Airlines pay close attention to maintenance cost for efficiency improvement and fuel consumption reduction. Due to the limitation and shortcoming of forecasting model using single mathematic method, a new idea of performing combination model instead of selecting model methods is proposed. The key point of a combination model is to determine the weight coefficients. In this paper, based on the rough sets theory, we employ the uniformity of rough sets and statistical methods to find out the significance of each attribute by its knowledge entropy. Then the weight coefficients can be calculated by determining the significant degree of each one. According to the experimental results, our proposed new model can obviously improve the prediction accuracy than the single model. This combination model has been partly adopted by airlines, which proved helpful in reducing the airlines costs.