AI Driven Framework for Personalized Diabetes Assessment using Multimodal Clinical Data

Abstract

This study presents a machine learning-based personalized diabetes assessment framework using multimodal clinical data including glucose level, BMI, insulin, blood pressure, age, and skin thickness. The dataset undergoes structured preprocessing steps such as mean imputation, outlier removal using the Interquartile Range (IQR) method, and Min–Max scaling. Logistic Regression, Linear Regression, and K-Nearest Neighbors (KNN) were implemented and evaluated. The KNN model achieved an accuracy of 86.45%, F1-score of 84.27%, and ROC-AUC of 86.13%, demonstrating balanced predictive capability. Personalized recommendations are generated based on key feature influences. This work closes gaps in existing research by integrating multimodal indicators, comparing multiple machine learning models, and delivering personalized insights for early diabetes detection.

Ähnliche Arbeiten