A Machine Learning Algorithm for Money Laundering Detection in Bank Melli Iran

In the study, different feature selection methods were initially studied to prevent and detect money laundering, and then a new method was developed and used in three stages for the selection of features effective in detecting money laundering using a cellular learning automata-based algorithm. In the first stage, the patterns were extracted using paired features through a complete graph. In the second stage, the extracted patterns were trained and classified on the basis of the impact rate of features using the cellular learning automata (CLA). Finally, in the third stage, the optimized feature was selected based on the impact rate of features. Selection of effective features using the proposed method improved the accuracy of data classification to detect money laundering. The Bank Melli Iran data set was utilized by entering into MATLAB to evaluate the proposed method and compare it with other methods. The results showed that the accuracy rate of classification in the proposed CLA method to detect money laundering was 94.19% and its runtime was 263.32 seconds. The proposed method was observed to have higher classification accuracy in detecting money laundering, as compared to the listed methods.