Fraud Detection on Financial Statements Using Data Mining Techniques

This study explores the use of data mining methods to detect fraud for on e-ledgers through financial statements. For this purpose, data set were produced by rule-based control application using 72 sample e-ledger and error percentages were calculated and labeled. The financial statements created from the labeled e-ledgers were trained by different data mining methods on 9 distinguishing features. In the training process, Linear Regression, Artificial Neural Networks, K-Nearest Neighbor algorithm, Support Vector Machine, Decision Stump, M5P Tree, J48 Tree, Random Forest and Decision Table were used. The results obtained are compared and interpreted.