基于数据挖掘CART算法的区域夏季降水日数分类与预测模型研究

夏季降水日数的准确预测，对于保障农业、运输业、电力等行业的有序进行具有重要现实意义.利用连云港市气象局提供的1951—2012年夏季降水数据对连云港地区的降水日数特征进行分析，难以直观地发现夏季降水日数随时间分布的规律.为进一步探索降水日数的发生规律，结合国家气候中心网站提供的多种气候因子数据，基于CART决策树算法构建了连云港地区夏季降水日数是否偏多与是否偏少的分类与预测模型.该模型可以发现在多种气候因子不同条件下，夏季降水日数是否偏多（偏少）的规律，模型的分类与预测都具有良好的效果.利用52 a的数据样本训练模型，模型的训练准确率为90.38%（86.54%），再用剩余10 a数据样本检验模型，测试准确率为80%（80%），并且得到规则集，方便气象业务人员使用以及决策服务人员参考.同时，为降水日数的预测提供了数据挖掘的新思路.

Model prediction of regional summer precipitation days based on CART algorithm

The accurate prediction of the number of summer precipitation days has important practical significance for industries such as agriculture,transportation,and electric power supply.The data of summer precipitation during 1951-2012 provided by Lianyungang Meteorological Bureau were used to analyze the interannual characteristics of summer precipitation days,yet no obvious temporal variation trends were found.Thus a model to predict the regularity of precipitation days is established based on analysis of climate factors listed by National Climate Center website,and CART decision tree algorithm.Year with positive/negative anomalies of summer precipitation days in Lianyungang is defined by various climatic factors,which is trained by sample data of 52 years with training accuracy of 90.38%/86.54%.The remaining data of 10 years are used to test the model,resulting in accuracy of 80% for positive/negative anomalies of summer precipitation days prediction.The rule set is provided for meteorological business and decision-making.