Resource Recommendation Based On Learner Feature Mining

With the development and popularization of the Internet,online learning represented by MOOC has become a new way of learning.During the COVID-19 outbreak,the activity of "stopping classes and learning" further promoted the application breadth and depth of MOOC learning.Although MOOC learning has the advantages of not being limited by time and space,low audience threshold,abundant learning resources,etc.,there are also many problems such as learners being submerged in massive resources,difficulty in finding learning resources that meet their own needs and high dropout rate.Therefore,how to provide personalized resource services for online learners is an urgent problem to be solved in the field of online learning.Through the analysis and mining of online learning behavior data,it is helpful to understand learners’ learning process and master learners’ learning state,so as to provide personalized services for learners.In the study of learning behavior,most of the existing research methods use the learning style scale as a subjective way to mine learners’ characteristics,which has some problems such as low accuracy.In the research of personalized resource recommendation,the recommendation method based on similarity usually only considers similar learners or similar resources,ignoring the knowledge mastery of learners;In addition,when modeling learners’ cognitive level,the influence of learners with similar characteristics is often ignored.In view of the shortcomings of current research,this paper proposes a method that considers both learners’ knowledge mastery status and similar learners,that is,using deep knowledge tracing model to track the changes of learners’ knowledge status,so as to obtain learners’ cognitive status characteristics,mining learners’ learning objectives and learning attitude characteristics from learning behavior data,modeling learners with these characteristics,and using collaborative filtering method to complete personalized resource recommendation tasks based on similar learners.The main research contents of this paper include:(1)According to learners’ answer data,the deep knowledge tracing model is used to track learners’ learning process,predict learners’ answer performance to other topics,and then obtain learners’ cognitive state characteristics.(2)Analyze and mine the behavior data of learners browsing the course video,and extract the knowledge points mainly contained in the video according to the video title and subtitle information,so as to determine the learning objectives of learners.Classify learners’ learning attitudes according to the times of clicking videos and the data of watching videos.With the passage of time,the cognitive state,learning objectives and learning attitudes of learners are constantly updated,and a dynamic learner model is established according to the characteristics of these three dimensions.(3)Using collaborative filtering,similarity calculation and other methods to achieve personalized video resource recommendation based on similar learners,and analyze the recommendation results.