基于正态信息扩散的新多变量模糊时间序列模型研究  

A New Multivariable Fuzzy Time Series Model Based on Normal Information Diffusion

作　　者：薛晔[1] 李肖肖 付恒春 Xue Ye;Li Xiaoxiao;Fu Hengchun(College of Economics &Management,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]太原理工大学经济管理学院

出　　处：《统计与决策》2018年第24期20-24,共5页Statistics & Decision

基　　金：国家自然科学基金青年项目(41101507);山西省高等学校人文社会科学重点研究基地项目(2017314);山西省高等学校优秀青年学术带头人支持计划项目(晋教财[2014]7号);山西省软科学项目(2017041025-2)

摘　　要：针对目前多变量模糊时间序列研究中小样本问题,文章基于信息扩散技术构建一个多变量模糊时间序列模型,讨论信息扩散系数的变化对模型精度的影响,进而选取2006-2016年人均GDP、能源消费总量的样本数据对中国SO2排放量进行预测,且与马尔可夫模型运行结果进行比较。结果表明:所建模型可以弥补小样本的不足;不同方式确定的信息扩散系数对模型精度存在影响,其中基于两点择近原则确定的h0最优;所建模型的平均绝对误差与平均绝对百分误差率均小于马尔可夫模型,即具有较高的预测精度。Aiming at the small sample problem in multivariable fuzzy time-series study, this paper constructs a multivariable fuzzy time series model based on the information diffusion technique, discusses the influence of information diffusion coefficient on model accuracy, and then selects the sample data of per capita GDP and the total energy consumption from 2006 to 2016 to predict SO2 emissions in China. Finally the paper also makes a comparison with the performance results of Markov model. The result shows that the proposed model can make up for the shortage of small samples; the information diffusion coefficients determined in different ways have an impact on the model accuracy, among which the [h0] determined based on the principle of two-point proximity selection is optimal; the mean absolute error and mean absolute percentage error of the proposed model are both lower than that of Markov model, thus the proposed model has a relatively higher prediction accuracy.

关 键 词：信息扩散 模糊信息推理 多变量模糊时间序列 二氧化硫排放量 

分 类 号：O159[理学—数学] C81[理学—基础数学]