Artificial Intelligence and Blockchain-based Financial Fraud Detection

Abstract

In this paper, we present an artificial intelligence(AI) and blockchain-based credit card fraud detection system for detecting fraud transactions in the dataset. This dataset has 284808 rows P with 31 columns Q and 0.17 % fraud class data. In the data preprocessing step, clean the data and normalized the feature. For a select important feature, we use explainable artificial intelligence(XAI) to get the highest priority feature in the dataset. Long short-term memory (LSTM) is used to detect fraud in the system and gives better accuracy. Blockchain is a decentralized system to secure the transaction of the system using smart contracts and an InterPlanetary File System(IPFS). After all processes, the LSTM gives 99.8% accuracy with using XAI. Also, present the comparison between two LSTM results with and without using XAI. Then we save the non-fraud transaction data using smart contracts and blockchain. Finally, we conclude our proposed system architecture with the results. Keywords: Explainable artificial intelligence, credit card frauds, deep learning, long short-term memory, fraud classification