Money laundering control in Mexico: A risk management approach through regression trees (data mining)

Purpose: This paper is aimed at developing a regression tree model useful to quantify the Money Laundering (ML) risk associated to a customer profile and his contracted products (customer’s inherent risk). ML is a risk to which different entities are exposed, but mainly the financial ones because of the nature of their activity, so that they are legally obliged to have an appropriate methodology to analyze and assess such a risk. Design/methodology/approach: This paper uses the technique of regression trees to identify, measure and quantify the ML customer’s inherent risk. Findings: After classifying customers as high- or low-risk based on a probability threshold of 0.5, this study finds that customers with 56 months or more of seniority are more risky than those with less seniority; the variables “contracted product” and “customer seniority” are statistically significant; the variables origin, legal entity and economic activity are not statistically significant for classifying customers; institution collection, business products and individual product are the most risky; and the percentage of effectiveness, suggested by the decision tree technique, is around 89.5 per cent. Practical implications: In the daily practice of ML risk management, the two main issues to be considered are: 1) the knowledge of the customer, and 2) the detection of his inherent risk elements. Originality/value: Information from the customer portfolio and his transaction profile is analyzed through BigData and data mining.