The Prediction Of National Examination Result For Junior High School Using Neural Network Backpropagation

Abstract

The national examination results can be used as a reference for the schools to determine school policy in the future. It also can be used as a basis for training and assistance to education unit in his attempt to improve the education quality as well. One of the prediction method can be used to predict the national examination result is neural network backpropagation. This study attempts to get the best architecture of artificial neural network backpropagation used to predict the value of national examination results and its accuracy. The simulation of national examination results prediction using the value of last semester examination and school examination as predictors is performed in R software. Data was obtained from SMPN 1 and SMPN 2 Lamongan. The obtained data set as much as 701 data was divided into 75% data as training set and 25% data as testing set. The result of processing data showed that the best architecture of artificial neural network used combination of 7 hidden layer and 0.9 learning rate with the average training error 4.397 and testing RMSE and MAPE were 7.28 and  0.55%. The best architecture of artificial neural network found from this research which is used combination of 7 hidden layer and 0.9 learning rate can be implemented for further research.