Custom-Churn-Prediction# Churn Prediction Using Deep Learning - First Attempt

📌 Project Overview

This is my first attempt at predicting customer churn using a Deep Learning model built with TensorFlow & Keras. The dataset used is Churn_Modelling.csv, and the goal is to classify whether a customer will exit (churn) or stay with the bank.

🎯 Key Highlights:

• Built a Sequential Neural Network with TensorFlow/Keras
• Achieved ~80% accuracy
• Applied One-Hot Encoding & Feature Scaling
• Used ReLU & Sigmoid activations
• Trained for 100 epochs

---

📂 Dataset Information

The dataset consists of customer details, such as Geography, Gender, Credit Score, Balance, etc.

🔑 Features Used:

Feature Name	Description
CreditScore	Customer's credit score
Geography	Country of residence
Gender	Male/Female
Age	Customer's age
Tenure	Years of account ownership
Balance	Account balance
NumOfProducts	Number of bank products used
HasCrCard	Whether the customer has a credit card
IsActiveMember	Whether the customer is active
EstimatedSalary	Estimated salary

🎯 Target Variable:

• Exited (1 = Churn, 0 = No Churn)

---

🔨 Model Architecture

# Build the neural network
model = Sequential([
    Input(shape=(X_train.shape[1],)),
    Dense(3, activation='relu'),
    Dense(11, activation='relu'),
    Dense(1, activation='sigmoid')
])

Hyperparameters

• Optimizer: Adam
• Loss Function: Binary Crossentropy
• Epochs: 100
• Validation Split: 20%

---

📊 Results & Observations

• Accuracy on Test Set: ~80%
• Validation Accuracy fluctuated across epochs
• The model could be further optimized by adjusting layers, neurons, and hyperparameters.

📉 Accuracy & Loss Curves:

---

📌 Next Steps

✅ Hyperparameter Tuning (Neurons, Layers, Regularization, Dropout) ✅ Reduce Overfitting (Better architecture) ✅ Increase Accuracy (Aim for 85-90%)

---

📜 Learning Takeaways

📌 Neural networks require careful tuning to improve performance. 📌 Feature scaling & encoding are crucial for better training. 📌 Churn prediction can benefit from additional feature engineering.

---

📎 Repository Structure

📂 Churn_Prediction_DL
├── 📄 Churn_Modelling.csv
├── 📄 churn_prediction.ipynb
├── 📄 requirements.txt
├── 📄 README.md  

---

📦 Installation & Requirements

To install dependencies, run:

pip install -r requirements.txt

requirements.txt

tensorflow
keras
numpy
pandas
matplotlib
scikit-learn

---

📌 Future Update (86% Accuracy Version) 🚀

I plan to improve this model with better hyperparameter tuning & architecture adjustments. Stay tuned for a follow-up post! 🎯

📌 Follow me on LinkedIn for updates! 😊