Using data mining techniques to predict students at risk of poor performance

Abstract

Abstract-The achievement of good honours in Undergraduate degrees is important in the context of Higher Education (HE), both for students and for the institutions that host them. In this paper, we look at whether data mining can be used to highlight performance problems early on and propose remedial actions. Furthermore, some of the methods may also form the basis for recommender systems that may guide students towards their module choices to increase their chances of a good outcome. We use data collected through the admission process and through the students' degrees. In this paper, we predict good honours outcomes based on data at admission and on the first year module results. To validate the proposed results, we evaluate data relating to students with different characteristics from different schools. The analysis is achieved by using historical data from the Data Warehouse of a specific University. The methods used, however, are fairly general and can be used in any HE institution. Our results highlight groups of students at considerable risk of obtaining poor outcomes. For example, using admissions and first year module performance data we can isolate groups for one of the studied schools in which only 24% of students achieve good honour degrees. Over 67% of all low achievers in the school can be identified within this group.

Toward an improved learning process: the relevance of ethnicity to data mining prediction of students’ performance

“…The study by Alharbi et al [38] carried out data mining analysis for early identification of at-risk students using the admission records and performance in their first year of study. The results of student-failure potential analysis create an opportunity for warning at-risk student of a potential failure early enough so that drastic intervention can be deployed [39].…”

Analysis and prediction of student placement for improving the education standards

“…The information was recovered from the data warehouse of a university, where data of the understudies and the results are gathered. Credits that identify with understudy demo designs and other preferences include [1]:…”

“…We expect to explore the accessible highlights that might be utilized for forecast, and in addition the kind of classifiers that may create the best outcomes. Understudies' scholarly achievement is evaluated based on the performance of the student in the exams conducted by the institutes or Universities [1]. In this paper, we propose a scheme where prediction of student final placement can be done based on the marks scored by them in the previous semesters.…”

Generalizing Predictive Models of Admission Test Success Based on Online Interactions

“…While this fact is often neglected, it is very important to obtain models that could scale beyond the course where they were developed to ensure the sustainability of prediction models [10]. For example, as it was mentioned before, predictive models can be used to identify learners at risk and to make impact on learning [11]. However, if models are only developed for a specific course and cannot be reused for other courses or different cohorts, then their applicability will be limited.…”