Multiple classifiers system for anomaly detection
The two major concerns of using credit cards are (i) fraud and (ii) default payment. Fraud and default payment may cause financial losses to both credit cardholders and banks. Thus, many researchers are trying to find various effective ways of addressing these two concerns by using data mining appro...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Published: |
2020
|
Subjects: | |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
id |
my-mmu-ep.11360 |
---|---|
record_format |
uketd_dc |
spelling |
my-mmu-ep.113602023-04-18T08:18:33Z Multiple classifiers system for anomaly detection 2020-10 Kalid, Suraya Nurain TK5101-6720 Telecommunication. Including telegraphy, telephone, radio, radar, television The two major concerns of using credit cards are (i) fraud and (ii) default payment. Fraud and default payment may cause financial losses to both credit cardholders and banks. Thus, many researchers are trying to find various effective ways of addressing these two concerns by using data mining approaches. However, dealing with credit card data sets is a challenging task for data mining researchers. It is challenging because credit card data sets generally exhibit the characteristics of (i) unbalanced class distribution, and (ii) overlapping class samples. Both characteristics generally cause low detection rates for frauds and default payments that are minorities in the data. On top of that, the weakness of general learning algorithms contributes to the difficulties of classifying these two minority classes as the algorithms generally bias towards the majority class samples. In this project, we proposed a Multiple Classifiers System (MCS) that can produce a more effective detection by employing the sequential combination strategy. Sequential combination strategy is a process where two or more classifiers sequentially classify data. The proposed approach was tested on the credit card fraud data set (CCFD) and the credit card default payment data set (CCDP). The result shows that the proposed approach outperformed the current research work, particularly in detecting frauds and default payments (the minority classes) for CCFD and CCDP with True Positive Rate (TPR) of 0.872 and 0.840, respectively. 2020-10 Thesis http://shdl.mmu.edu.my/11360/ http://erep.mmu.edu.my/ masters Multimedia University Faculty of Computing and Informatics (FCI) EREP ID: 10285 |
institution |
Multimedia University |
collection |
MMU Institutional Repository |
topic |
TK5101-6720 Telecommunication Including telegraphy, telephone, radio, radar, television |
spellingShingle |
TK5101-6720 Telecommunication Including telegraphy, telephone, radio, radar, television Kalid, Suraya Nurain Multiple classifiers system for anomaly detection |
description |
The two major concerns of using credit cards are (i) fraud and (ii) default payment. Fraud and default payment may cause financial losses to both credit cardholders and banks. Thus, many researchers are trying to find various effective ways of addressing these two concerns by using data mining approaches. However, dealing with credit card data sets is a challenging task for data mining researchers. It is challenging because credit card data sets generally exhibit the characteristics of (i) unbalanced class distribution, and (ii) overlapping class samples. Both characteristics generally cause low detection rates for frauds and default payments that are minorities in the data. On top of that, the weakness of general learning algorithms contributes to the difficulties of classifying these two minority classes as the algorithms generally bias towards the majority class samples. In this project, we proposed a Multiple Classifiers System (MCS) that can produce a more effective detection by employing the sequential combination strategy. Sequential combination strategy is a process where two or more classifiers sequentially classify data. The proposed approach was tested on the credit card fraud data set (CCFD) and the credit card default payment data set (CCDP). The result shows that the proposed approach outperformed the current research work, particularly in detecting frauds and default payments (the minority classes) for CCFD and CCDP with True Positive Rate (TPR) of 0.872 and 0.840, respectively. |
format |
Thesis |
qualification_level |
Master's degree |
author |
Kalid, Suraya Nurain |
author_facet |
Kalid, Suraya Nurain |
author_sort |
Kalid, Suraya Nurain |
title |
Multiple classifiers system for anomaly detection |
title_short |
Multiple classifiers system for anomaly detection |
title_full |
Multiple classifiers system for anomaly detection |
title_fullStr |
Multiple classifiers system for anomaly detection |
title_full_unstemmed |
Multiple classifiers system for anomaly detection |
title_sort |
multiple classifiers system for anomaly detection |
granting_institution |
Multimedia University |
granting_department |
Faculty of Computing and Informatics (FCI) |
publishDate |
2020 |
_version_ |
1776101400534581248 |