ATM Transaction Status Anomaly Detection Based on Unsupervised Learning

Abstract

This paper uses information technology to monitor the data obtained by ATM equipment in real-time (three indicators of traffic volume, transaction success rate and transaction response time) and constructs an anomaly detection scheme for unsupervised learning (K-means clustering and SOM neural network). After that, the data in a day was simulated to consider the above schemes. Both schemes can make decisions quickly and have sound anomaly detection effects.