🚀 Customer Churn Prediction – Model Training & Evaluation Pipeline

Welcome to the model training and evaluation phase of the Customer Churn Prediction project! This repo follows the data preprocessing pipeline from Customer Churn Prediction – EDA & Data Preprocessing Pipeline, where we prepared the data for churn modeling. Here, we focus on training machine learning models, evaluating their performance, and saving the trained models for future use.

🚀 This repo is updated weekly with:

Clean, progressive Jupyter notebooks
Raw & processed datasets
Practical steps using Python, pandas and scikit-learn
Real-world-style applied model Training & Evaluation for a customer churn analysis

📋 What's Inside?

This repo covers the complete model training and evaluation pipeline, built step-by-step:

Notebook	Description
`0_data_preparation.ipynb`	Preparing the data for model training and evaluation. It includes loading datasets and applying necessary transformations.
`1_base_model_training.ipynb`	Traning the base machine learning model for the analysis using Logistic regression, and plotting confusion_matrixes.
`2_kfold_validation.ipynb`	Performing K-Fold cross-validation to evaluate model performance, calculate metrics, and ensure generalization.
`3_multi_model_training.ipynb`	Training and evaluating multiple machine learning models to compare performance and select the best approach.
`4_hyperparameter_tuning.ipynb`	Optimizing model performance through hyperparameter tuning using search techniques to find the best parameter settings.
`5_threshhold_optimization.ipynb`	Adjusting the classification threshold to improve performance metrics and align predictions with specific objectives.

📁 Folder Structure:

📂 artifacts/ → Model training results, including training/test data (X, Y) saved as .npz files
📂 processed/ → Processed data used for model training
📂 raw/ → Raw input data and initial notebook for data preparation
📓 Notebooks → Notebooks to prepare data for training, testing and evaluation

🔧 Tools Used:

Python, Pandas, Scikit-learn
Matplotlib, Seaborn
NumPy
Jupyter Notebooks

🎯 Goals:

Train machine learning models on the churn prediction dataset
Evaluate models' performance using various metrics
Save and export model artifacts (X_train, X_test, Y_train, Y_test)
Provide a solid template for future machine learning projects

📌 Steps Followed from the Previous Repo

If you haven’t already gone through the Data preprocessing steps, make sure to check out the Customer Churn Prediction – EDA & Data Preprocessing Pipeline repo first. This repo focuses on preprocessing the data, including handling missing values, encoding features, and scaling the dataset, which are essential steps before model training.

🚀 Getting Started

To get started with this repo, clone the repository and install the required dependencies:

git clone https://github.com/deaneeth/churn-prediction-model-training.git
cd churn-prediction-model-training
pip install -r requirements.txt

🌟 Why You’ll Like It:

📚 Easy-to-follow structure for model building and evaluation
🧠 Consistent with the preprocessing steps from the previous repo
🧼 Learn how to build, evaluate, and save machine learning models in Python
💾 Continuous weekly updates with new models, techniques, and results

🤝 Contribute or Follow Along

This repo is updated weekly, with new models, evaluation metrics, and results. Star ⭐ the repo to stay updated, and fork 🍴 it to experiment with your own models. Contributions & feedback are always welcome — just make sure to check the contributing guidelines before submitting any pull requests.

👀 Want to continue building real-world models for churn prediction?

You're in the right place! Let's train some powerful models together and predict customer churn like a pro.

Created with ❤️ by deaneeth

deaneeth/churn-prediction-model-training