Use of Machine Learning to Measure the Influence of Behavioral and Personality Factors on Academic Performance of Higher Education Students

Quality of education and improvement of school achievement has been linked to students' cognitive, behavioral and personality trait factors. Several researchers have investigated the correlation between these factors and students' academic performance. Particularly, it is assumed that behavioral and personality factors, such as study habits and self-esteem, have a positive and high relationship with students' academic achievement. However, research studies have shown a weak and inconsistent linear correlation level. Hence, research about better representation methods is needed. In this article we present and discuss the results from two studies on the influence of study habits and self-esteem on the academic performance of 153 college freshman students. First, we analyzed the linear correlation between our target variables; similar to previous work, we found a weak positive correlation between academic performance and study habits (r=0.283) and self-esteem (r=0.214). In addition, multiple linear regression was used to explain the relationship between these variables; it was found that the independent variables only explain the academic performance in 6.18%. Second, we propose to use K-means, an unsupervised clustering algorithm, as a better method to explain the influence of behavioral and personality factors and students' academic performance. Through the use of this method: 1) a predictive model of the academic performance is proposed, 2) it was achieved a better representation about the influence among the target variables, and 3) a set of students' academic profiles was created: low, medium and high. We found that 80% of the students with a high level of self-esteem and study habits (high academic profile) obtained a good or outstanding academic performance; outperforming students within the medium and low academic profiles by a significantly margin.