Pelawi, Deni Ekel Ramanda Sembiring (2025) Evaluasi Deteksi Transaksi Fraud Atm Pada Bank Xyz Dengan Menggunakan Rule Based Classifier. Masters thesis, Institut Teknologi Sepuluh Nopember.
![]() |
Text
6032221170-Master_Thesis.pdf Restricted to Repository staff only until 1 April 2027. Download (3MB) | Request a copy |
Abstract
Penerapan arsitektur Fraud Detection System (FDS) pada Bank XYZ masih mengandalkan analisis berdasarkan data transaksi harian yang telah dibukukan dan diproses setelah melewati proses akhir hari. FDS dirancang agar dapat mendeteksi dan merespons transaksi mencurigakan secara real-time. sehingga dapat mencegah kerugian finansial yang signifikan. FDS akan menganalisis pola transaksi yang telah digunakan dalam deteksi penipuan dan akan membandingkan dengan transaksi ATM yang sedang dijalankan sebelum dibukukan. Penelitian ini bertujuan untuk menerapkan dan mengevaluasi model pendeteksi transaksi fraud di ATM menggunakan Rule-Based Classifier dan membandingkan kinerjanya dengan model Decision Tree. Berdasarkan hasil evaluasi, Model Decision Tree menunjukkan akurasi yang sangat tinggi yaitu 98%, recall yang cukup baik (79%), presisi model sebesar 75%, dan nilai F1-Score sebesar 77% yang mengkonfirmasi model ini efisien dalam mengidentifikasi kecurangan secara garis besar, namun ada keseimbangan yang perlu ditingkatkan antara presisi dan recall untuk mengoptimalkan kinerja. Model Rule Base Classifier, menunjukkan ketepatan yang lebih baik yaitu 97%, yang mengindikasikan bahwa hampir semua transaksi fraud yang diidentifikasi oleh model sesuai. Namun, akurasi umum yang lebih rendah (60%) dan recall (60%) menunjukkan bahwa model memiliki potensi gagal dalam mendeteksi fraud. Nilai F1-Score sebesar 74% menunjukkan perlu dilakukan peningkatan kemampuan model untuk mendeteksi semua kasus kecurangan tanpa meningkatkan false positive.
==================================================================================================================================
The implementation of the Fraud Detection System (FDS) architecture at XYZ Bank still relies on analysis based on daily transaction data that has been posted and processed after the end of the day process. The FDS is designed to detect and respond to suspicious transactions in real-time, thereby preventing significant financial losses. The FDS will analyze transaction patterns used in fraud detection and compare them with ATM transactions being processed before posting. This study aims to develop and evaluate an ATM transaction fraud detection model using a Rule-Based Classifier and compare its performance with a Decision Tree model. Based on the evaluation results, the Decision Tree model shows a very high accuracy of 98%, a fairly good recall (79%), a model precision of 75%, and an F1- Score of 77% confirming that this model is efficient in broadly identifying fraud, but there is a balance that needs to be improved between precision and recall to optimize performance. The Rule-Based Classifier model, showing better precision at 97%, indicates that almost all fraud transactions identified by the model a2 re accurate. However, a lower overall accuracy (60%) and recall (60%) indicate that the model has the potential to fail in detecting fraud. An F1-Score of 74% indicates that improvements are needed in the model's ability to det3 ect all cases of fraud without increasing false positives.
Item Type: | Thesis (Masters) |
---|---|
Uncontrolled Keywords: | Fraud Detection, Rule-Based Classifier, Decision Tree, Evaluasi Model, Performa Model, Transaksi ATM, |
Subjects: | T Technology > T Technology (General) > T58.6 Management information systems 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: | Pelawi Deni Ekel Ramanda Sembiring |
Date Deposited: | 07 Feb 2025 06:55 |
Last Modified: | 07 Feb 2025 06:55 |
URI: | http://repository.its.ac.id/id/eprint/118550 |
Actions (login required)
![]() |
View Item |