Hyperparameter Optimization on Support Vector Machine using Grid Search for Classifying Thalassemia Data

Thalassemia is a genetically inherited disorder that causes fewer hemoglobin and red blood cells in the body. Although thalassemia is not contagious, it cannot be cured and some types require a lifetime blood transfusion. However, the occurrence of thalassemia is prevented by increasing public knowledge and awareness followed by early detection with screening process. The purpose of this screening is to identify a carrier that does not appear to have symptoms through a number of examination procedures in the target population. The data used in this study were obtained from the Harapan Kita Children and Women's Hospital, Jakarta, Indonesia which consisted of 150 samples and 11 features. The Support Vector Machine (SVM) with hyperparameter optimization using Grid Search was proposed in this study to classify thalassemia data. Grid Search was used to optimize the C and gamma parameters on SVM with RBF kernel. The results of this study showed that the proposed method produced 100_% accuracy with 90% training data when the parameter value C = 428.13 and gamma = 0.0000183 using holdout validation. Furthermore, it produced 100% accuracy when C = 4832.93 and gamma = 0.0000183 using 10-fold cross validation. These results were a lot better compared to using the same RBF kernel on SVM but with default parameter values when C = 1 and gamma = scale, which only produced 73.33% and 57.14% accuracy with holdout and 10-fold cross validation respectively.