Abstract: For the characteristic of meteorological time series data set, under the framework of quotient space granular computing model, complicated meteorological information is analyzed at different layers with different grain-sizes. Additionally, the information is dealt with by the combination method in quotient space theory and the thinking in Multi-Side Increase by Degrees Algorithm. A model that combines grey model GM (1, 1) and structural machine learning method (Alternative Covering Algorithm) is presented for meteorological time series data mining (the forecast of yield). Finally, the experiments on data sets from real world (the yield forecast for winter wheat) are conducted by this model, and the result is satisfying.

Key words: grey model, quotient space, granular computing, structural machine learning method, meteorological time series

摘要： 本文从一种新的角度，针对气象时间序列的特点，在商空间粒度计算理论框架下，采用多种粒度，从不同的层次分析复杂的气象数据信息，利用商空间的合成技术，和多侧面递进算法进行综合信息处理。并提出了一种灰色模型GM(1,1)与构造性机器学习方法（交叉覆盖算法）结合的模型对气象时间序列进行数据挖掘（产量预测）。最后，通过本模型在真实数据上的实验（冬小麦产量预测），取得了令人满意的结果。

关键词: 灰色模型, 商空间, 粒度计算, 构造性机器学习方法, 气象时间序列