Abstract: For multivariate prediction problems, this paper constructs fusion prediction model of rough set and grey system theory. The model adopts theory of knowledge dependency to reduce multiple attributes, GM（1, N） model is established based on knowledge reduction. Annual water supply quantity of Xi’an is fitted and forecasted by using the established model, and fitted values is compared with values of DGM（1, 1） model. The experimental results indicate that this method has advantage over selecting the factors of influence by traditional grey incidence grade. Thereby, a kind of method suitable for water supply forecasts is provided.

Key words: rough set, knowledge dependency, grey correlation, GM（1, N） model

摘要： 对于多变量预测问题，构造了粗糙集和灰色理论的融合预测模型。该模型运用粗糙集的知识依赖度理论对多属性进行约简，在约简基础上建立GM（1，N）模型。用所建模型对西安市年供水量进行了拟合和预测，并与离散灰色GM（1，1）模型作比较。实验结果表明该模型的预测精度高于传统的用灰关联度选择影响因子建模，从而为供水量预测问题提供了一种新方法。

关键词: 粗糙集, 知识的依赖度, 灰色关联度, GM（1, N）模型