Background Accounting is an industry that the digital transformation process has significantly impacted, and the use of artificial intelligence (AI) is now considered one of the most crucial tools for enhancing fraud-detection capabilities. The ineffectiveness of older fraud-detection techniques is evident in their failure to curb complex schemes currently employed to manipulate accounting systems. Methods The machine learning models were compared with each other in the following aspects of the proposed framework: accuracy, F1-score, recall, and the errors committed. The artificial transaction data generated in this study to resemble actual financial transactions shows that using all four models provides optimal results for identifying fraudulent cases. Results Artificial Neural Networks (ANN) outperformed all algorithms in terms of accuracy with 99.19%, and the minimum error rate was 0.81%, as for the recall, whereas Random Forests was the best among all the algorithms, up to 98.38%, which makes it efficient for detecting fraud. The results obtained suggest that the proposed integrated AI-based framework yields better detection results than existing rule-based systems, as well as a decrease in the rate of false alarms. Conclusions The idea in this study is a great step ahead in the enhancement of accounting information systems, as it provides an efficient tool for minimizing fraudulent issues that affect financial institutions by automating the process of data analysis. .
Shaamood et al. (Thu,) studied this question.