Customer Segmentation Pipeline with MLflow

Developed an end-to-end Unsupervised Machine Learning pipeline for customer segmentation using Python and Scikit-learn. The project includes data cleaning, missing value imputation, feature scaling, and preprocessing before applying clustering algorithms.

Implemented both K-Means and DBSCAN clustering with automated hyperparameter tuning using Grid Search to identify the optimal model configuration. Evaluated clustering performance using Silhouette Score, Calinski-Harabasz Index, and Davies-Bouldin Index, while utilizing PCA for cluster visualization and interpretation.

Integrated MLflow for experiment tracking, enabling automatic logging of parameters, evaluation metrics, trained models, and visualizations to ensure reproducibility and efficient model management.

Key Features:

• End-to-end data preprocessing pipeline.

• K-Means and DBSCAN clustering.

• Automated hyperparameter optimization.

• Multiple clustering evaluation metrics.

• PCA-based visualization.

• MLflow experiment tracking and model logging.

• Clean, modular, and reproducible Python implementation.