With the method of dynamic programming,two spatial variables,the expected utility and the Probability of success for each offense,are used to model the criminal's lacation choices in urban areas.The modeling results show that a eriminal optimizes his crime location according to the expeeted utility and the probability of success during his crime planned period.A criminal usually commits his first offence in the district which has the highest probability of success but a lower expected utllity,and commits his last offence in the district where the expected utility is the highest and the success probability is lower.If a loeation has both a higher expeeted utility and a higher probability of success,the criminal will commit all his offeese in this place.The model also suggests that crime prevention measures should be adopted in the light of loeal conditons."Coering" measures,such as patolling,should be taken in the poor residential districts or delinquency districts,while more sophisticated and advanced measures should be introduced in the richer districts or the districts where career crimimals haunt.