支持向量机分类方法在天空云量预报中的应用

以2001年5月1日至2004年12月31日逐日武汉市地面、高空观测资料及欧洲中心24小时预报场等资料为基础,构建了不同的训练样本集,基于支持向量机方法进行了大量多因子的随机交叉验证,从而筛选出了包含最佳预报因子的训练样本集和相应的核参数g,建立了武汉市天空云量的预报模型。交叉验证结果表明预报模型是稳定性的、且具有较好的预报能力和推广应用能力。预报试验和实时预报的结果都显示出SVM方法对天空云量有一定的预报能力。

Application of SVM Method to Cloud Amount Forecast

Based on the data from the surface and upper air observations in Wuhan and numerical synoptic prediction data from EC during May 1st 2001 to December 31st 2004, the different samples are made. Based on SVM method, cross-validations are performed with randomly samples to find the samples with best factors and optimization parameter g, then the models of the cloud amount forecast are built, The stability, the forecast and generalization ability of the models are also revealed by cross-validations. The results of test and real-time forecast show the forecast ability of the cloud amount forecast models by SVM.