THE PREDICTION OF NATIONAL EXAM SCORES OF JUNIOR HIGH SCHOOL STUDENTS USING K-NEAREST NEIGHBOR (k-NN) ALGORITHM

Abstract

The prediction of the acquisition of national exam scores for Junior High School (JHS) students is intended to know the results of the student’s national exam early when students take the national examination. Knowledge gained from the results of this prediction will be important information for the school to take appropriate steps so that the acquisition of student national exam scores can be improved even better. The acquisition of student national  exam scores is low and there is no prediction model that is used to predict the achievement of student national exam scores is a problem that needs to be addressed. This paper propose a predicting student’s national exam scores for four national exam subjects (INDONESIAN, ENGLISH, MATHEMATICS and SCIENCE) using K-Nearest Neighbor (k-NN) as a prediction method and compare it with Decission Tree method. The results of the study showed that the prediction k-NN model had better performance than the prediction model of Decission Tree. Performance results obtained by evaluating using derivatives of the confussion matrix terminology to determine the value of accuracy, sensitivity (recall), and precission each subjects. To measure the performance of predictive methods used the value of accuracy in each method and each subject. The greater the accuracy value ( max 1 ), then the better performance of the prediction model used. Performance of k-NN in average accuracy=0.85, precision=0.87, recall=0.91 is better than Decission Tree method performance with accuracy=0.82, precision=0.85, and recall=0.89.