Application Of Over Cut-off Line Percentile Regression Model In University Admission Score Prediction Engineering

The college entrance examination is the main way of enrolling candidates in universities in China.At present,the common methods for predicting college entrance examination scores all ignore the relationship between candidates’ scores,rankings,and the current year’s score line,and ignore the impact of factors such as changes in the development trend of colleges and universities and the deviation of admission minimums from the normal on forecast accuracy,and affect the overall prediction accuracy.On the basis of previous studies,a prediction model for college admission scores combining online percentiles and linear regression was proposed,and the model was optimized by eliminating outliers.Firstly,the influence of factors such as the degree of difficulty of the question and the ranking of candidates on the prediction model is solved by constructing an online percentile.Secondly,use linear regression to find out the change trend of college admissions level,and then predict the percentile and admission score of the lowest score of college admissions next year.Finally,a minimum score outlier elimination algorithm was designed based on analysis of variance to eliminate the influence of abnormal minimum scores on prediction accuracy caused by factors such as point moves.The historical admission data of a group of undergraduate colleges and universities enrolled in Hebei Province in 2017 were selected,and four methods were used to predict the 2017 college admission score line.The results show that the construction of online percentiles solves the problem of college admissions level measurement.The linear regression based on the average percentiles on the average score line of the calendar year of colleges and universities obtains the average score of the next year and predicts the development trend of colleges.When using 4-year data prediction,the online percentile regression model and the improved model are compared with the average ranking method to compare the average squared error of the prediction score,and the prediction accuracy is improved by 39.70% and 45.89%.Figure 13;Table 24;Reference 52...