Prediksi Z-Score sebagai Indikator Kestabilan Perbankan Konvensional di Indonesia Menggunakan XGBoost

Angela, Michelle (2026) Prediksi Z-Score sebagai Indikator Kestabilan Perbankan Konvensional di Indonesia Menggunakan XGBoost. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002221014-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Kestabilan sistem perbankan merupakan salah satu faktor penting yang memengaruhi stabilitas sistem keuangan, khususnya di negara berkembang seperti Indonesia. Kestabilan perbankan dalam penelitian ini diukur menggunakan Z-Score, yaitu indikator yang menggabungkan profitabilitas (ROA), rasio ekuitas terhadap aset, dan volatilitas ROA. Penelitian ini menggunakan data yang bersumber dari Otoritas Jasa Keuangan (OJK), Badan Pusat Statistik (BPS), dan Bank Indonesia (BI). Tujuan penelitian ini adalah mengevaluasi performa model XGBoost dalam memprediksi nilai Z-Score perbankan konvensional di Indonesia serta mengidentifikasi faktor internal dan eksternal yang paling memengaruhi hasil prediksi. Data penelitian terdiri atas kelompok Bank Umum Konvensional (BUK) dan Bank Perekonomian Rakyat (BPR). Pemilihan model dilakukan melalui serangkaian eksperimen menggunakan validasi deret waktu Time Series Cross-Validation pada hyperparameter tuning dan evaluasi kinerja model. Hasil penelitian menunjukkan bahwa model XGBoost dengan pendekatan Direct Forecasting mampu memprediksi nilai Z-Score pada kelompok BUK maupun BPR, di mana performa terbaik secara konsisten diperoleh pada horizon t + 1 dan cenderung menurun seiring bertambahnya horizon prediksi. Hasil interpretasi menggunakan SHAP menunjukkan bahwa variabel historis Z-Score merupakan faktor yang paling dominan pada horizon jangka pendek untuk kedua kelompok bank. Pada BUK, kontribusi variabel historis berangsur berkurang pada horizon yang lebih jauh dan beralih ke kombinasi variabel internal serta eksternal seperti total aset dan inflasi, sementara pada BPR dominasi variabel historis Z-Score dan total aset tetap bertahan di seluruh horizon. Hasil proyeksi enam bulan ke depan menunjukkan bahwa nilai Z-Score BUK berada dalam rentang stabil antara 169 sampai 175, sedangkan pada BPR nilai Z-Score menunjukkan pola kenaikan dengan rentang yang lebih lebar, yaitu antara 60 sampai 109, dengan tetap mempertimbangkan tingkat ketidakpastian model yang meningkat pada horizon yang lebih tinggi.
===================================================================================================================================
Banking system stability is one of the key factors influencing financial system stability, particularly in developing countries such as Indonesia. In this study, banking stability is measured using the Z-Score, an indicator that combines profitability (Return on Assets/ROA), the equity-to-assets ratio, and the volatility of ROA. The study utilizes data obtained from the Financial Services Authority (Otoritas Jasa Keuangan/OJK), Statistics Indonesia (Badan Pusat Statistik/BPS), and Bank Indonesia (BI). The objectives of this study are to evaluate the performance of the XGBoost model in forecasting the Z-Score of conventional banks in Indonesia and to identify the internal and external factors that most strongly influence the prediction results. The dataset consists of two groups of banks: Conventional Commercial Banks (Bank Umum Konvensional/BUK) and Rural Banks (Bank Perekonomian Rakyat/BPR). Model selection was conducted through a series of experiments using Time Series Cross-Validation for hyperparameter tuning and model performance evaluation. The results indicate that the XGBoost model with the Direct Forecasting approach is capable of predicting the Z-Score for both BUK and BPR. The best predictive performance was consistently achieved at the (t+1) forecast horizon, with performance gradually declining as the forecast horizon increased. SHAP-based interpretation reveals that the historical Z-Score is the most influential predictor for short-term forecasting in both banking groups. For BUK, the contribution of the historical Z-Score gradually decreases at longer forecast horizons, shifting toward a combination of internal and external variables, particularly total assets and inflation. In contrast, for BPR, the historical Z-Score and total assets remain the dominant predictors across all forecast horizons. The six-month forecasting results suggest that the Z-Score of BUK remains within a stable range of 169 to 175, whereas the Z-Score of BPR exhibits an upward trend with a wider range of 60 to 109, while accounting for the increasing level of model uncertainty at longer forecast horizons.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	XGBoost, Stabilitas Perbankan, Z-Score, SHAP, XGBoost, Banking Stability, Z-Score, SHAP
Subjects:	Q Science > QA Mathematics
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Michelle Angela
Date Deposited:	31 Jul 2026 03:22
Last Modified:	31 Jul 2026 03:22
URI:	http://repository.its.ac.id/id/eprint/138072

Actions (login required)

View Item