Law enforcement agencies have started using social media for building community policing, i.e., establishing collaborations between the people in a community and local police departments. Both researchers and practitioners need to understand how the two parties interact on social media on a daily basis, such that effective strategies or tools can be developed for the agencies to better leverage the platforms to fulfill their missions. In this paper, we collected 9,837 tweets from 16 municipal police department official Twitter accounts within 6 months in 2015 and annotated them into different strategies and topics. We further examined the association between tweet features (e.g., hashtags, mentions, content) and user interactions (favorites and retweets) by using regression models. The models reveal surprising findings, e.g., that the number of mentions has a negative correlation with favorites. Our findings provide insights into how to improve interactions between the two parties.