Framework Manajemen Risiko Fraud Pada Transaksi QRIS: Pendekatan Unsupervised Anomaly Detection Dengan Isolation Forest Di Industri Perbankan

Kamal, Ahmad Rijaluddin Nisbil (2026) Framework Manajemen Risiko Fraud Pada Transaksi QRIS: Pendekatan Unsupervised Anomaly Detection Dengan Isolation Forest Di Industri Perbankan. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6032241068-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Perkembangan QRIS sebagai instrumen pembayaran digital di Indonesia meningkatkan kebutuhan terhadap penguatan manajemen risiko fraud, terutama karena penyimpangan transaksi tidak selalu muncul sebagai pelanggaran aturan eksplisit, tetapi dapat teridentifikasi melalui perubahan pola perilaku merchant. Penelitian ini bertujuan mengintegrasikan hasil deteksi anomali menggunakan algoritma Isolation Forest ke dalam Framework Manajemen Risiko Fraud yang mengacu pada ISO 31000:2018, COSO Fraud Risk Management Program (2023), dan POJK No. 12 Tahun 2024, serta mengevaluasi kontribusinya terhadap prioritisasi risiko dan investigasi merchant. Penelitian menggunakan pendekatan unsupervised anomaly detection berbasis segmentation-aware Isolation Forest terhadap 10.052.810 transaksi dari 4.952 merchant QRIS pada MCC 5411 dan 5499 selama periode 26 Januari–26 Februari 2026. Evaluasi dilakukan menggunakan indikator risiko berbasis proxy, metrik pemeringkatan, validasi ahli, simulasi kapasitas investigasi, dan selective alerting. Hasil penelitian menunjukkan AUC sebesar 0,8794 dan Lift@Top5% sebesar 6,5942, serta hasil validasi ahli menunjukkan bahwa sebagian besar merchant dengan tingkat anomali tertinggi dikategorikan suspicious atau memerlukan monitoring lanjutan. Skor anomali digunakan sebagai Relative Likelihood Proxy dan dikombinasikan dengan consequence berbasis potensi eksposur transaksi untuk menghasilkan final risk rating. Hasil penelitian menunjukkan bahwa pendekatan yang diusulkan dapat mendukung prioritisasi asesmen dan investigasi merchant secara lebih terarah serta menghubungkan hasil deteksi anomali dengan proses identifikasi, evaluasi, treatment, monitoring, dan pelaporan risiko. Dengan demikian, framework yang dikembangkan berfungsi sebagai decision-support berbasis risiko dan tidak digunakan untuk menetapkan merchant sebagai pelaku fraud secara langsung.
=================================================================================================================================
The rapid growth of QRIS as a digital payment instrument in Indonesia has increased the need to strengthen fraud risk management, particularly because transactional irregularities do not always manifest as explicit rule violations but may instead be identified through changes in merchant behavioral patterns. This study aims to integrate anomaly detection results generated using the Isolation Forest algorithm into a Fraud Risk Management Framework with reference to ISO 31000:2018, the COSO Fraud Risk Management Program (2023), and POJK No. 12 of 2024, as well as to evaluate its contribution to merchant risk and investigation prioritization. The study applies a segmentation-aware Isolation Forest approach for unsupervised anomaly detection to 10,052,810 transactions from 4,952 QRIS merchants classified under MCC 5411 and MCC 5499 during the period from January 26 to February 26, 2026. The model is evaluated using proxy-based risk indicators, ranking metrics, expert validation, investigation capacity simulation, and selective alerting. The results show an AUC of 0.8794 and a Lift@Top5% of 6.5942, while expert validation indicates that the majority of merchants with the highest anomaly levels were classified as suspicious or requiring further monitoring. The anomaly level is subsequently used as a Relative Likelihood Proxy and combined with transaction exposure-based consequence to derive the final risk rating. The findings demonstrate that the proposed approach can support more targeted merchant assessment and investigation prioritization while linking anomaly detection results to risk identification, evaluation, treatment, monitoring, and reporting processes. Therefore, the developed framework serves as a risk-based decision-support mechanism and is not intended to directly classify merchants as perpetrators of fraud.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Deteksi Anomali, Isolation Forest, Manajemen Risiko Fraud, Prioritisasi Investigasi, QRIS, Anomaly Detection, Fraud Risk Management, Investigation Prioritization.
Subjects:	T Technology > T Technology (General) > T174.5 Technology--Risk assessment. T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Interdisciplinary School of Management and Technology (SIMT) > 61101-Master of Technology Management (MMT)
Depositing User:	Ahmad Rijaluddin Nisbil Kamal
Date Deposited:	29 Jul 2026 03:36
Last Modified:	29 Jul 2026 03:36
URI:	http://repository.its.ac.id/id/eprint/139513

Actions (login required)

View Item