In parallel to the application of data mining for terrorism (post1, post2, post3), people also use it for discovering crime patterns. This is the case of Shyam Varan Nath in one of his paper. You can found information about this paper thanks to a comment from Ralf.
In this paper, Shyam explores clustering techniques for supporting identification of crime patterns. According to him, an important issue in data mining crime data is to convert free text in data mining attributes. This usually involves some experts in the crime field. Selected attributes can also have a weight according to the crime types. One of the limitations in his study, which I would more refer to as a conclusion, is that crime pattern analysis can support the detective although not replacing him. For more details, you can read the paper or visit his blog.