Performance assessment of PLA-SVM: A novel Gurobi-enhanced piecewise linear approximation based approach for diabetes prediction

Large population of the world suffers from Diabetes. The prevalence of diabetes is increasing internationally, which presents significant challenges for healthcare systems. Technology, in particular Machine Learning, is crucial to addressing these problems. This paper introduces a new technique called PLA-SVM for diabetes prediction and assesses its effectiveness in relation to various machine learning methods. The PIMA dataset, a well-liked dataset for predicting diabetes, is used in this work. Performance indicators such as F1 score, recall, accuracy, and precision are utilized to compare the performance of various algorithms. The proposed method has been effectively validated using the PIMA dataset, where PLA-SVM exhibits remarkable performance, achieving an accuracy of 88.8% and an exceptional F1 score of 85.71%. The proposed PLA-SVM outperforms SVM and Random Forest, which closely follow with accuracy rates of 82.4% and 78.2% respectively. These results underscore the superiority of PLA-SVM over traditional SVM and other currently employed methods in predicting diabetes, particularly in terms of both training speed and accuracy.