摘要
针对建筑节能气候数据质量差的问题,提出了基于K-means聚类算法与BP神经网络相结合的方法对建筑节能气候数据进行清洗。首先,针对传统的K-means聚类算法对离群点的处理不足,通过最小二乘法设定阈值提高聚类的效率;接着,将聚类后的数据集作为BP神经网络的训练样本进行网络设计和训练;最后,得到属性之间的映射关系,检测出异常值并修正,从而实现对建筑节能气候数据的清洗。实验结果表明,所提的基于聚类和神经网络算法对建筑节能气候数据的有效清洗率达到93. 6%,从而提高后续建筑节能设计和能耗模拟的可信度。
Concerning the poor quality in building energy-saving climate data, a method based on K-means clustering algorithm and Back Propagation （BP） neural network was proposed to clean the building energy saving climate data. As traditional K-means clustering algorithm is not suitable for outlier processing, the clustering threshold was set by least square method, the clustered data were used as training samples to design and train a BP neural network, the mapping relationship between attributes was got, and outliers were detected and corrected, therefore energy-saving climate data cleaning was implemented. The experimental results show that the effective cleaning rate of the proposed clustering and neural network algorithm for building energy saving climate data is 93.6%, so as to improve the reliability of subsequent building energy saving design and energy consumption simulation.
作者
李昌华
卜亮亮
刘欣
LI Changhua;BU Liangliang;LIU Xin(School of lnformation and Control Engineering,Xi＇an University of Architecture and Technology,Xi＇an Shaanxi 710055,China)
出处
《计算机应用》
CSCD
北大核心
2018年第A01期83-86,111,共5页
journal of Computer Applications
基金
国家自然科学基金资助项目(61373112)
关键词
K—means聚类
BP神经网络
气候数据
建筑节能
数据清洗
K-means clustering
Back Propagation （BP） neLLral network
climate data
building energy saving
data cleaning
