Situational theories postulate that opportunities offered to the burglar are very specific. Each necessitates a particular motivation, a compatible target and place, as well as the absence of guard or protector. Thus, burglaries occur not randomly. Only particular locations and premises are victimized during certain period of the day and through particular methods (i.e. modus operandi) by certain types of offenders. A classification system has been developed that allows, from information collected by scene of crime officers on burglary places, to recognize new events as belonging to a specific category. This method is particularly useful to a priori separate the huge quantity of data to be analyzed. We assume that the route followed by criminals shows specific characteristics that are specific to each category. However, exact date and time of burglaries are very rarely known. Thus, the establishment of the burglar's path is difficult. This study explores possible use of graph theory to help reconstruct those paths. Three important categories of house burglaries have been chosen in order to show how those spatio-temporal patterns may help for classification.