Improving Retention of Stem Students Using Data Mining Design: A Quantitative Study

Abstract

The persistent difficulty of retaining college students through graduation has become a global problem. The purpose of this quantitative, descriptive, and retrospective study was to apply data mining methods, tools, and algorithms to analyze enrollment data for issues affecting STEM students’ retention at an historically black college (HBCU). The data source was a Johnson C. Smith University (JCSU) database containing demographic data, background data, commitment behavior, and social data. Students’ enrollment data from JCSU were a useful data source to identify students who would most probably leave the institution. Results showed data mining approaches explained and predicted retention of all STEM students at JCSU. Results confirmed findings reported in the research literature that multiple factors affect undergraduate retention at HBCUs. The study showed that first semester GPA, number of class hours, and students’ financial situations were important factors. Other factors such as lack of family emotional support and too many activities of social life might place students at risk of failing to complete their degree. The research also showed that institutional assistance in solving students’ personal problems needs to be moderate; otherwise the effort might produce negative results. A secondary purpose of the study was to demonstrate the appropriateness and efficiency of data mining methodology applied to STEM students’ retention. By providing insight into why students drop out before completing their degree, successful identification of students at risk could result in a program of directed retention intercession services..