Intelligent multi-agent learning system applying educational data mining

Abstract

In this paper, we present a methodology for personalizing learning in accordance with the needs of individual students by using an intelligent, multi-agent learning system and data mining. Learning personalization is implemented on the basis of several methods. The Felder and Silverman Learning Styles model is used to create student profiles, and the probabilistic suitability indexes are identified to interlink learning components (i.e., learning objects, learning activities and learning environments) with the learning styles of individual students. Other technologies, which were proposed for creating the learning system, are ontologies, recommender system, intelligent software agents and educational data mining/learning analytics. Personalized learning units are referred to here as learning units composed of the learning components that have the highest probabilistic suitability indexes for particular students. In the paper, first, a systematic review on the application of intelligent software agents in learning is performed using the Clarivate Analytics Web of Science database. Second, we present the methods for personalizing the intelligent technologies of learning application, which are used to create optimized learning units for individual students. The developed student profiles and personalized learning units are further corrected by applying the methods and tools of data mining. The model of an intelligent, multi-agent learning system, based on the application of the aforementioned technologies, is presented in more detail. The principal success factors of the proposed methodology are the pedagogically sound vocabularies of learning components, an expert evaluation of the learning components in terms of their suitability for particular students as well as the application of ontologies, recommender systems, intelligent software agents and data mining.