Regresar

A Predictive Model for the Detection of Clients Suspicious Behavior

Abstract:

The purpose of this work is to identify the suspicious behavior of the clients of a financial institution. Financial institutions use rule-based systems to detect unusual transactions. These systems focus on individual transactions or simple transaction patterns. Due to this, the need arises to detect suspicious behavior using machine learning since many of the machine learning algorithms are designed to capture complex patterns. Descriptive analysis and predictive analysis are used to detect suspicious behavior. Within the descriptive analysis, outliers are sought in transactional movements. For the predictive analysis, we start from a set of alerts with the label of whether or not they were reported by the Compliance Unit. As a result of the descriptive model, a set of customers that have behaved in an unusual way is obtained and as a result of the predictive model, the alerts that should be reported are predicted. It is concluded that the techniques: descriptive analysis, descriptive analysis and the rule-based system can complement each other, since they focus on different aspects of the identification of unusual transactions and should not be considered as exclusive alternatives.