Predicting students' performance: Educational data mining approach
Year: 2013
Type: article
Abstract: Applying data mining on data gathered from educational environments is a new, growing research area also known as educational data mining (EDM). It is focused on developing models and methods for exploring data collected from educational environments. EDM considers different aspects of education: students, teachers, teaching materials, organization of classes in order to better understand and improve educational process. In this paper we use different data mining algorithms in order to find the best suited model for prediction of students' success at the end of their studies. These models are generated and evaluated on students' personal, high school, admission and first year grades data from Faculty of Organizational Sciences, University of Belgrade, who studied Information Systems and Technologies study program. Specifically, artificial neural networks, linear regression and support vector machines are applied on students' aforementioned data to generate the model, which can be used to predict the students' average grade at the end of their studies. Similarly, several attribute selection techniques are applied in order to identify which attributes contribute the most to prediction of students' performance. Experiments showed that genetic algorithm attribute weighting technique gave best results where absolute error for linear regression and support vector machines were 0.2528. Also, personal data does not influence the final grade average. On the other hand first year grades, except Economy course, admission and high school data are considered important.
Source: Inovacije u nastavi
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: diamond
OpenAlex ID: W2507354133
