Haze has a significant impact on human health, ecological environment, transportation and other aspects. When haze comes, public concerns are widely expressed on micro-blogs and other media, which contain netizens' psychological reflections on haze. It is helpful for city managers to adjust public opinion propaganda and resource allocation in haze situation by grasping the relevant reactions of micro-blog netizens to haze. Considering that the higher the co-occurrence frequency of the same words in micro-blog text is, the more likely the same topic is addressed, a community-based co-word network method was used in this paper to mine micro-blog topics. Firstly, keywords like "haze", "air" were used to collect micro-blog data in winter 2017. Secondly, TF-IDF algorithm was applied to extract keywords from pre-processed micro-blog data, from which the co-occurrence relationship between keywords was obtained. Then the co-word network was constructed based on the co-occurrence relationship, and the topic community was detected according to the Louvain algorithm. Finally, taking seven typical cities in North China, East China and Northeast China with the most serious haze pollution as the research areas, the difference degree of topic communities in co-word network of each city was analyzed combined with the Pagerank value of keyword nodes. From the experimental results, it can be shown that there are similarities and differences in different cities' attention to haze events. The results of the study have a certain theoretical guidance for the urban differentiated management in the haze situation.