An Effective Diabetes Mellitus Classification Model for Diabetes Predictions and Diagnosing Using Machine Learning Techniques

This paper applies onerous machine learning methods to enhance the precision and potency of diabetes mellitus (DM) analysis and classification. Then, it presents HOMED, a novel hybrid online model using Adaptive Principal Component Analysis (APCA) and Incremental Support Vector Machine (ISVM) algorithms to overcome typical overfitting, missing data imputation, and computational inefficiency problems experienced by traditional models. The model was evaluated on 768 records from the Pima Indian Diabetes Dataset (PIDD), and on 1,099 records on the Diabetes Treatment Dataset (DTD). When compared to well-known models such as Decision Trees (DT), Random Forests (RF), Naï ve Bayes (NB), and k-Nearest Neighbors (k-NN), HOMED yielded higher accuracy (80.34%), higher sensitivity (92.67%), and higher specificity (98.43%). The ability to deal with high dimensional data while preserving precision and data integrity provides evidence for its potential to be a robust data tool for real time medical diagnosis. HOMED takes advantage of machine learning by bringing it to the support system for making clinical decision, to make healthcare practices more scalable, and more adaptive to early disease detection and effective patient management. The findings can be key for policymakers and healthcare practitioners to understand the need for using the most advanced algorithms to support sustainable, data driven medical practices. This research speaks to the use of technology to improve care outcomes worldwide.