Fraud prediction model

Design of the predictive model of Web Frauds carried out by Bluetab for a leading financial institution in the Spanish market

It consists of developing a predictive algorithm capable of classifying browsing sessions according to their similarity to sessions in which a fraudulent transfer has occurred. The output of this algorithm will be a scoring (or a probability) that will allow ordering these sessions by their probability of being fraudulent.

For this, information from the management of alerts and claims of fraud in Remote Banking Transfers is used, as well as the daily information of the browsing sessions on the web and in the APP; therefore, the target audience or population on which this model will be executed is the one that has started a session by one means or another.

Our methodology is based on the construction of the target based on the crossing between fraud and navigation information (the most critical process of a predictive model), in the diagnosis of inconsistencies among fraudulent operations that do not have an associated session, outliers with some wrong record on date.