摘要： 针对降水量影响因子多、彼此之间关系复杂、预测难度大的特点，提出一种基于多侧面集成学习方法的降水量预测模型。采用多侧面递进算法对数据进行特征提取，得到多个特征子集(多个侧面)，且在特征子空间上建立预测模型，对未知样本进行分类预测，并利用专家投票的大多数规则对预测结果进行组合。实验表明，该方法具有较高的预测准确率。

关键词: 多侧面递进算法, 多分类器融合, 月降水量预测

Abstract: Due to the factor for rainfall forecast is more, and the relation between them is very complex, monthly rainfall forecast is a difficult field in climate prediction, this paper proposes a rainfall forecast model based on the fusion of multiple classifier. The Multi-side Increase by Degrees Algorithm(MIDA) is introduced to extract features, obtaining several features subsets(several sides), and then some sub-forecast models are built in the feature subspace to classify the unknown sample. The majority voting principle based on expert is adopted to combine the class results from the forecast sub-model. Experimental results show that a higher accuracy rate of rainfall forecast can be achieved.

Key words: Multi-side Increase by Degrees Algorithm(MIDA), fusion of multiple classifier, monthly rainfall forecast

中图分类号: 

• TP39