Research On The Difficulty Level Of Programming Questions Based On The Online Judge System

With the development of computer education and computer technology,intelligent computer-aided instruction is more and more popular in computer education,and more and more attention is paid to the cultivation of practical ability in computer technology.The training of practical ability lies in the cultivation of programming ability.Obviously,the programming ability of learners is various.Some of them are good at,while others are not.How to recommend a suitable topic according to their ability is the key to efficient computer-aided teaching,or where it was wrong.More and more people use Online Judge system to program and study,so Online Judge system has saved a lot of evaluation data,analyzed these data from big data's point of view,and created the difficulty classification model of programming topics according to the experimental data.It is of great significance to the intelligent recommendation function of the online evaluation system in the future.This paper obtains a large number of programming test questions from the Codeforces system question bank of the famous online program evaluation system,and uses Python to write scripts,crawling the corresponding user submission information on the Codeforces system,on the basis of big data.The feature vectors which affect the difficulty of the test questions in the information submitted are sorted out and extracted.Firstly,the data from the obtained samples is analyzed by combining the decision tree algorithm in machine learning.A binary decision tree,which can automatically predict the difficulty of a subject,is trained.Based on the binary decision tree,a stochastic forest classification model is established,which is based on the binary decision tree and combined with a machine learning stochastic forest algorithm.Then the two automatic classification models are compared,in which their advantages and disadvantages are analyzed,After that the optimal classification model is finally determined,and the programming topics are automatically classified.