Abstract: Objective To improve the preventive strategy of diabetes mellitus,reduce tbe cost of screening diabetes mellitus.Methods With neural network as the primary tool,the value of OGTT 2h blood glucose is adopted as the expected out-put.the fitting results of neural network based on age,DM family record,height,weight,waistline and seat circumference as the practical results,compared the differences between them.Results 90% DM can be found by testing the value of blood glucose of 54.9% of target population.Conclusion Using neural network as the primaxy screening tool can greatly decrease the cost of DM screening,which is a new promising epidemiological method.