Supervised Machine Learning Model in the University Business Consultancy

Abstract

Although progress has been made in university business consultancies aimed at supporting the management of micro and small enterprises (MSEs), there remains a need to strengthen decision-making processes, which in many cases still rely on the experience and intuition of advisors. The purpose of this study was to evaluate a boosting classification model capable of predicting the social stratum of users served by the University Business Consultancy. A total of 570 users from a public university in Colombia participated in the research. The study followed an applied research design with a quantitative approach, employing statistical and machine learning techniques within a non-experimental, cross-sectional, and correlational framework. The results reveal an unequal distribution among social stratum classes, affecting the uniform representation of the dataset. The ROC curves corresponding to the main classes show an area under the curve (AUC) greater than 0.80, whereas in minority strata the AUC is below 0.70, indicating the need for model optimization. It is concluded that the use of quantitative metrics provides a comprehensive understanding of model performance, facilitates evidence-based decision-making, and offers valuable insights for designing differentiated strategies according to the social stratum of users in the University Business Consultancy.