Features of Low and Highly Susceptible Individuals in Retail Investment Fraud: A Machine Learning – Based Analysis

Abstract

Investment fraud/scam is defined as the intentional misinterpretation, concealment, or omission of facts regarding promised goods, services, or other expectations by putting funds into investments that are not real, unnecessary, never intended to be fulfilled, or intentionally distorted for the purpose of monetary gain. We present in this paper, an analysis of individuals' features/characteristics of those who are highly susceptible to retail investment scamming using machine learning (ML) methods. Purposive sampling is applied in data collection, asking only those who've at least experienced being scammed in a retail investment. Participants' demographic profile, emotional intelligence scores, personality traits scores and financial literacy levels are collected as parameters for the analysis. The data (N = 177) is first submitted to a Boruta algorithm for feature selection and out of nineteen (19) input features, only seven (7) features are confirmed to be important in determining low or high likelihood of susceptibility in retail investment scamming. Afterwards, a 2 - cluster solution is revealed using the $k$ - means clustering. Cluster 1 is composed of individuals having higher number of times being scammed - characterized by higher social class, higher income, higher emotional intelligence scores, higher levels of agreeableness, openness and extraversion, and lower financial knowledge. Cluster 2 is composed of individuals having lesser number of times being scammed - characterized by lower social class, lower income, lower emotional intelligence scores, lower levels of agreeableness, openness and extraversion and higher financial knowledge. Findings of this study may serve as basis for prevention, protection and enforcement against retail investment frauds.