An improved K-means algorithm and its application in the evaluation of air quality levels

This research introduces an improved k-means algorithm which combines hierarchical method with k-means method and the application in the evaluation of air quality levels. In present, evaluation of air quality relies on a tedious index computing based on a formula. A comprehensive analysis on single pollutant cannot be seen. The new clustering method gives more proper initial points calculated by hierarchical method and chooses a more precise distance computing method that makes the coefficients the smallest for association between every sample. It has a more proper clusters result as well as less iterations. This paper collects 392 days´ air pollutants records and does the clustering by the new method. An evaluation including five levels indicates every day´s pollutant features. This application will help researchers do analysis more convenient and give possibility for the automated of evaluation in the big data era.