Perbandingan Jarak Metrik pada Klasifikasi Jamur Beracun Menggunakan Algoritma K-Nearest Neighbor (K-NN)

Abstract

Mushrooms are organisms from the kingdom fungi that have a fleshy body structure and can be consumed, but there are some species of mushrooms that are not safe to eat and have specific characteristics, so distinguishing between edible and poisonous mushrooms can be tricky due to the almost identical appearance of various mushroom species. Errors in identifying edible mushrooms can impact the health of consumers who consume the mushrooms. Evaluating the performance of various methods on a dataset is a key step in determining the most suitable classification method. This research is about how to measure the performance of classification methods on toxic mushroom datasets using the K-Nearest Neighbor algorithm with several metrics such as euclidean, manhattan and minkowski, which is a method for classifying new data based on proximity to existing training data. The results obtained in this study with several distance metrics can be concluded that the accuracy value of the manhattan metric is better than the euclidean and minkowski metrics. Because the manhattan metric gets the highest accuracy result of 99% with K = 100 and the lowest 82% with K = 3000, while the euclidean metric gets accuracy results with a value of 98% with K = 100 and 72% with K = 3000, and the minkowski metric gets accuracy results with a value of 96% at K = 100 and 64% at K = 3000.