Penerapan Algoritma K-Nearest Neighbor untuk Klasifikasi Persediaan Kain Sasirangan pada UMKM

Abstract

An inventory classification system plays an important role in supporting stock management among Sasirangan fabric MSMEs in Banjarmasin, South Kalimantan. This study aims to develop a web-based Sasirangan fabric inventory classification system using the K-Nearest Neighbor (KNN) algorithm. The system classifies fabric inventory into three categories: low, medium, and high. This study used 198 inventory data records from MSMEs in Banjarmasin, consisting of 158 training data and 40 testing data. The research stages included data collection, preprocessing, splitting the data into training and testing sets, implementing the KNN algorithm, and evaluating model performance using a confusion matrix and classification evaluation metrics. The evaluation results showed that the KNN model achieved accuracy, precision, recall, and F1-score values of 0.9500 based on the weighted average. Based on the macro average, the KNN model achieved precision, recall, and F1-score values of 0.9484. These results indicate that KNN can classify Sasirangan fabric inventory accurately, stably, and consistently. In addition, KNN demonstrated better performance than Naive Bayes, which achieved an accuracy value of 0.8000. Therefore, improving the quantity and quality of data, conducting feature selection, and optimizing the value of k through parameter tuning are necessary to improve the accuracy of Sasirangan fabric inventory classification.