Significance of LASSO Regression as Feature Extraction Techniques in Diabetic Detection

Diabetes Mellitus (DM) is a global health concern, with early and accurate identification playing a crucial role in personalized treatment. In this study, we utilize Lasso regression-based feature extraction to reduce high-dimensional gene data (2870x10) to a more manageable (51x1) per patient. Different classifiers techniques include, Gaussian Mixture Model (GMM), Softmax Discriminant Classifier (SDC), and Support Vector Machine (SVM) with different kernel’s as linear, polynomial, and radial basis function (RBF) kernels, evaluate by using different performance metrics such as Acc, F1 score, MCC as Matthews Correlation Coefficient, ER as Error Rate, and the FM metric as Fowlkes-Mallows. Results show that the SVM-RBF Kernal classifier attains an accuracy of 92.86% and a low error rate of 7.14%, demonstrating the potential of our approach in accurately distinguishing the Non-Diabetic Patients (NDP) and Diabetic Patients (DP) individuals from source of microarray gene data.