Research On Forgeable Deep Knowledge Tracing Model Based On IRT

Knowledge tracing can realize the purpose of personalized learning by tracing the state of learners’ knowledge mastery.Since knowledge tracing plays an important role in realizing personalized learning and promoting intelligent education,establishing an effective,high-performance and interpretable knowledge tracing model has become an important research topic in the field of knowledge tracing.This paper mainly aims at the problems that most of the existing knowledge tracing models ignore the role of forgetting factors or forget modeling is not sound,poor interpretability and low accuracy.The key technologies of item response theory,feedforward neural network,memory enhancement neural network,residual learning network and long and short memory neural network are used to carry out research.The main research work is as follows:(1)In view of the fact that most of the existing knowledge tracing methods are not sound enough in modeling forgetting factors or ignore the learning ability of learners,this paper proposes a deep knowledge tracing model that integrates learning ability and forgetting behavior.Firstly,the model comprehensively considered learners’ forgetting behavior,embedded three kinds of forgetting factors into the input data,calculated the memory erasing vector and memory updating vector,and modeled the forgetting behavior.Then,according to the state of knowledge mastery of learners,the learning ability network layer is established.Then,on the basis of Dynamic Key-Value Memory Networks for Knowledge Tracing,a deep knowledge tracing model combining forgetting behavior and learning ability was established Long Short-Term Memory network is used to further optimize the model.Finally,experiments conducted on two online education data sets show that compared with other knowledge tracing models,the deep knowledge tracing model that integrates forgetting behavior and learning ability has better explanatory and predictive performance,which proves the effectiveness of the knowledge tracing model that integrates forgetting factors and learning ability of learners.(2)In order to further improve the performance of the knowledge tracing model and improve the modeling of forgetting behavior,a knowledge tracing model based on deep forgetting was proposed.The model took the knowledge tracing model based on the key-value memory network as the framework,used the long and short term memory neural network to improve the long-term dependence of the model,considered three kinds of forgetting factors,and used the residual network to realize the deep modeling of forgetting factors.Experiments show that the knowledge tracing model based on deep forgetting modeling can effectively model the forgetting behavior and has a high prediction accuracy in the experimental data set.(3)In order to improve the interpretability and predictive performance of the deep knowledge tracing model,a forgeable deep knowledge tracing model based on item response theory is proposed.First of all,the model takes into account three factors of forgetting to realize the in-depth modeling of forgetting behavior.Then,the feedforward neural network was used to improve the item response network.The difficulty of exercises was divided into two parts,and the probability of correct answers was predicted together with the learning ability.Finally,experiments on three public education data sets show that the item response network-based forgettable deep knowledge tracing model can predict the forgetting line in the experimental data set with high accuracy and good interpretability,which verifies the effectiveness of deep forgetting behavior modeling and item response network modeling.