Machine Learning in FinTech for Financial Fraud Data Detection
TL;DR
This research takes advantage of multiple machine learning techniques, such as k-nearest neighbors (kNN) and decision trees, by utilizing data fusion to detect financial fraud accurately, and shows that the proposed method outperforms each individual method.
Abstract
Financial fraud keeps rising these days. Organizations attempt to stop this trend by using various methods, such as distributing guides on how to avoid scams and frauds and automatically generating alerts when suspicious activities occur. However, this passive approach does not mitigate the problem, as the trend is worsening, and it is usually too late when victims realize they have been scammed. Therefore, active approaches must be employed before scams reach the victims. A wide variety of preventive methods, such as neural networks and data mining, have been used to detect financial fraud data, but none have proven entirely effective in combating scams. Each method has its pros and cons. This research takes advantage of multiple machine learning techniques, such as k-nearest neighbors (kNN) and decision trees, by utilizing data fusion to detect financial fraud accurately. The data fusion function used here is self-adjusting through learning. During the training phase, the system is repeatedly applied to the dataset until an optimal detection rate is achieved. Experimental results from credit card transactions show that the proposed method outperforms each individual method. Parameter or threshold values for the data fusion are set heuristically. Future research will focus on developing reconfigurable data fusion by automatically adjusting the values.