Machine Learning Models to Predict Students’ Study Path Selection

Abstract

Selecting a proper study path in higher education is a difficult task for many students. They either have a lack of knowledge on the study pathoffered or are unsure of their interest in the various options. The current educationalsetups enable us to collect valid and reliable data on student success andlearning behaviour. This study explores and solves the problem of what path toselect by proposing possible study paths with the help of machine learning algorithms.Learning analytics (LA) and educational data mining (EDM) are technologiesthat aid in the analysis of educational data. In this quantitative study,we applied a questionnaire to collect data from students at the Business InformationTechnology Department (Bite) at the Haaga-Helia University of AppliedScience. We managed to collect 101 samples from students during 2017–2018.We used various machine learning algorithms and prediction models to assessthe best approach for study path selection. We applied three performance scoresof accuracy, Cohen’s Kappa, and ROC curve to measure the accuracy of thealgorithm results. KNIME analytics was selected as a proper tool to pre-process,prepare, analyse, and model the data. The results indicate that Random Forest(94% accuracy) and Decision Tree (93% accuracy) are the best classificationmodels for students’ study path selection. The contribution of this study is foreducational data mining research to assess the comparison of various algorithms.Furthermore, this is a novel approach to predict students’ study path selection,which educational institutes should develop to assist students in their studypath selection