| In recent years,the rapid development of emerging information processing and data transmission technologies such as data mining,machine learning,and fifth-generation mobile communication systems has accelerated the process of educational informatization reform,making the traditional offline classroom gradually transformed into a new model of integrated development of online education and offline teaching.With the widespread popularity of the Internet and the strong support of national policies,the number of people receiving online education and the market demand for online education have increased year by year.At the same time,smart education that uses modern information technology to promote teaching and learning has become a hot spot in the global education field.At present,the research on smart education mainly includes two aspects:the assessment of learners’ knowledge level and the recommendation of personalized test question resources.The test method used to evaluate the knowledge level of learners has been developed into a cognitive diagnostic model that integrates high-order learning ability characteristics.However,the existing cognitive diagnostic models still have areas that can be improved in the consideration of high-order feature parameters.The research on personalized test item resource recommendation methods mainly includes test item recommendation method based on cognitive diagnosis,test item recommendation method based on collaborative filtering,and test item recommendation method based on deep learning.The research work of this paper is carried out on the basis of the above,as follows:(1)On the basis of the existing cognitive diagnosis model,this paper studies the C&RM cognitive diagnosis model constructed by introducing highorder features of the effort level,and improves the test item knowledge point correlation matrix in the model.The discrete test item knowledge point correlation matrix is modeled as a test item knowledge point dependency matrix of continuous representation.Finally,comparing C&RM with MIRT,HO-DINA and FuzzyCDF models,the test results show that the accuracy of C&RM remains in the range of 77.7%-84.2%under 30%to 90%of the training set data ratio,and the best in the latter is 72.3%-81.9%.Compared with it,C&RM has better performance.(2)Existing test item recommendation methods based on collaborative filtering often ignore learners’ personality characteristics(knowledge level),test item recommendation methods based on cognitive diagnosis ignore the common characteristics between similar learners,and test item recommendation methods based on deep learning need large amount of training data and lack of interpretability of recommendation results.In view of the shortcomings of the above three methods,this paper proposes a PMF-C&RM test item recommendation method that combines cognitive diagnosis and collaborative filtering.The experimental results show that the PMF-C&RM method has an average error rate of 0.192 under the proportion of 30%to 90%of the training set data ratio,while the PMF method is 0.284,and the C&RM method is 0.206.Compared with them,the PMF-C&RM method has better performance.(3)Taking the C&RM cognitive diagnosis model and the PMF-C&RM personalized test item recommendation method as the core,this paper designs a personalized learning evaluation and recommendation system for teachers and learners.The system has a friendly interface and includes rich functions such as learning ability assessment,knowledge level assessment,and personalized test question recommendation. |
