Customer Churn Prediction Using Neural Networks

About

This project focuses on predicting customer churn (whether a customer will leave or stay with the company) using a deep learning model. The dataset contains various features about customers, including demographic details, account information, and service usage. The goal is to train a neural network model to classify whether a customer will churn (Exited = 1) or not (Exited = 0).

The model uses a fully connected neural network (also known as a dense network) built with TensorFlow and Keras. The project also includes data preprocessing, exploratory data analysis (EDA), and evaluation using various metrics like accuracy, F1-score, and confusion matrix.

Dataset

The dataset used for this project is Churn_Modelling.csv, which contains the following columns:

• RowNumber: Row number in the dataset (not relevant for prediction)
• CustomerId: Unique customer ID (not relevant for prediction)
• Surname: Customer's surname (not relevant for prediction)
• CreditScore: Credit score of the customer
• Geography: The country where the customer is located (France, Germany, Spain)
• Gender: Gender of the customer (Male/Female)
• Age: Age of the customer
• Tenure: Number of years the customer has been with the company
• Balance: Customer's account balance
• NumOfProducts: Number of products the customer has purchased
• HasCrCard: Whether the customer has a credit card (1 = Yes, 0 = No)
• IsActiveMember: Whether the customer is an active member (1 = Yes, 0 = No)
• EstimatedSalary: Estimated annual salary of the customer
• Exited: Target variable (1 = Churned, 0 = Stayed)

Features

• Data Preprocessing: Cleaned the dataset by removing irrelevant columns, handling categorical features (Gender, Geography), and scaling continuous variables (Balance, EstimatedSalary).
• Exploratory Data Analysis (EDA): Visualized customer tenure and salary distributions to gain insights into how they relate to churn.
• Model: Built a fully connected neural network model with two hidden layers using TensorFlow/Keras.
• Model Evaluation: Evaluated the model using accuracy, precision, recall, F1-score, and confusion matrix.

Project Structure

.
├── Churn_Modelling.csv       # The dataset used for training
├── churn_prediction_model.py # Python script with data preprocessing, model creation, and evaluation
└── README.md                 # This README file

Requirements

To run this project, you will need to install the following libraries:

• Python 3.x
• TensorFlow
• pandas
• matplotlib
• scikit-learn
• seaborn

You can install the dependencies using pip:

pip install tensorflow pandas matplotlib scikit-learn seaborn

Usage

Step 1: Load the dataset

The Churn_Modelling.csv dataset should be available in the same directory as the script.

Step 2: Preprocess the data

The dataset is cleaned by:

• Removing irrelevant columns.
• Encoding categorical columns (Gender, Geography).
• Scaling numerical features.

Step 3: Split the data

The dataset is split into training and test sets, with a stratified split to ensure equal distribution of the target variable (Exited).

Step 4: Train the model

A deep neural network model is built using TensorFlow's Keras API, with the following architecture:

• Input layer (12 features)
• 2 hidden layers (100 units each, ReLU activation, Batch Normalization)
• Output layer (1 unit, sigmoid activation)

The model is compiled using the adam optimizer and binary_crossentropy loss function.

Step 5: Model Evaluation

The model is trained for 100 epochs. After training, the model's performance is evaluated on the test set using metrics like accuracy, precision, recall, F1-score, and the confusion matrix.

Step 6: Visualization

A confusion matrix heatmap is displayed using Seaborn to visualize the true vs. predicted churn values.

Results

After training the model, the performance metrics (accuracy, precision, recall, and F1-score) provide a clear understanding of how well the model is performing. The confusion matrix also highlights the misclassifications made by the model (false positives and false negatives).

Potential Improvements

• Handling Class Imbalance: Use techniques such as oversampling the minority class or adjusting class weights to handle the class imbalance in the dataset.
• Hyperparameter Tuning: Perform grid search or random search for hyperparameter optimization to improve model performance.
• Alternative Models: Try other machine learning models (e.g., Random Forest, XGBoost, Logistic Regression) and compare performance.

License

This project is licensed under the MIT License.

---