introduction
2. Dataset Overview
3. Project Structure
4. Installation
5. Data Analysis (EDA)
6. Preprocessing
7. Feature
8. Modeling
9. Evaluation
10. Conclusion
11. Acknowledgements

---

Customer churn is a issue in the telecommunications industry. Understanding why customers leave and predicting churn can help companies come up with new strategies to retain customers. This builds a machine learning model to predict customer churn based on various data provided in the csv file.

---

Dataset

The dataset contains information about a fictional tele communication company's customers (CX). It includes customer demographic info, account details, and the services they have subscribed to.

• Features Include:

  • CustomerID
  • Demographic: Gender, Senior Citizen status, Partner, and Dependents.
  • Account Information: Tenure, Contract type, Payment method, Monthly charges, and Total charges.
  • Services: Phone service, Multiple lines, Internet service, Online security, Online backup, Device protection, Tech support, Streaming TV, and Streaming movies.

• Target Variable:

  • Churn: If the customer churned (Yes) or not (No).

---

file tree

ML-churn/
├── data/
│   └── telco_churn.csv
├── images/
│   ├── churn_distribution.png
│   ├── numerical_distributions.png
│   └── correlation_matrix.png
├── src/
│   ├── __init__.py
│   ├── data_loading.py
│   ├── eda.py
│   ├── preprocessing.py
│   ├── feature_engineering.py
│   ├── modeling.py
│   ├── evaluation.py
│   └── visualization.py
├── main.py
├── requirements.txt
└── README.md

• data/: dataset
• images/
• src/: modules
• main.py: main logic
• requirements.txt: packages needed to run this
• README.md

---

TO START

1. clone the repo

   git clone https://github.com/INAHIDC/ML-churn.git
   cd ML-churn

2. Dependencies ;skip if you have it already

make sure to use a venv for this!!!!!!!!

pip install -r requirements.txt

3. run it

   python main.py

---

Analysis

Understanding the data is important before going into modeling.

1. Churn Distribution

We start by examining the distribution of the target variable.

![Churn Distribution]

• Observation: The dataset is imbalanced, with a higher number of customers who did not churn.

2. Numerical Features Distribution

We plotted histograms for numerical features like tenure, MonthlyCharges, and TotalCharges.

![Numerical Distributions]

• Observation:
  • Tenure: lots of customers are either new or have been with the company for a long time.
  • MonthlyCharges: a wide range of monthly charges.
  • TotalCharges: Similar to tenure, reflects the cumulative amount charged.

3. Correlation Matrix

analyzed correlations between numerical variables.

![Correlation Matrix]

• Observation: correlation between TotalCharges and tenure, which makes sense as longer tenure typically results in higher total charges.

---

Data Preprocessing

preprocessing include handling missing values, encoding categorical variables

1. Handling Missing Values

• Replaced empty strings with NaN.
• Dropped rows with missing values.

2. Encoding Categorical Variables

• Label: binary categorical variables.
• One Hot: categorical variables with more than two categories.

3. Feature Scaling

• standardization: added to numerical features to normalize the data.

---

Feature Engineering

• TotalServices: Summed up the number of services a customer has subscribed to.

---

Modeling

two models to predict customer churn.

1. Logistic Regression

simple model to establish a baseline.

• Pros: Easy to understand.
• Cons: does not show complex relationships

2. Random Forest Classifier

An ensemble model to improve performance.

• Pros: nonlinear relationships.
• Cons: not understable/ or intuitive

---

Evaluation

evaluated the models using classification reports, confusion matrices, and ROC curves.

1. Logistic Regression Results

• Accuracy: Moderate.
• Precision and Recall: the model's performance on predicting churned CX.
• Confusion Matrix and ROC Curve: shows true positives, false positives etc.

2. Random Forest

• Accuracy: Improved compared to Logistic Regression.
• helps find which features contribute most to predictions.
• Confusion Matrix and ROC Curve: Showed better performance in distinguishing churned customers.

---

Conclusion

• What do i think?

  • Forest model outperformed Logistic Regression in predicting customer churn.
  • Features like contract type, tenure, and monthly charges are important predictors.

• Implications:

  • Cx Retention: Focus on customers with month to month contracts and high monthly charges.
  • Service Improvement: improve services that are linked to higher churn rates.

---

source

• Dataset Source: IBM Cognos Analytics Dataset.