Research On Knowledge Tracking Model In Intelligent Education System

With the vigorous development of big data and information technology,millions of students into intelligent learning online courses in education system,how to realize the personalized and intelligent education platform,how to scientifically,targeted to each student's knowledge state tracking effectively,how to find the students' knowledge breakpoints and weakness,how to tailor a personalized learning path for students,it is a hot research topic in the field of current knowledge to track.Although a lot of knowledge tracking models have been proposed,it is not possible to realize a personalized education platform in the real sense.The challenges in the knowledge tracking field are as follows:(1)none of the existing knowledge tracking models can dynamically model a student who acquires knowledge at different speeds.(2)in the actual complex educational environment,practice questions may appear several times in succession,which is not taken into account in the current knowledge tracking model.(3)there are a large number of practice questions on the platform,so it is almost impossible for students to finish all the practice questions within a certain time.It is very important to recommend a personalized sequence for students.However,until now,all knowledge tracking models have been unable to automatically find the dependencies between practice questions,which were previously correlated by experts.In order to solve the difficulties faced by the current knowledge tracking model,the research content and main contributions of this paper are as follows:(1)from the perspective of students obtain knowledge speed,knowledge of the current most advanced tracking model(Dynamic key-value memory network)was improved,based on mastering the knowledge of the speed tracking model is put forward,the model when calculating the delete vector and vector,consider the content of the memory of the current,and using distributed in external memory matrix.(2)in the real data set,students may encounter repeated exercises continuously.When no new exercises are encountered,it is not necessary for the model to carry out reading and writing operations on each time step.Because students repeat the exercises,they are read and written to the same memory.In view of this phenomenon,the first model proposed in this paper is improved by using the idea of hierarchical cyclic neural network,and a knowledge tracking model based on long-term dependence is proposed,which not only improves the prediction effect but also can remember students' long-term behaviors.(3)put forward independent reading knowledge dependency tracking model for a long time,the model of creative each exercises with two relative weights,a weight is used to read operation,another to write operation,let exercises read and write different fragments of memory,so it can automatically discover dependencies between exercises.Finally,the effectiveness and feasibility of the three knowledge tracking models are evaluated on four real knowledge tracking data sets.In addition,it shows how the model in this paper finds similar practice questions and the dependencies between practice questions,and predicts students' performance in an end-to-end way,which cannot be achieved by the current knowledge tracking model.