Research On Key Technologies Of Personalized Exercise Recommendation Based On Exercise’ Difficulty

Personalized exercise recommendation can effectively improve the effectiveness of students’ exercise-based self-learning,which is deemed as an important research topic in the field of intelligent education.Existing personalized exercise recommendation technologies do not fully consider the impact of difficulty of recommended exercises on the effectiveness of students’ self-learning which makes them quite difficult to help students achieve efficient step-by-step self-learning following the rule of from the easy to the difficult.In order to cover the shortage of existing research technologies,this thesis takes exercises’ difficulty model as the cut-in point,and studies several key technologies of personalized exercise recommendation based on exercise’s difficulty,aiming at improving the effectiveness of students’ exercise-based self-learning.Specifically,this thesis mainly covers the following two research aspects.(1)This thesis proposes a personalized exercise recommendation technology(named as ReDi)based on exercise’s difficulty.ReDi models an exercise’s difficulty as two parts,namely the subjective difficulty derived based on students’ knowledge level and the objective difficulty obtained based on exercise’s attributes.Moreover,prerequisite dependencies among knowledge points and the knowledge graph technology are also used to refine the exercise’s difficulty model.Based on the proposed exercise’s difficulty model,ReDi designs an effective personalized exercise recommendation algorithm,which recommends personalized exercises to students in easy to difficult order.The experimental results show that the recommended personalized exercises of ReDi greatly improves the effectiveness of students’ self-learning,compared with state-of-the-art technologies.(2)This thesis proposes a Q matrix validation technology based on DINA model.The quality of Q matrix poses an influence on the accuracy of the quantified knowledge level of a student based on cognitive diagnosis,and further impacts the effectiveness of personalized exercise recommendation based on exercise’s difficulty.In view of this,using students’ testing results as the input,this thesis designs an effective Q matrix validation technology based on the popular DINA model from cognitive diagnosis.The proposed technology picks out potential errors in defining the relationship between exercises and their corresponding knowledge points in the Q matrix at first,and then validates each potential error based on a grouping analysis of students.Experimental results derived from real educational datasets show that the proposed Q matrix validation technology has increased the validation accuracy by10% than related Q matrix validation technologies.Personalized exercise recommendation based on exercise’s difficulty is a problem worth further study,having important theoretical and practical significance.This thesis studies exercise recommendation technology based on exercise’s difficulty and Q matrix validation technology based on DINA model.Experiments conducted based on real classroom experiments and real educational datasets verify the effectiveness of the proposed key technologies in question for recommending personalized exercises based on exercise’s difficulty.