Understanding the geographic and environmental factors that affect local criminal activity is important for crime prevention. In this study, we use data from Geo-Twitter to analyze the reasons underlying the occurrence of local criminal activity. In recent years, abundant location-based social network (LBSN) (e.g., Foursquare, Geo-Twitter) data has become easily available at a low cost. Therefore, many studies have used LBSNs data to model and understand human mobile behavior, such as patterns of human travel and activity. However, few studies on local criminal activities have been reported. In this paper, a new methodology is proposed to identify the reasons underlying local criminal activity from the view point of geographic and environmental factors. Our methodology consists of the following steps. First, we collect geo-tagged data from Twitter. In particular, we extract a large corpus with geo-tags, called tweets, from major cities in the United States. Second, we measure the sentiments expressed in tweets posted from a specific area using the fastText model . Third, we apply a simple clustering technique called latent Dirichlet allocation (LDA) to identify the topics that are clustered in each area using sentimental analysis. Lastly, we analyze the reasons for crime by comparing the topics and the information on all crime using the data portal.