Classification of Suspicious Financial Transactions using Light Gradient Boosting Machine Method (LGBM) based on Social Network Analysis (SNA) Indicators

Abstract

Money laundering is an act committed by individuals or a group to conceal or disguise the origin of wealth obtained from illegal activities into assets that appear to have been acquired through legal means. Generally, there are three money laundering processes: placement, layering, and integration. The complexity of these money laundering processes described above makes it difficult to trace suspicious financial transactions and identify the parties involved and which transactions are connected to the suspected money laundering network. To address this issue, Social Network Analysis (SNA) is implemented to generate SNA features. In the following stage, these SNA features are employed as indicators to detect suspicious financial activities. The gathered indicator data is utilized to build a classification model using the Light Gradient-Boosting Machine (LGBM) approach. The results of this study show that the model created using SNA and LGBM methods achieved an accuracy of 97%. The precision, recall, and F1-Score values for non-suspicious transaction data were 98%, 97%, and 97%, respectively, while for suspicious transaction data, they were 97%, 98%, and 97%, respectively. The achieved accuracy values were quite high indicating that the used approach was capable of effectively classifying suspicious financial activities. We believe that the findings of this study could be an alternative method for detecting suspicious financial transactions in order to avoid money laundering operations.