This paper presents a method for analyzing patterns of criminal activity that occur in space and time. The method uses the fuzzy C-means algorithm to cluster criminal events in space. In addition, a cluster reorganization algorithm is included to preserve the order of fuzzy partitions from one time step analysis to another. Order preservation is possible since crime forms relatively stable patterns due to the fixed shape of urban spaces and routine activities of people. The method provides a novel way to analyze criminal directionality, since it generates time series from clustering. A sample database of robberies in San Francisco, USA, is used to test the algorithm. Results show that criminal patterns might be tracked in a simple way.