Structure Learning of mbox{PM}_{2.5} Distribution Using Sparse Graphical Models

We consider the structure learning problem of the mbox{PM}_{2.5} pollution data over 31 provincial capitals in China. Specifically, we make use of the graphical model tools to study the hubs and the community structures of the mbox{PM}_{2.5} pollution networks. The results show that the hubs in the mbox{PM}_{2.5}pollution networks are always seriously polluted cities, and the mbox{PM}_{2.5} pollution networks have significant community structures which consist of cities which in some sense can be regarded as blocks with similar cause of pollution. In view of the results, we suggest that the government should strengthen theeffort to treat the seriously polluted areas and western China areas. Moreover, the management of the mbox{PM}_{2.5} pollution should be region-dependent.