基于模糊C-均值聚类与支持向量机的PMV指标预测系统  

PMV index forecasting system based on fuzzy C-means clustering and support vector machine

作　　者：徐巍[1] 陈祥光[1] 彭红星[1] 刘春涛[1] 

机构地区：[1]北京理工大学化工与环境学院,北京100081

出　　处：《系统工程理论与实践》2009年第7期119-124,共6页Systems Engineering-Theory & Practice

基　　金：北京理工大学校基础研究基金(20070542009)

摘　　要：为了更好地预测室内热舒适度PMV指标,在分析模糊C-均值聚类方法与支持向量机方法的优势和互补性后,探讨了二者的结合方法,提出了一种基于模糊C-均值聚类预处理的支持向量机PMV指标预测系统。该方法把复杂的数据集看作多个群体的混合,每个群体采用单一的回归模型进行描述,使得大规模数据集的回归估计问题变成了一个多模型估计问题。将该系统应用于PMV指标预测中,与标准支持向量机方法相比,得到了较高的预测精度,从而说明了基于模糊C-均值聚类方法作为信息预处理的支持向量机学习系统的优越性。In order to commendably estimate indoor thermal comfort, advantage and complementarity of fuzzy C-means clustering algorithm （FCM） and support vector machine （SVM） is analyzed. A kind of SVM forecasting system based on FCM data preprocess is also proposed. In the proposed method, the large dataset is viewed as a mixture of multiple populations, and each population is represented by a single regression model. The problem of regression estimation for large dataset is viewed as a problem of multiple regression model estimation. In using forecasting PMV index, this approach has achieved greater forecasting accuracy comparing with the method of standard SVM. It is denoted that the SVM learning system has advantage with the information preprocessing based on FCM algorithm.

关 键 词：模糊C-均值聚类 支持向量机 室内舒适度 PMV指标 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]