engleski

Comparison of Machine Learning Algorithms for Students Performance Prediction Based on LMS Data

Huge volumes of data created by students` activities on learning management systems (LMS) urged an opportunity to extract meaningful information from data. Development of data mining field yielded algorithms making possible to analyse data with the aim to improve quality of the educational processes. In this paper, we are comparing four data mining methods based on different machine learning algorithms to predict academic performance of IT students based on data about their activities at the LMS. Aim of the research were twofold: (i) to predict students' academic achievement and identify most important predictors of academic success, (ii) to compare different machine learning algorithms and identify which fits the best on LMS data. Research results indicated frequency of students` discussions as the most important predictor of success. Neural networks provided most accurate and reliable model.

LMS data ; data mining ; CRISP DM ; neural networks

nije evidentirano