Application of K-Nearest Neighbor Method for Prediction of Best-Selling Fruit Sales at Ziel Kiosk

Abstract

Ziel kiosk sells various types of high-quality fresh fruits. Unfortunately, there is currently no system that manages fruit sales prediction, so there is often a buildup of goods, damaged and rotten goods, or even a shortage of goods, resulting in losses for the kiosk. The data collected is less accurate and effective because the current system is operated manually. This research conducts a data mining process on fruit sales data from Ziel Kiosk from January - December 2023. In sales prediction, Fruit Kiosks can use data mining techniques to be more proactive in managing stock items. This not only avoids the accumulation of fruit stock that can cause spoilage and damage, but also reduces the risk of stock shortages that can affect customer satisfaction. The purpose of this research is to ensure that Ziel Kiosk can see the sales rate for each product sold, so that they can avoid the accumulation of goods and concentrate on the most sold products. With an 80:20 data split, the K-Nearest Neighbor model has high accuracy. This algorithm can also predict fruit sales with an accuracy rate of 97.22% by determining the categories of fruit sales that are in demand or not in demand.