ANN and SVM algorithm in Divorce Predictor

Abstract

Classification is a technique used to predict group membership or label for data samples (instances). In order to predict the result, the classification algorithm processes the training set, which contains a set of attributes and corresponding results. One of these classification technique is implemented in order to predict divorce in Turkey. This research is executed by Yöntem, M. K. et al. in 2019. In this , M. K. concluded that the ANN algorithm combined with correlation-based feature selection has the best performance with an accuracy of 98.82% and Kappa value of 0.9765. Nevertheless, unlike any previous research, ANN is not considered very good in terms of the required training time. In several previous studies, it was also concluded that other classification algorithms, such as SVM, have better prediction accuracy compared to ANN. In this study, prediction accuracy and Kappa value between ANN and SVM algorithms are compared using the same dataset and feature selection as the research done by Yöntem, M. K., to ensure a fair comparison between both of the algorithms. The result obtained from comparing both algorithms is that the SVM algorithm performs better than ANN with an accuracy of 99.8235 and a Kappa value of 0.9964. The training time required by SVM is also better than the ANN training time.