Perbandingan Metode Random Forest Regression (RFR) dan Support Vector Regression (SVR) dalam Memprediksi Risiko Kredit pada Bank XYZ

Tualeka AC, Nur Aisha Al Zahra (2024) Perbandingan Metode Random Forest Regression (RFR) dan Support Vector Regression (SVR) dalam Memprediksi Risiko Kredit pada Bank XYZ. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006201075-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (4MB) | Request a copy

Abstract

Meningkatnya animo masyarakat terhadap pengajuan pinjaman kredit di lembaga keuangan telah diimbangi dengan peningkatan kredit bermasalah yang berpotensi menimbulkan kerugian dan mempengaruhi rasio Non-Performing Loan (NPL). Penelitian ini bertujuan untuk membandingkan efektivitas metode Random Forest Regression (RFR) dan Support Vector Regression (SVR) dalam memprediksi risiko kredit pada Bank XYZ. Pemilihan parameter dilakukan dengan metode Grid Search, sementara evaluasi model menggunakan metrik Mean Absolute Percent Error (MAPE) dan Mean Squared Error (MSE). Hasil prediksi risiko kredit menggunakan metode RFR menunjukkan akurasi tinggi pada data pelatihan dengan MAPE sebesar 0,0125%, namun performanya menurun pada data pengujian dengan MAPE sebesar 14,86% dan MSE sebesar 0,3766. Sebaliknya, hasil prediksi risiko kredit menggunakan metode SVR dengan beberapa kernel menunjukkan bahwa kernel RBF memberikan hasil terbaik dengan MAPE sebesar 11,63% dan MSE sebesar 0,2486, mengungguli kernel Linear, Polynomial, dan Sigmoid. Perbandingan metode RFR dan SVR dengan kernel RBF menunjukkan bahwa meskipun RFR menunjukkan akurasi sangat tinggi pada data pelatihan, performanya yang buruk pada data pengujian menunjukkan overfitting yang signifikan. Sebaliknya, SVR dengan kernel RBF menunjukkan kinerja yang konsisten baik pada data pelatihan maupun pengujian, dengan nilai MAPE dan MSE yang lebih rendah. Kesimpulan dari penelitian ini menunjukkan bahwa SVR dengan kernel RBF merupakan pilihan yang lebih baik untuk prediksi risiko kredit, memberikan keseimbangan terbaik antara akurasi prediksi dan kemampuan menjelaskan variasi dalam data. Penelitian ini diharapkan dapat mendorong perusahaan perbankan untuk meningkatkan kualitas kredit melalui manajemen risiko yang efektif dan pemanfaatan teknologi machine learning dalam memprediksi risiko kredit.
============================================================
The increasing public interest in applying for credit loans at financial institutions has been accompanied by a rise in problematic loans, which have the potential to cause losses and affect the Non-Performing Loan (NPL) ratio. This study aims to compare the effectiveness of the Random Forest Regression (RFR) and Support Vector Regression (SVR) methods in predicting credit risk at Bank XYZ. Parameter selection was conducted using the Grid Search method, while model evaluation utilized the Mean Absolute Percent Error (MAPE) and Mean Squared Error (MSE) metrics. The results of credit risk prediction using the RFR method showed high accuracy on the training data with a MAPE of 0,0125%, but its performance decreased on the test data with a MAPE of 14,86% and an MSE of 0,3766. Conversely, credit risk prediction results using the SVR method with various kernels indicated that the RBF kernel provided the best results with a MAPE of 11,63% and an MSE of 0,2486, outperforming the Linear, Polynomial, and Sigmoid kernels. Comparing the RFR and SVR methods with the RBF kernel shows that although RFR demonstrated very high accuracy on the training data, its poor performance on the test data indicates significant overfitting. In contrast, SVR with the RBF kernel showed consistent performance on both training and test data, with lower MAPE and MSE values. The conclusion of this study indicates that SVR with the RBF kernel is a better choice for credit risk prediction, providing the best balance between prediction accuracy and the ability to explain variations in the data. This research is expected to encourage banking companies to improve credit quality through effective risk management and the utilization of machine learning technology in predicting credit risk.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Credit Risk, Grid Search, Machine Learning, NPL, Random Forest Regression, Risiko Kredit, Support Vector Regression
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HA Statistics > HA31.3 Regression. Correlation H Social Sciences > HD Industries. Land use. Labor > HD61 Risk Management H Social Sciences > HG Finance > HG3751 Credit--Management. Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Mathematics, Computation, and Data Science > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Nur Aisha Al Zahra Tualeka AC
Date Deposited:	31 Jul 2024 20:31
Last Modified:	31 Jul 2024 20:31
URI:	http://repository.its.ac.id/id/eprint/110579

Actions (login required)

View Item