Artificial Intelligence for Fraud Detection and Criminal Liability in Indonesian Cooperatives based Deep Neural Network Approach

Abstract

Cooperatives are legally recognized entities that play a vital role in Indonesia's national economic development by promoting mutual assistance and collective welfare in accordance with the country's socio-cultural values. However, the rapid growth of cooperatives has also been accompanied by increasing legal challenges, particularly fraudulent activities committed by cooperative executives and managers, especially within savings and loan cooperatives. This study aims to identify the predominant forms of fraud in cooperatives and to analyze the criminal liability of perpetrators from both legal and technological perspectives. A Deep Neural Network (DNN) model was developed to classify potential fraudulent criminal acts using data collected from 275 respondents. Experimental results demonstrate that the DNN model achieved its highest classification accuracy of 98.5% after 140 training epochs, while a model trained for 40 epochs attained an accuracy of 93.9%. Furthermore, incorporating additional input features reduced the classification error to 0.188%, highlighting the effectiveness of deep learning in recognizing complex fraud patterns within cooperative institutions. The findings underscore the importance of strengthening internal governance, enhancing member-based oversight, reinforcing governmental supervision, and implementing more robust legal regulations and criminal sanctions against cooperative officials who abuse their authority. The integration of artificial intelligence with legal oversight offers a promising strategy for improving fraud detection, supporting law enforcement, and preventing financial crimes within Indonesia's cooperative sector