Knowledge-based Recommender System of Conceptual Learning in Science

With the advancement of technology and the development of IT, such as big data, artificial intelligence (AI), and optimization theory have triggered revolutions in many fields. Along with the massive digital dataset, it also gives the reform motivation to promote traditional teaching and learning. How to find the appropriate information that students are interested, valuable, and easy to be understood in the vast amount of information, and bring their creative ability, is still a difficult thing. A Recommender System (RS) for e-learning based on learners’ Science Knowledge (SK) is a powerful tool to solve such problems, which can guarantee the quality and efficiency of the science teaching and learning in the international context. It is one of science research directions with great research value. This paper discusses the overall framework design of a wisdom RS for conceptual learning (CL) in science, which analysis models are based on SK of the students, using a big data software platform. The research sample is made of totally 621 junior students who take the subject of an introductory in science concepts (SC) in Computer Studies of Program, Macao Polytechnic Institution, from 2006 to 2021 academic year. According to the students’ SK, their CL methods have been classified into six types of thinking modes, which has three different corresponding understanding levels for each mode. The appropriate recommended materials to the students who are interested in the information provided, such as, text-based teaching materials, computer assisted materials, simulation tools and game activities, can increases the motivations of the students to learn, considered their different characteristics and understanding. The performance of this RS has been tracked over a period of 16 years, which can effectively improve the personalized teaching quality in the area of computer science, since it may be useful in heightening students’ motivation and interest in CL in science. The recommendation algorithm presented in this paper may be applied for a similar fashion across different domains, topics and contexts.