The prediction of scholarship recipients in higher education using k-Nearest neighbor algorithm

Abstract

This article aims to implement the algorithm model of k-Nearest Neighbor (k-NN) in analyzing, predicting, and classifying students who have potentials to get scholarships in universities. The k-NN algorithm works by making a prediction based on the closest data points between the old data history as training data and the new data as testing data. The data collected totals 1018 students with 24 scholarship receiver candidate students are used as the dataset for the test purposes. The attributes used in the prediction process are a semester, parents' income, number of family dependents, and Cumulative Grade Point Average. The distance calculation of the value from testing attribute to each training attribute uses Euclidean Distance equation, while the test of the model accuracy value is calculated using Confusion Matrix. The results of the simulation of the prediction model show that the determining factor of training data from both the number and the variation of different values can improve the performance of the k-NN algorithm with the best accuracy rate of 95.83 percent in predicting students who have the greatest chance of getting the scholarship.