K-NN Based Air Classification as Indicator of the Index of Air Quality in Palembang

Abstract

Good air quality is something that is wanted by every human who lives in big cities. Clean air and no pollution is one of the proper environmental requirements. One of the most severe causes of air pollution is due to large-scale forest fires due to the long dry season or is carried out by irresponsible persons which they commonly refer to as land clearing in an easy and inexpensive way by utilizing the reason of the dry season. The purpose of this study is to classify air quality in Palembang using a data mining approach. Then use the results of the classification as an indicator of the level of air quality in the city of Palembang. The data mining approach that researchers use is the K-Nearest Neighbor algorithm. Based on the test results of K-NN calculations and measured using a confusion matrix produce an accuracy of 80 percent, 82.3 percent for precision, and 93.3 percent for recall. The measurement results show that the calculation using the K-NN algorithm can be used as an indicator in measuring air quality, of the 20 that have been trained and tested only 4 inaccurate data, this inaccuracy occurs because the source data has unbalanced classes such as unhealthy and very unhealthy healthy have 1 sample each. So it proves that the performance of classifiers using the K-NN algorithm relevant as an indicator of air quality levels in the city of Palembang.