数据挖掘，认知诊断，个性化推荐，系统设计

ABSTRACTWith the development of technology,our life style and leaming style are changing.The increase ofresources and the change of ways have a subtle influence on traditional education.More and more leamersuse these resources,and produce a large amount of data,which contains great analysis and market value,and further promote the development of personalized services in education.The goal of personalizedrecommendation is to select the right content for users according to their individual characteristics.Personalized learning can analyze the cognitive level,leaming ability,leaming style and other characteristicsof leamers through a large amount of data produced by leamers.On the basis of professional analysis of alarge number of relevant data generated by leamers,personalized leaming can make an in-depth analysis offactors such as leaming style,leaming ability and cognitive level,so as to organically choose the appropriatelearning style for leamers.In this way,on the basis of recommending appropriate leaming resources,theefficiency of their knowledge acquisition can be improved,and at the same time,they can leam moreknowledgeIn the aspect of detecting the leaming effect of leamers,the methods used are often quite many.Howto test the learning effect of students?More use,and the most convenient way is to practice.The key problemof this paper is how the online leaming system provides students with appropriate exercises according to theircurrent leaming status.In the case that leamers'answers are taken as the important basis,and the combinationof cognitive diagnosis model is taken as the important premise,the modeling of leamers'leaming level isimplemented.In the traditional process of education and teaching,it mainly depends on the teacher to selectthe test questions manually.With the wide application of online education websites and question banksystems,a large number of practical resources make it difficult to automatically select and recommendindividual exercisesThe Internet not only brings the increase of resources,but also enriches the data.In data,there is a lot ofinformation that can be mined.Based on the research of personalized recommendation algorithm,combinedwith cognitive diagnosis model,this paper proposes a personalized exercise recommendation algorithm basedon cognitive diagnosis.According to the leamers'answers,combined with the cognitive diagnosis model,thelearning level of leamers is modeled.Based on the learning level of leamers combining theproblem-knowledge point matrix and the difficulty information labeled by experts,the neural network is used