PREDIKSI RISIKO GAGAL BAYAR PREMI MENGGUNAKAN ALGORITMA GRADIENT BOOSTING: STUDI TRAVEL INSURANCE PREDICTION

Qowiyyul Amin Siregar; Revi Meliyani; Hikmah Rahmah; M. Hamal Musito; Maria Amelia Claudia Secu

doi:10.64626/jukomtek.v3i2.565

Penulis

Qowiyyul Amin Siregar Universitas Mitra Bangsa
Revi Meliyani Universitas Mitra Bangsa
Hikmah Rahmah Universitas Mitra Bangsa
M. Hamal Musito Universitas Mitra Bangsa
Maria Amelia Claudia Secu Universitas Mitra Bangsa

DOI:

https://doi.org/10.64626/jukomtek.v3i2.565

Kata Kunci:

Gradient Boosting, Risiko Gagal Bayar, Machine Learning, Asuransi, Prediksi

Abstrak

Prediksi risiko gagal bayar premi merupakan salah satu aspek penting dalam pengelolaan risiko perusahaan asuransi. Ketepatan model prediksi memungkinkan perusahaan mengidentifikasi calon tertanggung yang berpotensi menunggak pembayaran premi sehingga langkah mitigasi dapat dilakukan sejak awal. Penelitian ini bertujuan menerapkan algoritma Gradient Boosting Classifier (GBC) dalam memprediksi risiko gagal bayar dengan menggunakan dataset publik Travel Insurance Prediction. Variabel target direlabel menjadi Default Risk sebagai representasi risiko gagal bayar. Proses penelitian meliputi pre-processing data, encoding variabel kategorik, penyeimbangan data dengan SMOTE, dan evaluasi model menggunakan metrik AUC, akurasi, precision-recall. Hasil penelitian menunjukkan bahwa Gradient Boosting menghasilkan performa terbaik dibandingkan Logistic Regression dan Random Forest, dengan nilai AUC tertinggi dan stabilitas prediksi yang baik pada data tidak seimbang. Penelitian ini memberikan kontribusi pada pengembangan model risiko berbasis machine learning di industri asuransi.

Referensi

Chen, T. and Guestrin, C. 2016 . XGBoost: A scalable tree boosting system, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 785–794.

Chawla, N.V., Bowyer, K.W., Hall, L.O. and Kegelmeyer, W.P. 2022 ‘SMOTE: Synthetic minority oversampling technique revisited’, ACM Transactions on Knowledge Discovery from Data 16(4). 1–14.

Feng, Y., Miao, J. and Wu, Y. 2020 . Credit risk prediction using machine learning models, Expert Systems with Applications, 162, pp. 1–12.

Fernández, A. et al. 2020 . Learning from imbalanced data sets, Springer Nature. 1–39.

Gao, L. and Yan, Y. 2021 . Insurance fraud detection using ensemble learning. Journal of Information Security and Applications 59. 1–10.

Henckaerts, R. et al. 2021 . Boosting insights into insurance tariff plans with tree-based machine learning methods. North American Actuarial Journal. 25(4), pp. 508–529.

Henckaerts, R. et al. 2022. Balancing accuracy and transparency in actuarial pricing with boosting models. Expert Systems with Applications 205. 1–13.

Khairuddin, N., Hassan, M., and Ali, N. 2023 . Performance analysis of boosting methods in financial risk prediction. Journal of Finance and Data Science 9(2). 134–142.

Kong, X. et al. 2023. A hybrid XGBoost approach for insurance risk prediction. Applied Intelligence 53. 2451–2465.

Liu, M., Li, C. and Zhang, W. 2021. Comparison of machine learning and statistical models for financial risk prediction. Information Sciences 546. 1–16.

Michel, C., Salhi, N. and Thomas, D. 2022. Risk scoring with interpretable machine learning for financial services. Decision Support Systems 160. 113766.

Naitzat, G., Levina, E. and Yuan, M. 2021. Boosting algorithms and neural networks for non-linear risk prediction. Machine Learning 110(6). 1475–1494.

Powers, M. and Wunsch, D. 2020 . Insurance risk modelling: Trends and opportunities with machine learning, Journal of Insurance Regulation 39(2). 45–62.

Sun, J. et al. 2022. Credit scoring using explainable boosting machines. Knowledge-Based Systems 239. 1–12.

Zhang, T. and Guo, Z. 2023 Explainable machine learning models for underwriting risk prediction. Artificial Intelligence Review 56(7). 6235–6258.

Pamungkas, MH. Rahmah, H. Aziezah, N. Widodo, B. Pengaruh Aspek Tampilan Terhadap Tingkat Kepercayaan Pengguna Robot Siram Otomatis (Rosio). TEKTONIK: Jurnal Ilmu Teknik 1 (2), 205-210