AutoML for All: Democratizing Machine Learning Model Building with Minimal Code Interfaces

In the fast-moving area of artificial intelligence, AutoML has emerged as a disruptive concept aimed at eliminating the technical complexities associated with traditional machine learning workflows. This position paper reports on a full paper "AutoML for All: A Large-Scale Study into the Limitations and Potential of Automated Machine Learning" that discusses how low-code, user-friendly AutoML tools can empower non-experts- from educators and clinicians to domain experts - to build accurate predictive models without requiring expert knowledge of modeling methodology. We demonstrate how to use H2O. ai in a code-lite environment using the Student Performance Dataset from the UCI Machine Learning Repository. It involves automating critical steps, such as data pre-processing, feature engineering, model selection, hyperparameter tuning, and ensemble learning. The approach focuses on easy set up and reproducibility, by means of a mixture of regression algorithms like GBM, GLM, DRF, and stacked ensembles. Performance evaluation using metrics such as RMSE, MAE, MSE shows that ensemble models which are developed with low user intervention bring out the best in accuracy values. The results show how AutoML can build models of high accuracy and interpretability like conventional pipelines with a significant reduction in development time and complexity. By democratizing model construction, everyone can be included and participate in AI-powered problem solving across different disciplines. Our work underscores an applied dimension in educational analytics and the potential of AutoML as a mature, ethical, and democratizing technology in AI development.