Functional clustering with application to air quality analysis

Abstract

School of Mathematics and Statistics, Nanjing University of Information Science & Technology, \u00a0 Nanjing 210044, China (Received March 21 2019, accepted June 20 2019) Based on the air quality status of 161 cities in China, this paper studies the temporal and spatial distribution characteristics of PM2.5 concentration of major pollutants affecting air quality index (AQI). We use \u00a0improved \u00a0functional \u00a0clustering \u00a0analysis \u00a0methods \u00a0and \u00a0add \u00a0priori \u00a0information \u00a0about \u00a0location \u00a0and \u00a0human factors to make the clustering results more accurate. The improved functional clustering \u00a0model is compared with \u00a0the \u00a0basic \u00a0sparse \u00a0data \u00a0function \u00a0clustering \u00a0method, \u00a0k-centres \u00a0functional \u00a0clustering \u00a0method, \u00a0functional principal component analysis and traditional K-means clustering method by repeated simulation. Finally, we use the PM2.5 concentration of selected 161 cities in China as an illustrative example.