Autoencoders for Anomaly Detection Based on Isolation Forest Algorithm

With the increased usage of credit cards in this, credit card fraud and the loss to credit card users have increased. Supervised learning has is widely been used to discover anomalies in records of credit card transactions under the presumption that the pattern of fraud would depend on past transactions. Unsupervised learning does not undermine the fact that fraudsters change their approaches from time to time according to customers’ behaviors and patterns. Anomaly detection has become the most critical aspects within several applications, involving the security of networks, fraud detection, and quality control. Several conventional methods failed to find an anomaly in nature accurately within these complex and high-dimensional data. The present study is aimed at finding a better approach in such anomaly detection by integrating the autoencoder neural network with the Isolation Forest algorithm.