Research On Application Of Deep Learning Model In College Admission Score Prediction Engineering

The common prediction methods for college entrance examination scores are wired difference method,average ranking method,etc.There are two main problems in the above method: one is to use scores or rankings as the forecast basis,without considering the impact of changes in enrollment scale and provincial control line on the level of college admissions scores;and two,it does not consider the impact of college development trends on admissions scores.First,the online percentile measurement,eliminating the impact of enrollment scale and provincial control line scores on the college admissions score measurement,and realizes the normalization of admission scores.Then,an LSTM prediction model was constructed based on the long-and short-term memory neural network in deep learning.Collected score data of undergraduate batches and batches of liberal arts from 1053 universities in Hebei Province from 2010 to 2017,and divided them into 5559 training samples and 251 test samples.The hidden layer dimensions is 32,mini-batch is 100,the learning rate is 0.0001 and the epoch is 12000,the predicted lowest score on the model output line is converted into the predicted lowest score through the scale of one point and one table,and compared with the actual lowest score of the year in the university.The results show that the LSTM model predicts the lowest the sum of the squared average error is 68.03,the average ranking method is 146.14,and the online percentile regression algorithm is 88.12.The prediction accuracy of the LSTM model is 53.45% higher than the average ranking and 13.74% higher than the online percentile regression algorithm.The experiment proves that the LSTM prediction model can independently learn the influence of factors such as the development trend of colleges and universities on admission scores,thereby predicting college admissions scores.However,the LSTM model has a large deviation when predicting admission scores of top universities such as Peking University,and it can be solved by combined prediction in the future.Figure 27;Table 23;Reference 50...