Machine Learning Approach for Detecting Money Laundering Transactions

Money laundering is a widespread criminal activity that presents difficult detection and prevention concerns. Machine learning algorithms have grown into effective tools in the past few decades to help in the detection of money laundering activities. The purpose of this paper is to apply ensemble methods, XGBoost and Linear Regression, to the detection of money-laundering-related suspicious financial activities. To increase overall forecast accuracy, ensemble methods integrate the predictions of various models. A complex gradient-boosting technique called XGBoost has demonstrated promising results in several applications, including the detection of money laundering. Further research is necessary to determine its precise advantages over other ensemble techniques in money laundering detection. With the use of feature engineering and model interpretability, traditional algorithms like linear regression have also been employed to find instances of money laundering. They may, however, be less efficient in detecting complicated patterns and nonlinear correlations associated with sophisticated money laundering operations. This paper intends to fill in these gaps by comparing the performance and effectiveness of XGBoost and Linear Regression in detecting suspicious financial transactions. To train the models, a Synthetic Financial dataset of classified suspicious and non-suspicious transactions will be compiled. Following that, the models will be tested using a variety of performance metrics such as accuracy, precision, recall, F1 score, and AUC-ROC. The findings of this study will shed light on the advantages and disadvantages of ensemble approaches and traditional algorithms in the identification of money laundering. This study fosters an improved understanding of the best ways for detecting abnormal financial transactions, thereby supporting financial institutions and law enforcement authorities in their efforts to successfully combat money laundering.