摘要
异常数据是对建筑能耗预测结果存在影响的主要因素,结合回归分析特点,提出基于回归分析的建筑能耗异常数据识别算法,实现多属性建筑能耗异常数据的可靠识别。采用梯度提升回归树挖掘建筑能耗数据,通过优化聚类中心和数据特征融合,获取差异化属性的数据特征,将其作为自回归模型的输入,结合残差平方和理论,通过判断输入数据是否为异常数据,完成建筑能耗异常数据识别。测试结果表明,上述算法可获取数据集中不同数据特征分布集,实现不同属性的特征分类聚类,实现多属性数据中已有的异常数据和引入的异常数据的识别,保证异常数据的可靠识别。
Abnormal data influence the prediction results of building energy consumption.In this regard,this paper puts forward a building energy consumption abnormal data recognition algorithm based on regression analysis,in order to realize the reliable recognition of multi-attribute building energy consumption abnormal data.Firstly,gradient lifting regression tree was used to mine building energy consumption data.Secondly,the optimized clustering center and data features were combined to obtain the data features of differentiated attributes.Then,the data features were used as the inputs of autoregressive model.Concurrently,the sum of squares theory of residuals was also introduced.Then,the abnormal data in the judgment input data were judged,and finally,the identification of abnormal data of building energy consumption was achieved.The results show that the algorithm can obtain different data feature distribution sets in the data set,achieving the feature classification and clustering of different attributes and the identification of existing abnormal data and introducing abnormal data in multi-attribute data,thus ensuring the reliable identification of abnormal data.
作者
赵山
苏一帆
ZHAO Shan;SU Yi-fan(School of Civil Engineering and Transportation,North China University of Water Resources and Electric Power,Zhengzhou Henan 450045,China)
出处
《计算机仿真》
北大核心
2022年第5期499-503,共5页
Computer Simulation
关键词
回归分析
建筑能耗
异常数据识别
差异化属性
残差平方和
Regression analysis
Building energy consumption
Abnormal data identification
Differentiation attribute
Sum of squares of residuals
