Prediction Of College Entrance Examination Batch Line Based On Gray Neural Networks

At present,people are paying more and more attention to education.As an important part of education,the college entrance examination is the most impor-tant thing for senior high school student.There are ten thousands of candidates participating in the college entrance examination of china every year.However,the voluntary reporting of the college entrance examination determines the different destiny of each candidate.Therefore,it is necessary and meaningful to study the ad-mission line of colleges and universities.With the development and popularization of the Internet,citizens can search the Internet for information they need.During the volunteer reporting period,The search behavior of Candidates’ can reflect the college entrance examination from the side.This article takes the search index of college-related keywords as an innovation point to study and realize the prediction of the college entrance examination batch line.Firstly,this article analyzed the influence factors of the college admission line.On the basis of enrollment plan,rankings and regional line,a new feature of online search index was introduced,which reflected the popularity of colleges and univer-sities.The Java language was used to write reptile program.The information was obtained from the relatively authoritative education website.According to the do-mestic market share of search engine,Baidu index is selected as the data source of the search index,Baidu index of college-related keywords is obtained from the Baidu index query platform,and the principal component analysis method is used to reduce dimension and synthesize the composite network search index.Secondly,many methods were used to predict the college admission line.Ac-cording to the conditions of the independent variables were the college enrollment plan,composite network search index,regional batch line and ranking and the de-pendent variable was real line difference,the multivariate linear regression method was used for prediction.In order to eliminate the effect of size and year,taking the line difference as the research object,using the gray prediction theory to achieve the prediction;In order to improve the accuracy of model prediction,the combination prediction model was constructed.The support vector machine method and BP neural network algorithm were used to combine with the grey prediction theory,and then the regression prediction model and grey neural network prediction model would be get.Finally,the information of a college in Hunan province was analyzed as an ex-ample to compare the accuracy of four models.The gap between 2011 and 2016 was set as the training set,and the difference of the admission line in 2017 was set as the test set.The verification results show that the MLR1 model,which takes network search index into account,has a good prediction effect,and its forecasting effect is better than the grey forecasting model that takes the line gap as the re-search object.In addition,the prediction results of GNN model and SVR model were superior to the multiple linear regression method and grey prediction model,and the prediction result of GNN prediction model was better than the SVR prediction model.When the number of hidden layer nodes was nine,the relative error between the predicted value and the real value of the lowest input line of GNN model was 6.14%.While the relative error was 4.03%when the number of hidden layer nodes was eight.Therefore,the GNN model has highly accuracy in predicting the college admission line,which has important evaluation value for the students’ voluntary reporting.