In hospital, a high quality of nurse rostering schedule can reduce medical cost and improve staff morale and enthusiasm. So the models of this study consist of two parts. First, we count kinds of patients and nurses manpower requirement by patient classification system. Second, we focus on building a nurse rostering schedule to maximize the total nurses satisfaction. We applied integer programming to construct the shifts and day-off assignment models. Finally, the models were developed to obtain an initial optimal schedule which will be as a base for practical nurses scheduling. The results show that the models were better than traditional shift scheduling method and can operate more efficiency in real applications.