特征提取和小样本学习的电力工程造价预测模型  

Cost forecast model for power engineering based on feature extraction and small-sample learning

作　　者：彭光金[1] 俞集辉[1] 韦俊涛[1] 杨光[1] 

机构地区：[1]重庆大学输配电装备及系统安全与新技术国家重点实验室,重庆400030

出　　处：《重庆大学学报（自然科学版）》2009年第9期1104-1110,共7页Journal of Chongqing University

基　　金：重庆市自然科学基金资助项目(CSTC2006BA6015);国家电力公司科技项目(04207520070603)

摘　　要：通过特征提取和小样本学习的结合,提出一种全新的基于混合算法的电力工程造价预测模型。利用主成分分析对原始指标进行预处理,消除原始指标之间的相关性,并提取潜在的综合独立指标,将新指标作为输入集构造基于最小二乘支持向量机的预测学习模型,将其预测结果和神经网络模型预测对比分析。并通过不同主成分数目预测结果的比较,确定最优的主成分个数,达到理想的预测效果。实例预测结果表明:该方法可以有效提取原始指标的信息量,在小样本学习方面表现突出,能够达到期望的预测效果。A novel power engineering cost forecast model was proposed by combining feature extraction and small-sample learning. The initial data was preprocessed with principal component analysis to remove the correlation among the original indexes and get the potential independent indexes. The new indexes acted as the input set to build a new forecast model based on least squares support vector machines. The results of this model were compared with the forecast results getting from artificial neural network. By comparing the forecast results with different principal components number, the optimal number was determined to achieve the desired forecast effect. The prediction results indicate that the method can extract the feature of initial data effectively and is good at small-sample learning . The expected forecasting results can be reached.

关 键 词：电力工程 预测模型 主成分分析 最小二乘支持向量机 小样本学习 

分 类 号：TM743[电气工程—电力系统及自动化]