Application of k-means clustering to an air pollution monitoring design