Weighting of the attribute importance weights using improved multiple linear regression and modified digital logic for process based fraud detection

Fraud detection becomes one of the solutions to overcome the fraud case that occurred in banks. In the Fraud detection process, each PBF (Process Based Fraud) attribute has different effects to indicate fraud. The MDL (Modified Digital Logic) method is used to weight the PBF attributes. MDL method produces the attribute’s importance weight that matches the impact of PBF attributes. However, the role of the expert is still very significant to assess the attribute’s importance weight. This study aims to modify the weighting procedure of the attribute’s importance weight in MDL method by adding Improved Multiple Linear Regression method (IMLR). By replacing the input previously given by the expert to the automatically weighting procedure. Then the results of both methods were evaluated using confusion matrix. Based on the experimental results, MLR method shows that the classification using all attributes importance weights has a better result with an accuracy of 99.5%.