No Detail Found!

Boruta feature selection and particle swarm optimization for fraud detection on pawn transaction

Author(s)
MASDRA WAHYU KUNCORO Computer Science Department, Master of Computer Science, Bina Nusantara University, Jakarta, 11480, Indonesia
SUHARJITO Computer Science Department, Master of Computer Science, Bina Nusantara University, Jakarta, 11480, Indonesia
Abstract
The pawn company is one of the largest financial service companies in Indonesia, pawning is become an alternative for Indonesia people to obtain credit other than Bank. The number of pawn transaction raises the potential for fraud action. The use of PSO has shown promising result for improving classification accuracy. This will be a problem if the dataset used has many attributes. Previous research on fraud detection is usually carried out on credit card transaction, there has not been any research on fraud detection in pawn transaction. This research proposes based on data mining model to combining Boruta Feature selection and Particle Swarm Optimization (PSO). For classification using Gradient Boosted Tree (GBT) and C5.0 to measure the level of accuracy. The research is comparing several classifications and to know the highest accuracy of some classification. Pawn transaction data has been taken from pawn company in Indonesia. There are 216 transactions in 2019 until 2020. Among them, 26 transactions detected as fraud and 191 are no fraud. The attributes used is 24, among other is name of the customer, address, type of work, age, address, loan destination, identity number, collateral category, estimated value of collateral, loan money, credit time, type of product, type of transaction, class of collateral, maximum loan money, weight of collateral, and others. The results indicate that the combination of C5.0 optimizing by PSO and Boruta feature selection gives the highest classification accuracy of 96.82% and the GBT optimizing by PSO and Boruta feature selection reaching accuracy 93.57%.
Publication Details
Page(s) 2578-2585
DOI DOI not available
Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 8, Year: 2022

Request the full text from PASTIC

Keywords
Particle Swarm Optimization, Fraud Detection, Gradient Boosted Tree, Boruta
0

Citations

0

Downloads

46

Views