Learning Path Generation Research Based On Knowledge Graph

With the rapid development of online education, more learners are willing to obtain knowledge through online education platform. However,some researchers have found that "Information Overload" and "Study Lost"are easily occurred during online learning. In course picking, it is very likely that the courses we pick contain a large scale of repetitive content,which is the major cause of the "Information Overload". When we finally pick up the courses we learn, still we don't know how to range a learning path, which is the major cause of the “Study Lost". In recent years,the rearch and application of knowledge graph has been developed rapidly. In the area of education, knowledge graph can be used in the representation of knowledge. Other features like the knowledge structure, the level of nodes and other information are also very important. With these background, this dissertation presents the following research objectives:research on the generation of learning path technology based on the knowledge graph, in order to solve the "Information Overload" and the"Study lost" problem.This dissertation will discuss the course selection and course sequencing in the construction of learning path through concrete experiments, and point out that the existing methods are too dependent on the data generated by the learners. This dissertation will design a knowledge structure for online courses based on knowledge graph, the course selection and course sequencing problems will be solved through a set of features.This dissertation built a knowledge graph with 23 IT knowledge structure from StuQ, through the experiment of the research method, we draw the following conclusion: Using the vector space model to establishthe connection between the course and the knowledge points can decompose the knowledge structure effectively, establish connections between courses, the feature of knowledge can be easily transferred into the feature of courses. The coverness, superclass, overlap of the knowledge is the key to course selection, The appropriate weight of these three features can enable learners to learn fewer courses, achieve higher learning effects,and solve the learner information overload problem . The use of genetic algorithm can quickly converge a global optimal course in the course sequencing based on course importance, basicity and relevance, which can guide the learners to learn the whole course collection and solve the problem of study lost. The learning path generation based on the association of courses can discover the knowledge structure of the course,visualize the whole learning path. The learning path graph generation based on Bayesian statistics uses the learning path constructed by experts, and the construction result can be simulated just like the way experts build the learning path. When we use the mix way to build the learning path, it can not only simulate the way experts construct the learning path, but also can show the knowledge structure of the courses, and have a better understanding of the learning path.