The Prediction Model Of Difficulty And Differentiation Of College Entrance Examination Mathematics Questions And Papers Based On Entropy Weight Method And BP Neural Network

As a large-scale examination to select talents,the college entrance examination has always been a hot topic in education research,especially the difficulty of the exam questions,which often attracts the public’s keen attention in the society.In addition,for the selective exam,the degree of differentiation is also an important index to evaluate the exam,which is generally believed to be affected by the difficulty.The traditional method uses the score after the test to calculate the difficulty and differentiation of the test questions,which makes the evaluation of the test questions lag behind and can not play a guiding role in the proposition before the test.Based on such thinking,this paper first establishes the difficulty prediction model of college entrance examination mathematics questions and papers.Then,on this basis,it constructs the differentiation prediction model of college entrance examination mathematics questions and papers.The specific research contents and conclusions are as follows:(1)Reconstruct the difficulty evaluation system of college entrance examination mathematics questions.Based on the theory of mathematical problem solving and the comprehensive difficulty model of college entrance examination,the difficulty evaluation system of college entrance examination mathematics questions is reconstructed,which includes eight factors,including background factor,whether to include reference,operation level,reasoning ability,knowledge content,cognitive level,reading amount and condition content.(2)Optimize index weight.Firstly,entropy weight method was used to obtain the initial weights of the eight factors,and then BP neural network algorithm was used to further optimize the weights.The final weight distribution was relatively uniform and reasonable.(3)Construct the difficulty prediction model of college entrance examination mathematics questions and papers.Using BP neural network algorithm to build a three-layer BP neural network model,taking 8 sets of high school quality inspection mathematics questions as the research object,76.2% of the samples as the training set,to build the potential relationship between 8 factors of the difficulty of mathematics questions and the score rate after the test,using this relationship can achieve the purpose of predicting the difficulty of questions and papers from the inherent attributes of the questions.23.8% of the sample was the test set.After the data fitting test,there was no significant difference between the scoring rate and the estimated difficulty,indicating a good test effect.(4)Construct the differentiation prediction model of college entrance examination mathematics questions and papers.Using the data fitting method,it is known that there is a quadratic function relationship between the score rate after the test and the differentiation degree after the test.This relationship further realizes the prediction of the differentiation degree of the test questions and papers on the premise of realizing the difficulty prediction.There is no significant difference between the differentiation degree and the prediction degree after the test,indicating that the test effect is good.(5)Verify the difficulty prediction model of college entrance examination mathematics papers.Taking the 2019 College Entrance examination of Science and Mathematics in three sets of papers I,II and III as the research object,the difficulty prediction model of college entrance examination mathematics papers constructed in this paper was verified.The test showed no significant difference in the difficulty prediction and score rate after the test,indicating that the prediction effect was good.