Deteksi Fraud pada Jaringan Telekomunikasi Tetap dengan Metoda Discretized Interpretable Multi Layer Perceptron

Susanto, Eri (2005) Deteksi Fraud pada Jaringan Telekomunikasi Tetap dengan Metoda Discretized Interpretable Multi Layer Perceptron. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 2203205008-Master_thesis.pdf Restricted to Repository staff only Download (25MB)

Abstract

Fraud dalam telekomunikasi dapat dianalisa dari data rincian percakapan sentral lokal. Deteksi fraud dilakukan dengan cara mengklasifikasi historis pola-pola panggilan berdasarkan pola-pola panggilan normal dan pola-pola kasus fraud yang telah terjadi. Hasil klasifikasi pola adalah fraudulent atau dugaan fraud. Metoda yang digunakan untuk klasifikasi adalah Discretized Interpretable Multi Layer Perceptron (DIMLP) dengan pelatihan terbimbing. Hasilnya dibandingkan dengan metoda Multi Layer Perceptron (MLP) dan analisis berdasar ambang batas. Hasil-hasil penelitian menunjukkan rata-rata sensitifitas DIMLP 76%, rata-rata sensitifitas MLP 67% dan sensitifitas analisis berdasar ambang batas 73%. Rata-rata suseptibilitas DIMLP 0,81%, rata-rata suseptibilitas MLP 0,53% dan suseptibilitas analisis berdasar ambang batas 2,1%. Hasil-hasil tersebut dicapai dari deteksi fraud pada kasus pemalsuan identitas pada saat mendaftar sebagai pelanggan.
===================================================================================================================================
Fraud in telecommunication can be analyzed from call data records output by local exchange. Fraud detection can be done by classifying historical call patterns based on normal call patterns and fraud call patterns that have happened. The result is fraudulent or fraud suspect. Method used for classification is Discretized Interpretable Multi Layer Perceptron (DIMLP) with supervised learning. The result compared with Multi Layer Perceptron (MLP) and threshold base analysis methods. Results achieved by the research are DIMLP sensitivity average 76%, MLP sensitivity average 67% and threshold base analysis sensitivity 73%. DIMLP susceptibility average 0.81%, MLP susceptibility average 0.53% and threshold base analysis susceptibility 2.1%. These results obtained from fraud detection in telecommunication on subscription fraud case.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	fraud, klasifikasi pola~ DIMLP,sensitifltas, suseptibilitas,.fraudulent , fraud, pattern classification, DIMLP,sensitivity, susceptibility, fraudulent.
Subjects:	C Auxiliary Sciences of History > CE Technical chronology. Calendar
Divisions:	Faculty of Industrial Technology > Electrical Engineering > 20101-(S2) Master Thesis
Depositing User:	magang .
Date Deposited:	30 Sep 2026 04:08
Last Modified:	30 Sep 2026 04:08
URI:	http://repository.its.ac.id/id/eprint/145073

Actions (login required)

View Item