Customer Churn Prediction Using Machine Learning

📌 Project Overview

This project predicts customer churn for a bank using machine learning models.
The goal is to identify customers who are likely to leave so that the business can take proactive retention actions.

---

📊 Dataset

• Source: Churn Modelling Dataset
• Records: 10,000 customers
• Features: Credit score, age, balance, tenure, salary, geography, etc.
• Target: Exited (1 = churn, 0 = not churn)

---

🛠️ Tools & Technologies

• Python
• Pandas, NumPy
• Matplotlib, Seaborn
• Scikit-learn

---

🤖 Machine Learning Models Used

• Random Forest Classifier (Best Model)
• Logistic Regression
• Support Vector Machine (SVM)
• K-Nearest Neighbors (KNN)
• Gradient Boosting

---

📈 Model Performance

Model	Accuracy
Random Forest	86.5%
Gradient Boosting	86.7%
KNN	83%
Logistic Regression	81%
SVM	80%

---

🏆 Best Model

• Random Forest Classifier
• Accuracy: 86.5%
• Precision (Churn): 0.75
• Recall (Churn): 0.47

---

📊 Key Insights

• Model performs better on non-churn customers
• Struggles to detect churn cases (class imbalance issue)
• Feature engineering improved model performance

---

⚠️ Limitations

• Class imbalance affects recall
• Some models underperform on minority class

---

🚀 Future Improvements

• Apply SMOTE for imbalance handling
• Use XGBoost / LightGBM
• Hyperparameter tuning (GridSearchCV)
• Adjust decision threshold for better recall

---

📂 Files in this Repository

• churn_prediction.ipynb → Full analysis and model building
• churn_prediction.html → Report version of notebook
• Churn_Modelling.csv → Dataset

---

👨‍💻 Author

Sahel Ahmed