Detection of fraudulent transactions in the securities market using machine learning methods

Abstract

Financial fraud is a serious problem and an increasing threat to the economy. It has serious consequences for the stability and sustainability of the economy. In this case financial institutions are forced to constantly improve their systems for detecting fraud and suspicious transactions. This article is devoted to the use of machine learning methods to identify suspicious transactions in money laundering and financing of terrorism, provide in the area of security markets. The purpose of the article is to study the effectiveness of various machine learning models in the analysis of transactions in the stock market to identify various typologies of money laundering. The article consistently analyzes such machine learning methods as K-nearest neighbors (K-nearest), Decision Trees (DT), Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), Naive Bayes Classifier (NB), Discriminant Linear Analysis (LDA). Considerable attention is paid to the process of preparing data for research, which includes several stages such as selecting operations, encoding parameters for loading into the model, normalizing the parameters and combining them in a specific format. Transactions made on the stock exchange in the anonymous trading mode and non-trading operations (crediting/debiting funds and securities to the brokerage and depositary accounts) were used as initial data. The best results were obtained using the SVM, LR and LDA models, but the combined model is the most effective.