Comparative Analysis of Restock Needs Bottled Water Using K-Nearest Neighbor (K-NN), Support Vector Machine (SVM), and the Naïve Bayes Algorithm

Abstract

Restocking goods is essential for bottled drinking water to ensure smooth production and maintain a stable product supply. This research aims to compare the K-Nearest Neighbor, Support Vector Machine, and the Naïve Bayes algorithm to predict the need to restock bottled water. The data set for training and training data is taken from Adimaru's Agent. The comparative analysis with three algorithms gives the results of the prediction analysis for the accuracy value of K-NN is 88.20%, SVM is 84.51%, and Naïve Bayes is 66.20%. The AUC values of the three result algorithms include Good Classification. The comparison of the K-NN and SVM with T-Test algorithms get obtained the best performance with an alpha value is 0.102. From this accuracy value, the classification method of the K-Nearest Neighbor algorithm has the best predictive model results for restocking needs of bottled water goods.