Machine Learning-Based Trustworthy Diabetes Prediction Model

Abstract

Diabetes is a condition that develops when the blood sugar is above average level. Diabetes shoots up the risk of eye, kidney, nerve, and heart problems. By adopting suitable measures to prevent or manage blood sugar level may reduce the chances of getting diabetes-related health issues. Medical data collection is not done on a regular basis but is determined by the patient's condition and other clinical or administrative factors. A trustworthy model is free from missing values, faulty data and makes reliable predictions. Factors like data quality, anomalies, and model selection play a significant role in building a trustworthy model. Hence, this work concentrates on the prediction of diabetes by considering the following issues: 1) missing values in the dataset, 2) an imbalanced dataset and the presence of anomalies, and 3) diabetes prediction using Machine learning (ML) algorithm. The main objective of this work is to propose a trustworthy diabetes prediction model for an incomplete and imbalanced dataset using ML-based imputation, techniques for balancing datasets, and anomaly detection.