Abstract

In Machine Learning, AutoML has advanced by automating the creation of better models and learning methods. This research takes it further, demonstrating that AutoML can automatically discover machine learning algorithms using basic mathematical operations. The novel framework minimizes human bias and leverages a generic search space. Despite this vast space, evolutionary search methods allow for discovering two-layer neural networks trained by backpropagation. The project focuses on developing an AutoML web application tailored for classification tasks, built using Python and leveraging pandas for data manipulation, streamlit for interactive web interfaces, and pyCaret for automating machine learning workflows.