Quasi Opposition-based Learning in Beluga Whale Optimization Feature Selection Approach for Diabetes Prediction in IoT System

The prediction and classification of the Non-Communicable Disease (NCD) like diabetes is significant for enhancing the survival rates of the patients.An integration of Internet of Things (IoT) and Machine Learning (ML) approaches contributes an efficient approach for the early diagnosis and prediction of the NCD.However, relevant feature selection from the diabetes data and integrating it into structured form is challenging, so it leads poor prediction performance.Hence, this research proposes the Quasi Opposition-based Learning with Beluga Whale Optimization (QOBL-BWO) for the prediction and classification of diabetes.Through considering the QOBL, the BWO has potentially covers a wide range of feature subsets, enabling a more detailed exploration of the search space.Initially, data is collected by using Pima Indian Diabetes Dataset (PIDD) to validate the effectiveness and which is pre-processed by using data cleaning, min-max normalization and handling class imbalance problem using Synthetic Minority Over-Sampling Technique (SMOTE).Then, feature selection is done by using QOBL-BWO and it is predicted and classified by using the Hinge Loss Function based Support Vector Machine (HLF-SVM).The experimental results shows that the proposed QOBL-BWO approach attains better accuracy of 99.05%, precision of 98.56%, sensitivity of 97.45%, specificity of 97.39% and F1-score of 97.38% as compared to the existing methods like AdaBoost Ensemble Technique (AdaBoost-ET) and SVM.