Optimasi Hyperparameter Algoritma Support Vector Machine dalam Klasifikasi Penyakit Beta-Thalassemia

Abstract

Beta-thalassemia, a type of thalassemia disease caused by genetic variations, forces sufferers to receive regular blood transfusions for survival. Therefore, classification of this disease is important to reduce the number of births of beta-thalassemia patients in the future. 5066 beta-thalassemia carrier patient data from the Punjab Thalassemia Prevention and Program (PTPP) case study was taken as the source in this study which was accessed through the github.com website. Preprocessing is done for class alignment to avoid data imbalance, feature selection to streamline model performance, and normalization of feature values to a certain scale. In this research, the main focus is on applying the performance of the Support Vector Machine (SVM) algorithm to obtain classification results. Before entering the final model, hyperparameter tuning is required to obtain suitable parameter values to be entered into the model. hyperparameter tuning that will be carried out include "C" (regularization parameter), "kernel" (kernel type), "gamma" (kernel parameter for non-linear kernels), and "degree" (polynomial degree for polynomial kernels), carried out before the model is evaluated. The accuracy results were evaluated using confusion matrix, resulting in precision of 99.24%, recall of 99.62%, f1-score of 99.43%, and accuracy of 99.42% after hyperparameter tuning where the best parameters are "'C': 1, 'gamma': 100, 'kernel': 'rbf'" with an average test score of 0.993494149.