Deteksi Process-Based Fraud Dalam Aplikasi Kredit

Huda, Solichul (2017) Deteksi Process-Based Fraud Dalam Aplikasi Kredit. Doctoral thesis, Institut Teknologi Sepuluh Nopember.

Text 5113301003-Disertation.pdf - Accepted Version Download (2MB)

Abstract

Fraud dalam bidang perbankan semakin banyak terjadi khususnya di aplikasi kredit. Umumnya fraud diketahui setelah terjadi pencairan kredit. Penelitian disertasi ini mengembangkan metode deteksi fraud dalam proses bisnis, sehingga pencairan kredit dapat dibatalkan. Fraud yang terjadi pada proses bisnis disebut Process-based Fraud (PBF). Kesulitan dalam mendeteksi PBF adalah mengidentifikasi indikator atau atribut fraud dan cara deteksi yang perlu dikembangkan. Oleh karena itu kontribusi disertasi ini adalah identifikasi indikator atau atribut PBF dan cara mendeteksi PBF yang lebih akurat. Untuk mengatasi perbedaan pandangan yang terjadi diantara ahli fraud terkait perbedaan dalam pembobotan PBF, disertasi ini mengembangkan metode Fuzzy Multi Attribute Decision Making (MADM) untuk mendeteksi PBF. Pemecahan kedua masalah tersebut dilakukan sebagai berikut, pertama menganalisis indikator atau atribut yang diidentifikasi dari proses yang melanggar Standard Operating Procedure (SOP). Metode analisis atribut yang digunakan antara lain analisis skip, analisis throughput time, analisis resource, analisis duty, analisis pattern, analisis decision, dan analisis event parallel. Selanjutnya, berdasar atas indikator/atribut tersebut ditentukan kreteria/attribute value dan bobot penting atribut. Selain kedua kreteria tersebut, penelitian disertasi ini juga memasukkan bobot kepatuhan pelaksana proses (originator) dalam menentukan bobot fraud. Penelitian disertasi ini dilakukan dengan tahapan sebagai berikut, pertama sistem menganalisis pelanggaran SOP, lalu attribute value ditetapkan berdasar pada pelanggaran dan bobot kepatuhan pelaksana proses. Kemudian, bobot penting atribut ditentukan menggunakan metode Modified Digital Logic(MDL). Selanjutnya bobot fraud ditentukan menggunakan metode decision vector, inferensi fuzzy dan MADM. Terakhir, bobot fraud ditetapkan sebagai fraud atau tidak berdasarkan threshold.iii Hasil penelitian disertasi ini kinerjanya dievaluasi menggunakan Receiver Operating Characteristic (ROC). Evaluasi metode dengan Fuzzy MADM tanpa model kepatuhan memperoleh akurasi, sensitivitas dan spesifisitas masing-masing 0.98, 0.77 dan 0.99. Sedangkan evaluasi metode deteksi PBF dengan mengintegrasikan bobot kepatuhan memperoleh akurasi, sensitivitas dan spesifisitas masing-masing 0.99, 1 dan 0.99
====================================================================================================================================
Fraud especially in credit applications is getting increasedly widespread. Generally fraud is known after draw done of the credit. This study attempt to detect fraud in processes, hence draw done of credit may be canceled. Fraud which occurs in a business process is called Process-based fraud (PBF). Difficulty in detecting fraud are indentification of indicators or attributes of fraud and detection methods which need to be developed. Therefore, the contribution of the dissertation are identification of indicators or attributes of PBF and method for detecting PBF more accurately. To evercome of weighting PBF by experts, the dissertation develop Fuzzy Multi Attribute Decision Making (MADM) for detecting PBF. Solving to these problem is done as follows, first, identify indicators or attributes which identified from processes deviate the Standard Operating Procedure (SOP). Skip analysis, throughput time analysis, resource analysis, duty analysis, pattern analysis, decision analysis, and event parallel analysis used to identify indicators or attributes. Base on indicator or attribut, the method decide criteria / attribute value and important weights of attributes. Further, the dissertation also include behavior of originator for weighting PBF. The dissertation is performed by stages as follows, first, the system analyzes a deviation of SOP. Based on deviation and behavior of originator, the attribute value is determined. Afterwards, the weight important attributes is decided using Modified Digital Logic method. Furthermore, using the decision vector, fuzzy inference and MADM, the weight of fraud is establised. Based on threshold, the weight of fraud is determined as fraud or not. Performance of the methods is evaluated using Receiver Operating Characteristic (ROC). Evaluation of the method resulted accuracy, sensitivity and specificity of 0.98, 0.77 and 0.99 respectively. Using behavior model for detecting PBF, the method can obtain accuracy, sensitivity and specificity of 0.99, 1 and 0.99 respectively

Item Type:	Thesis (Doctoral)
Additional Information:	RDIf 658.403 Hud d-1
Uncontrolled Keywords:	Process-based Fraud, bobot fraud, identifikasi fraud, deteksi fraud, fuzzy MADM, attribute value, bobot penting atribut, PBF rating, modified digital logic, suspicious fraud, bobot relasi, process mining, bobot kepatuhan.
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD30.23 Decision making. Business requirements analysis. T Technology > T Technology (General) > T58.62 Decision support systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55001-(S3) PhD Thesis
Depositing User:	Mr. Tondo Indra Nyata
Date Deposited:	21 Jun 2023 02:31
Last Modified:	21 Jun 2023 02:31
URI:	http://repository.its.ac.id/id/eprint/98158

Actions (login required)

View Item