Research On Learner Knowledge Tracking Method Based On Deep Learning

In recent years,education has become a hot topic for the whole people.In online education platforms,such as China university MOOC and little ape tutoring,learners can study without being limited in space and time.At the same time,they have access to better educational resources.However,there are some problems in online education.First of all,learners need to have good self-control,to ensure that they can complete the learning tasks in accordance with the course requirements and answer the after-class exercises with high quality.Secondly,because the online learning platform is unable to track learners' learning status in real time,there is a lack of personalized teaching recommendations and puzzles in teaching.Therefore,the loss of learners in the online education platform is serious.In order to solve the above problems,it is necessary to track learners' knowledge.Firstly,this paper introduces the research background and significance of knowledge tracing and relevant studies at home and abroad,and then introduces the Bayesian knowledge tracing model(BKT)and Deep knowledge tracing(DKT),which are mainly used in knowledge tracing.It also describes the deep learning techniques used in the deep knowledge tracking model,and explains the concepts of Recurrent Neural Network(RNN)and Long Short Term Memory Network(LSTM).Finally,it introduces how to improve the DKT model in this paper: the k-means algorithm is used to dynamically cluster the learner data input into the model,so as to extract the difference information contained in the learner data,in order to solve the problem that Deep knowledge tracing(DKT)ignores the difference of learners in the input data.Secondly,by optimizing the structure of the model,the Attention mechanism is introduced to select a large number of historical data in the process of model learning,so as to improve the DKT model structure.Through the experimental comparison,it is found that the operation effect of the improved model is better than that of the original DKT model.In addition,according to the experimental results,clustering the input data of the model,so as to get better hierarchical difference information,has a more obvious effect on improving the model.However,the time of model fitting can be reduced by optimizing the structure of the model.In addition,this paper found that combining model optimization and integrating learner difference information can not only reduce model fitting time but also further improve the running effect of the model.Compared with other DKT algorithms,the running effect of the model is improved.