📊 Customer Churn Prediction App

📌 Overview

This project is an interactive Streamlit web application that predicts whether a customer will churn or not using a Naive Bayes classification model.
The app allows users to:

• Review the dataset
• Adjust input features (Age, Tenure, Gender) via a sidebar
• See real-time prediction results and probabilities for churn

The machine learning model is trained using customer data and saved for deployment.
With this app, decision-makers can quickly identify customers at risk of leaving and take proactive measures.

---

📂 Repository Structure

• StreamlitApp.py → Main Streamlit application script for deployment.
• churn_dataset.xlsx → Customer churn dataset used for training and prediction.
• lab2(classification).ipynb → Jupyter Notebook containing the full data preprocessing, model training, evaluation, and saving steps.
• naivebayesClassifier.pkl → Pre-trained Naive Bayes model file saved using Joblib.
• requirements.txt → Required Python libraries for running the app.

---

🛠 Project Workflow

1️⃣ Data Preprocessing

• Loaded customer churn dataset (churn_dataset.xlsx) using Pandas.
• Checked data quality and handled categorical variables:
  • Converted "Sex" into numeric values (Male = 1, Female = 0).
  • Converted "Churn" target column into numeric values (Yes = 1, No = 0).

2️⃣ Model Training

• Selected Naive Bayes (GaussianNB) as the classification model.
• Split the dataset into training and testing sets using train_test_split.
• Trained the model on customer features (Age, Tenure, Sex) and target (Churn).
• Evaluated performance using accuracy score and confusion matrix.

3️⃣ Model Saving

• Saved the trained model using joblib.dump as naivebayesClassifier.pkl for later deployment.

4️⃣ Streamlit App Development

• Created an interactive Streamlit app to:
  • Display a preview of the dataset.
  • Accept user inputs for Age, Tenure, and Gender via sidebar controls.
  • Convert gender selection to numeric values for prediction.
  • Load the saved Naive Bayes model and make predictions.
  • Display prediction probabilities for both churn and non-churn cases.