| The rapid development of the modern communication technology, Internet andmultimedia technology has enabled users to obtain unprecedented abundantinformation and learning resources very conveniently. But it also brought aboutproblems: how to support users finding knowledge and learning resources fromInternet so as to help them to complete their typical learning process? How can thelearners find appropriate learning resources suited to their learning states andknowledge level? Therefore, personalized course generation technology and methodis one of the main research topics in personalized learning. Since1980, there has beendone research on course generation in the field of artificial intelligent and education,intelligent tutoring, adaptive hypermedia, and web-based learning systems.This work presents the topic of course generation based on recommendationtechnology as well as course evolution based on genetic algorithms. The main focuslies on the design and implementation of personalized course generation and anevolution model. The course generation framework enables teachers to constructonline courses automatically according to their teaching goals. The course evolution isused to update learning content of the course according to the learners’ changinglearning states and the knowledge level during the whole learning process. In such asetting, the data collection and updating is very important. The evolution model isimplemented in two phases. The first phase takes place before learning a course, whenthe personalized course is generated from the knowledge base according to theteaching outline from teachers. The second phase takes place during the learningprocess. The learning content is dynamic updated and evolved based on geneticalgorithms, while the learning states are changed.In detail, this thesis makes the following contributions: 1. The thesis presents a personalized course generation and evolution model fornormal learning in the context of large scale learning. Combined with the theoryof knowledge maps, this paper describes the concept model of course generationand its formal description, and presents a personalized course generation andevolution model for normal learning with large numbers of online learners. Itrealizes personalized learning through a series of personalized generated courses.This model provides a good general-purpose and scalable framework model forpersonalized learning in large-scale online learning environment.2. The thesis introduces the concept of a layered recommendation system (LRS)based on multi-dimensional feature vectors to implement personalized coursegeneration model and algorithms. In this work, we present a personalized coursegeneration algorithm based on the multi-dimensional feature vectors (PCG-LRS)and hybrid applications by content-based recommendations and collaborativefiltering recommendation algorithm to generate personalized curriculums. Basedon this algorithm, we introduce the teaching outline as the basis of the initialgenerated course and the final learning goals. The knowledge base of the coursescan be constructed from the teaching outline. The initial personalized knowledgemodels of students are generated by pre-tests. These personalized knowledgemodels are the base of personalized course generation. This algorithm not onlyhelps teachers to develop the overall curriculum teaching plan and to generatethe curriculum automatically, but also meets the learning requirements of eachindividual student with different knowledge and abilities. Additionally, thelayered recommendation algorithm recommends content within a large-scaleknowledge base repository and resource base implement at different levels. Thepersonalized recommendation algorithm is divided into a number of steps, whichachieves an effective dimensionality reduction, reduces the amount ofcomputation, and improves the courses generated algorithm.3. The thesis presents an evolution of personalized courses based on geneticalgorithms (PCE-GA). The genetic algorithms are successfully applied in thedynamic update process of the course during the whole learning process. In order to converge to an optimal solution of the multi-objective problems, due todirected acyclic concept maps combined with a topological sorting algorithm, wepresent a layered topological sorting algorithm to generate an initial solution.Under this framework of this algorithm, the target user model updatesdynamically, and the courses evolve during the process. It provides a goodgeneral-purpose and scalable framework that addresses the personalizedcurriculum generation in an online learning environment.4. The thesis presents a personalized user profile model to better fit the learningrequirements in an online learning environment. A vector space model is appliedin the algorithm to construct a personalized user profile model based on thecurrent state of the knowledge map. Extended triple elements are introduced todescribe the personal characteristics of learners. Respectively, the level ofknowledge, ability level, and target characteristics measure the user’s personalitytraits and match the character vectors of the conceptual model and the learningobject model. The model plays a crucial role in the implementation of the abovealgorithms.5. Based on the above findings, a course generation and evolution system basedPCG-LRS and PCE-GA is implemented in the field of E-learning. For large-scaleonline teaching, it provides teachers the automatic course generation of aprocess-based teaching program, including the construction of knowledgestructures, where learning activities add structured courses generation and enableother functions. We apply a layered recommendation and genetic algorithms tocurriculum formation and evolution process, allowing students’ knowledge stateto evolve according to the changes resulting from the learning process. Itpromotes students’ interest in learning, and improves learning efficiency andeffectiveness.This paper is supported by the European Framework Project FP7:“ResponsiveOpen Learning Environment（ROLE）”Grant agreement no.231396... |
PDF Full Text Request