Efficient Machine Learning Pipeline Automation Using Tpot And Pycaret

Abstract

The paper falls under the domain of Automated Machine Learning (AutoML) and Data Preprocessing. The existing system employs the Tree-based Pipeline Optimization Tool (TPOT), a Python-based AutoML framework that automates tasks such as algorithm selection, hyperparameter tuning, and data preprocessing, primarily for regression tasks. However, its functionality is limited to regression-based optimization. To overcome this limitation, the proposed work integrates PyCaret, a more versatile AutoML library that supports both classification and regression. PyCaret enhances the machine learning pipeline with robust preprocessing features, including feature engineering, error handling, and class imbalance management. It enables users to train multiple models and automatically selects the best-performing one, streamlining the entire workflow and making machine learning more accessible. The system achieves an impressive accuracy of 97.8%. Future work may include extending support to unsupervised and deep learning tasks, integrating cloud-based scalability, and incorporating real-time data processing for broader applicability.