Google Vertex AI Automl Application for Early Diabetes Risk Prediction

Abstract

AutoML, or Automated Machine Learning, has become an essential approach for simplifying the design and implementation of machine learning models. By automating model selection, hyperparameter tuning, and evaluation, AutoML allows researchers and practitioners to focus on problem formulation and data quality, rather than algorithmic details. This study examines Google Vertex AI as a cloud-based AutoML platform that integrates model preparation, training, evaluation, and deployment in a single environment. The study covers the fundamental requirements for using Vertex AI, such as account setup, data formats, and computing resources, as well as its benefits and limitations compared to other AutoML frameworks. A case study of diabetes risk prediction using tabular clinical data is presented, achieving an AUC-ROC of 0.884 and a PR-AUC of 0.894, demonstrating the applicability of Vertex AI AutoML in real-world medical settings. The results highlight the accessibility and versatility of Vertex AI, while acknowledging the constraints associated with cost and dependence on internet connectivity.

Ähnliche Arbeiten