Analysis of Accuracy Improvement in K-Nearest Neighbor using Principal Component Analysis (PCA)
Alida Lubis, Poltak Sihombing, Erna Budhiarti Nababan · Journal of Physics Conference Series · 2020
Abstract This research conducted to simplify and eliminate attributes or features that are less relevant without reducing the intent and purpose of the original data using Principal Component Analysis (PCA) to improve the performance (accuracy) of the K-NN (K-Nearest Neighbour) classification method. Modifying the method yields an average accuracy of 88%, where the K value is at K = 3 to K = 9. While the lowest accuracy value generated from K-Nearest Neighbors Conventional with PCA + K-NN classification models for the Pekanbaru Air Quality dataset has an average accuracy of 87% where when the K value is at K = 2, K = 4, K = 6 and K = 8. The highest average error is 0.11% where when the value of K is at K = 2, K = 4, K = 6 and K = 9. While the lowest average error value generated from K-Nearest Neighbors Conventional with PCA + K-NN classification model of the Pekanbaru Air Quality dataset is 0.08% where when the K value was at K = 5.