Analisis Risiko Pinjaman dengan Metode Support Vector Machine, Artificial Neural Network dan Naïve Bayes

Bandung Pernama (Universitas Kristen Satya Wacana)
Hindriyanto Dwi Purnomo (Universitas Kristen Satya Wacana)

Jurnal JTIK (Jurnal Teknologi Informasi dan Komunikasi)
Vol 7 No 1 (2023): JANUARY-MARCH 2023

Banking is an industrial institution that is influential in the economy of a country. Banks are engaged in finance that collect funds from the public in the form of deposits and distribute loans to the public. It is undeniable, that in making loans to the public, problems will inevitably arise, such as the borrower being late in making installment payments or misuse of funds for other purposes, the borrower failing to build his business, thereby hampering installment payments. In this study, we will predict loan risk using a machine learning approach using several methods such as Support Vector Machine (SVM), Artificial Neural Network (ANN), and Naive Bayes. From the results of research that has tested the three methods using cross validation, confusion matrix, and ROC curves, the Support Vector Machine (SVM) method, which is the method with the best results, is 92.0% accuracy, then the second method is Artificial Neural Network. (ANN) of 91.2% and the lowest accuracy is the Naïve Bayes method with an accuracy of 81.2%.