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.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
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)
