支持向量机方法在单站降水预报中的应用探讨

The Discussion of Precipitation Predict in Wuhan Based on SVM Method

将武汉天空云量预报的81个预报因子运用到该站中等以上强度的降水预报中,基于SVM方法进行了交叉验证和预报试验。结果表明用81个预报因子建立的5～9月和全样本的降水预报模型有较好稳定性、且对降水都有正的预报技巧。因此天空云量的预报因子可以用来做降水的预报因子,同时也证明了这些预报因子在天空云量和降水预报中是协调的。SVM方法为天空云量和降水的预报提供了客观参考依据。

81 predictors from cloud amount forecast are used to predict the precipitation. Based on Support Vector Machine（SVM） method, cross-validations and test are performed. The results show that the stability and the forecast ability of the precipitation model are revealed. Therefore, the predictors from cloud amount forecast could be able to predict precipitation, and the predictors are harmony between cloud amount and precipitation predict. There is an objective way in cloud amount and precipitation predict by SVM