Implementation of Simple Linear Regression for Predicting of Students’ Academic Performance in Mathematics

Abstract

Predicting student academic performance is an interesting thing to research. Student academic performance can be used to determine the level of student mastery of the subject matter that has been delivered. This research uses academic and personal data of secondary students on mathematics subject scores in Portugal with 395 data records. The purpose of this research is to study how linear regression is applied in order to determine the predictive results of students' academic performance. Prediction evaluation is done by calculating attribute correlation with class, Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results of this research test the accuracy obtained linearly with correlation results, namely class failure (d) has the smallest RMSE and MAPE with an RMSE value of 1.148 and a MAPE of 9.82% of students' academic performance in mathematics. The results of the data analysis show that the failure variable has a positive effect on G3, where the probability value of the F test, the significance value for the simultaneous failure effect on G3 is 0.006 <0.05 and from the analysis of the coefficient of determination it is known that the failures variable is significant to the dependent variable with a large influence 63,8 % (model 2).