Accurate Diabetes Detection Using Neutrosophic and ML Approaches

Resumen

Type 2 diabetes diagnosis is physically more difficult because of inconsistent variables such as blood glucose levels (BGL), BMI, and family history. Threshold-based models usually confuse the model with ambiguous cases, while neural networks are good at resolving contradictions but make it difficult to interpret. Hence, the new hybrid diagnostic framework may offer an insight combining Neutrosophic and Plithogenic logic with machine learning (ML) techniques to help in diabetes risk assessment. Neutrosophic logic makes it easy to categorize clinical information as exactness (T), indeterminacy (I), and falsity (F), and also permits classifications as Over-Sets Under-Sets or Off-Sets. With extension to this, Combining Neutrosophic Logic and Plithogenic Logic uses the intensity of opposite clinical information in defining the risk limits and gives such information random utilities in terms of simple and natural numbers. Application of this joint model in the analysis of real diabetes cases enabled proper neutrosophic patient categorization and consequent plithogenic risk evaluation. Comparative study reveals 86% for classical and 79% for fuzzy systems, while the embedded proposed enhanced ML model associated with neutrosophics, has: 99% accurate policies with superior capabilities in handling grey and contrasting data. The Off-Set classifier provided increased generalization and robustness improving other consistent explanation model including the conventional and fuzzy logic model. This article shows that the fusion of Neutrosophic and Plithogenic paradigms with ML results in an intelligent and clearer alternative for making differentiated diagnoses within complicated medical narratives.