基于因素分解的月最高气温预测

The Prediction of Monthly Maximum Temperature Based on Factor Decomposition

摘要: 利用因素分解方法中的Holt-Winters三参数指数平滑法对月最高气温预测进行了研究。该类方法主要考察序列外部会影响序列波动的确定性因素，查看序列有无各明显特征并按其固定特征进行因素分解。根据Holt-Winters三参数指数平滑法并利用所获得的黑龙江克山气象站1952年到2017年各月的最高气温作为支撑数据构建相应模型进行预测，最终将预测出的未来的12个月即2017年每个月的最高温度与2017年实际每月最高温度进行对比且计算出平均绝对百分误差值。

Abstract: The Holt-Winters three parameter exponential smoothing method in factor decomposition methods is used to carry on research about the prediction of the monthly maximum temperature. This kind of method mainly investigates the certainty factors outside the series that will affect the fluctuation of the series, checks whether there are obvious characteristics of the series, and conducts the factor decomposition according to the fixed characteristics. According to the Holt-Winters three parameter exponential smoothing method, and using the obtained monthly maximum temperature of Keshan meteorological station in Heilongjiang Province from 1952 to 2017 as the supporting data to build the corresponding model for conducting prediction, and finally use the values after predicting for the next 12 months, which are the maximum temperature of each month in 2017, compared with the actual maximum temperature of each month in 2017 and calculated the value of average absolute percentage error.