Prediction of primary education HSEE (LGS) science achievement through artificial neural networks

Abstract

This study aimed to investigate the prediction level of science success in high school entrance exams (HSEE /LGS) using artificial neural networks (ANNs) by associating students’ success in science classes from the 4th-grade elementary level. The Pearson Moment Product Correlation analysis results were analyzed using the SPSS program to examine the connection between students' performance on the LGS science subtest and their academic achievement in science. In the MATLAB program, artificial neural network modeling performance was examined to understand the level of prediction. The data sets were obtained from the electronic school system of the Ministry of National Education for the 4th to 8th- grades of 1027 students graduating from 24 schools in 17 districts of Bursa in the 2017-2018 academic year and entered the 2018 LGS and LGS result documents without personal information. When the correlations between the LGS science sub-test and science test success were examined (p <.001), the highest and lowest correlations were found in the 8th-grade science exams (r = .70) and the 4th-grade ones (r=.57), respectively. The highest performance values were learning R=.80, verification R=.74, and test R=.75, RMSE=2.35 at the network architecture, generated in the second sub-problem and the trained network with 845 student data. A high-level relationship with r=.75 (p<.001) using the data of 182 students was obtained, while the correct numbers of simulated real LGS science sub-tests compared with the correct numbers predicted by the computer after the training process of the network. In addition, it was obtained that 124 students out of 182 were correctly estimated in the [+ 2, -2] error value range.