COMPARATIVE ANALYSIS OF APPLYING LEARNING ANALYTICS TOOLS TO CREATE PERSONALISED LEARNERS’ PROFILES: ARTIFICIAL NEURAL NETWORKS VS CASE-BASED REASONING AND BAYESIAN NETWORKS

Many techniques have been developed for user profiling, which vary from statistical keyword analysis to social filtering algorithms and different machine learning techniques. In the paper, several methods dealing with learning analytics (LA) tools, namely, Artificial Neural Networks (ANN) and Case-Based Reasoning (CBR) together with Bayesian Networks are comparatively analysed to be applied to create personal learners’ profiles. Learning analytics is the measurement, collection, analysis and reporting of data about learners and their contexts, for purposes of understanding and optimising learning and the environments in which it occurs. A related field is educational data mining (EDM). First of all, the author performed systematic literature review on application of ANN, CBR and Bayesian Networks to create learners’ profiles in Clarivate Analytics (formerly Thomson Reuters) Web of Science database. After that, methodology of ANN, CBR and Bayesian Networks’ application to create personalised learners’ profiles are analysed. ANNs in e-learning are mostly being used to suggest learning material to students, and for this purpose supervised training method is employed. A wide array of psychological theories is being used as predictors of learning material suitability, all with demonstrable success. Second most promising area of research is prediction of student performance, which can be ultimately used to customise their learning experience. By comparing the predicted student psychological profile with the one determined by the psychological questionnaire the agent will be able to identify the students who have changed, irresponsibly filled out the questionnaire or are by their nature outliers. After identifying these students, the system can adapt to them by giving them an option to choose two alternative learning paths or by asking them to revisit the questionnaire. This paper also analyses a technique that integrates Case-Based Reasoning and Bayesian Networks to build learners’ profiles incrementally. CBR provides a mechanism to acquire knowledge about learners’ actions that are worth recording to determine his habits and preferences. Bayesian Networks provide a tool to model quantitative and qualitative relationships between items of interest. Both presented methods and their proper application in different pedagogical situations are helpful to enhance learning quality and effectiveness.