Fraudulent Electronic transaction detection using KDA Model

TL;DR

This paper presents a KDA-based model for detecting fraudulent electronic transactions, achieving high detection rates through dynamic and offline analysis, and introduces software supporting real-time fraud detection.

Contribution

It proposes a novel dynamic KDA model and software system that improve fraud detection accuracy and coverage in electronic transactions.

Findings

Detects 68.75% of frauds online

Detects 81.25% of frauds offline

Provides software for dynamic fraud detection

Abstract

Clustering analysis and Datamining methodologies were applied to the problem of identifying illegal and fraud transactions. The researchers independently developed model and software using data provided by a bank and using Rapidminer modeling tool. The research objectives are to propose dynamic model and mechanism to cover fraud detection system limitations. KDA model as proposed model can detect 68.75% of fraudulent transactions with online dynamic modeling and 81.25% in offline mode and the Fraud Detection System & Decision Support System. Software propose a good supporting procedure to detect fraudulent transaction dynamically.