ANALYSIS OF FACTORS AFFECTING CLIENT PROFITABILITY

This thesis investigates the effects of identifying and quantifying key financial and behavioral features on the prediction of Gross Income (GI) for micro and small segment enterprises at Raiffeisenbank. For this purpose, machine learning regression models are employed, utilizing client-level data on various financial activities, transactional behavior, and static attributes, along with the target variable of average gross income. The analysis focuses on identifying the most influential features and evaluating the predictive performance of different models. The three main findings are that (i) machine learning models can effectively predict a significant portion of the variability in GI, (ii) certain client financial and behavioral features, such as average asset and liability volumes, income, and international payment activity, emerge as strong positive predictors of GI, and (iii) a carefully selected subset of the most important features can achieve comparable or even slightly better predictive performance than models utilizing the entire feature set. This thesis represents an initial data-driven investigation into the drivers of Gross Income within Raiffeisenbank's micro and small enterprise client base using modern machine learning techniques.