Penerapan Algoritma K-Nearest Neighbor untuk Menentukan Potensi Ekspor Komoditas Pertanian di Provinsi Sulawesi Tengah

Abstract

Agriculture is a highly robust sector in Indonesia. This is evidenced in Central Sulawesi Province, where the gross domestic product (GDP) from the agricultural sector, based on constant prices from 2018 to 2021, continues to experience growth. Such conditions suggest that commodities in the agricultural sector have the potential to become export products, enabling a greater economic boost for the region. Before engaging in exports, it is necessary to identify which commodities have potential. One way to determine this is by applying Klassentypology. To simplify the process, it can be implemented in machine learning using the K-Nearest Neighbor algorithm. K-Nearest Neighbor is chosen because this algorithm can handle data containing noise and has good adaptability when given new data. In this research, two machine learning models were developed. The first model is used to classify whether a commodity is advancing or lagging, while the second model is used to classify commodities that grow rapidly and slowly. The highest accuracy obtained from the first model is 96.23%. Meanwhile, the highest accuracy from the second model is 93.49%.

Keywords

• k-nearest neighbor
• agriculture
• klassen typology
• confusion matrix