📉 Churn Forecasting Model using Supervised Learning.

This project is focused on predicting customer churn using a machine learning pipeline. Churn prediction helps telecom companies retain customers by proactively identifying those likely to leave the service.

🧠 Project Overview

The goal of this project is to:

• Understand customer behavior through data
• Perform feature engineering and preprocessing
• Train and evaluate multiple machine learning models
• Build a predictive system to classify whether a customer is likely to churn

---

📁 Dataset Overview

• The dataset contains customer details like tenure, monthly charges, total charges, and various service subscriptions.
• The Churn column is the target variable (Yes/No).

Key preprocessing steps:

• Removed irrelevant column (CustomerID)
• Handled missing values in TotalCharges
• Addressed class imbalance using SMOTE (Synthetic Minority Oversampling Technique)

---

🔍 Exploratory Data Analysis (EDA)

Numerical Features:

• Distribution and outliers analyzed using boxplots
• Correlation heatmap created to understand relationships

Categorical Features:

• Countplots used to understand distributions
• Label encoding applied for model compatibility

---

🛠️ Data Preprocessing

• Label Encoding of categorical features
• Train-test split for evaluation
• SMOTE applied to handle class imbalance

---

🤖 Model Training & Evaluation

Multiple classification models were trained:

• Logistic Regression
• Decision Tree
• Random Forest
• XGBoost

🔍 Key Finding:

Random Forest gave the best accuracy among all models with default parameters.

---

💾 Deployment Ready

• The best performing model is saved using joblib
• A predictive system is built to classify new customer data using the saved model

---

📌 Requirements

To run this notebook, install the following:

pip install pandas numpy seaborn matplotlib scikit-learn imbalanced-learn xgboost