支持向量机在短期气候预测中的应用

支持向量机（SVM，Support Vector Machines）是基于统计学习理论框架下的一种新的通用机器学习方法。可以解决样本空间中的高度非线性分类和回归等问题，是一种处理非线性分类和非线性回归的有效方法。气候变化诸多因子的复杂性和非线性决定了预报因子与预报对象问的非线性关系，SVM为解决短期气候预测提供了一种可行的有效途径。利用Nino区海温、南方涛动指数、副高面积指数、亚洲区极涡面积指数等15个预报因子，建立了阳泉夏季降水正、负距平的SVM非线性分类模型，同时也建立了阳泉夏季降水的SVM回归模型，并进行了相应的预报试验，结果显示，对应的SVM分类模型和回归模型均具有良好的预报能力。

Application of Support Vector Machine to Short-term Climate Forecast

The support vector machine (SVM), a new general machinery study method based on the frame of statistical study theory, may solve the problems of non-linear classification and regression in sample space and be a effective method of processing the non-liner classification and regression. The complexity and non-linearity of factors of climate change decide the non-linear relationship between the forecast factors and the forecast object, and the SVM provide an effective and feasible way to forecast short-term climate. A SVM non-linear classification model of positive/ negative departure in summer precipitation is developed according to the 15 forecast factors, including the sea surface temperature of Nino area, the southern Oscillation index, the subtropical high area index and the polar vortex area index of Asian region, meanwhile, a SVM regression model of summer precipitation in Yangquan is developed as well.