Semi-Automated Diabetes Prediction Using AutoGluon and TabPFN Models

Since the previous decade, the number of diabetic patients has increased significantly, which raises the risk of additional complications such as heart attacks, renal failure, decreased vision, and nerve damage. If this illness is detected and treated in its early stage, patients can be saved from a life-threatening disease. The discipline of artificial intelligence (AI), which is rapidly expanding, has possibilities that could revolutionize how this serious ailment is diagnosed and managed. At present, AI can only forecast diabetes using manually entered data. This paper proposes a semi-automated AI model that allows us to measure in automated and manual ways. Additionally, the two distinct AI models, AutoGluon and TabPFN, are employed to train AI models and are assessed using statistical metrics. Additionally, it also compares with the four traditional models w.r.t evaluation performance. Feature importance is then utilized to determine which feature is more advantageous.