Financial statement fraud detection in Indonesia listed companies using machine learning based on meta-heuristic optimization
Published in IEEE 2020 International Workshop on Big Data and Information Security (IWBIS), 2020
Financial statement is a critical document which form the basis of decisions of various stakeholders in the capital market. Ironically, the phenomenon of fraud on the company’s financial statements is not a practice that never happened. Data reported by ACFE in 2020 showed that financial statement fraud is the costliest category of occupational fraud with median loss of $954.000. This study utilizes several machine learning approaches based on meta-heuristic optimization to develop robust fraud prediction models in financial statements. Two classification methods were used, namely, Back Propagation Neural Networks and Support Vector Machines. The best classifier in this study is a Support Vector Machine, which parameters are optimized with Genetic Algorithm resulting in 96.15% accuracy.