组合预测模型构建及在合肥市PM_(2.5)浓度预测中的应用  

Construction of Combined Prediction Model and Its Application in the Prediction of PM_(2.5)Concentration in Hefei

作　　者：孙飞燕 袁宏俊[1] 张圣梅[1] SUN Fei-yan;YUAN Hong-jun;ZHANG Sheng-mei(School of Statistics and Applied Mathematics,Anhui University of Finance and Economics,Bengbu Anhui,233030)

机构地区：[1]安徽财经大学统计与应用数学学院,安徽蚌埠233030

出　　处：《山西大同大学学报（自然科学版）》2023年第1期54-61,共8页Journal of Shanxi Datong University(Natural Science Edition)

基　　金：安徽财经大学研究生科研创新基金项目[ACYC2021398];安徽省哲学社会科学规划项目[AHSKY2020D42];安徽省高校自然科学重点项目[2022AH050602];安徽财经大学科研基金资助重大项目[ACKYA21004]。

摘　　要：Dice系数是衡量两个序列接近程度的指标,将广义诱导有序加权平均(GIOWA)算子和Dice系数相结合,提出了基于GIOWA算子和Dice系数的组合预测模型,并且给出了优性组合预测的定义。将该模型应用到合肥市PM_(2.5)浓度预测上,选取2014-2021年共84期的PM_(2.5)浓度及核心影响要素的月度数据进行分析,将组合预测结果与单项预测结果进行对比,并分析了组合预测模型中参数取值发生变化时对组合预测权重、预测效果评价指标和Dice系数影响情况。实证结果表明所建立的组合预测模型的预测效果要优于单项预测模型。Dice similarity coefficient is an index to measure the closeness of two sequences.Combining the generalized induced ordered weighted average(GIOWA)operator with Dice similarity coefficient,the combination prediction model based on GIOWA operator and Dice similarity coefficient is constructed,and the definition of optimal combination prediction is given.The model is applied to the prediction of PM_(2.5)concentration in Hefei.Monthly data of PM_(2.5)concentration and some core influencing factors during 84 periods from 2014 to 2021 are selected for analysis.The results of the combination prediction are compared with single prediction results,and the influence of parameter changes on weight of portfolio prediction,evaluation index of prediction ef-fect and Dice similarity coefficient of combination prediction model is analyzed.The empirical results show that the combination prediction model is better than the single prediction model.

关 键 词：组合预测 PM_(2.5)浓度 GIOWA算子 Dice系数 

分 类 号：X513[环境科学与工程—环境工程]