Student Performance Patterns in Engineering at the University of Johannesburg: An Exploratory Data Analysis

Abstract

Globally, the increased demand for engineers is not matched by an increase in graduates. This is further exacerbated by the fact that student dropout rates in engineering are higher than in other disciplines. Understanding engineering students’ performance patterns and potential influences can lead to developing interventions to improve engineering students’ success. Recent advances in data science and educational data mining have made it possible to extract valuable information from historical data, which can supplement interventions. This study sought to extract insights and information from real-world data, analyse correlations in the dataset’s variables and better understand the influences of student performance. Exploratory data analysis was applied to the dataset to visualise the dataset and infer the correlations between variables provided in the dataset on student performance patterns. We used Python for data analysis and visualising the correlation between variables. The results show gender disparity in engineering enrollments, with only a quarter of female students enrolled. The study also indicates that the completion rates could be much higher. Another finding is that most students who drop out do so because of choosing the wrong qualifications. Furthermore, when comparing the percentages, female students performed slightly better than their male counterparts. The correlation analysis shows no relationships between gender, race, admission point score, mathematics marks and science marks with student performance in engineering. Understanding student performance patterns can reduce dropout rates by correctly advising students to enrol on the most suitable programmes, and aid support interventions are needed to improve student success in engineering.

Keywords

• Engineering education
• exploratory data analysis
• extended programmes
• mainstream programmes
• student performance