Klasifikasi Berat Badan Lahir Rendah (BBLR) menggunakan Metode Support Vector Machine dengan Teknik SMOTE

Abstrak

One of the main causes of infant mortality is associated with an increase in the Neonatal Mortality Rate (AKN) with low birth weight (LBW). LBW needs to be identified and predicted to prevent death when knowing the risk of LBW. In this study, a classification system was built as the initial identification of LBW. The data used comes from medical record data for childbirth at the Ardimulyo Public Health Center, Malang Regency for the January-August 2021 period in the form of imbalanced class. In this study, the method used is the Support Vector Machine (SVM) by combining the Synthetic Minority Oversampling Technique (SMOTE) technique. The performance of the SVM method without the SMOTE technique and the SVM method with the SMOTE technique using a linear kernel and RBF are compared in this study. Tests were carried out using 3-fold cross validation on kernel and parameter testing to find the best method and independent data testing of all methods. to compare the two. Based on testing the evaluation results obtained are less than optimal because they get low results. By testing the 3-fold cross validation test, the best results are obtained on the RBF kernel with the parameter lamda of 0.1, gamma of 0.001, complexity of 20, maximum iteration of 100, and epsilon of 0.001. Meanwhile, the results of data testing show that the best method is the RBF kernel SVM method without using SMOTE, which results in accuracy of 0.75, precision of 0.5, recall of 0.2, and f-measure of 0.2857.