摘要
城市用水量受到多重因素的影响,且各因素之间存在相关性。将逐步回归技术引入偏最小二乘(PLS)用水量预测模型影响因子的筛选过程,可对PLS回归建模过程进行改进,在保证拟合精度的条件下,有效解决了自变量间的多重相关性问题;同时实现测定指标的降维,达到了简化、精炼模型的目的。将所提理论和方法应用于某城市用水量预测中,运用R软件进行求解,并将耦合逐步回归的PLS模型与单一的PLS回归模型进行比较分析。结果表明,模型的拟合和预报精度较好。
This paper intends to present a newly developed urban wa- ter consumption prediction method based on a partial least-square model with a stepwise regression analysis procedure. Veracious forecast of water consumption is very significant for the urban water resource use in a well-planned way. We also know that the urban water consumption is actually influenced by numerous factors, among which there stands the problem of relativities. To solve the problem of relativities, we have introduced a stepwise regression method into the partial least-square (PLS) regression analysis process for choosing the likely influential factors. The advantages of the two methods can be synthesized to improve the accounting power so that it can be used to predict the urban water consumption. Moreover, the improvement may allow the problem of mutual influence among the independent variables to be solved effectively without decreasing its accuracy. What is needed to do is to simplify the given model and refine it through dimension reduction. In so doing, the model innovated by us can be worked out by using a software named R. And make a com- parative analysis between the simulated results of PLS model with the stepwise regression analysis by using the main influential factors and traditional PLS model with all its independent variables. The research shows that the precision of the former model in simulation can be higher than the latter, for it has enhanced its accounting power. Thus, the improved method can be made suitable for dealing with the problems of more variable numbers. The validating analysis on the water consumption forecast of the city proves that the forecast results are in conformity with the actual de-facto, which demonstrates its usefulness for water resource planning and management.
出处
《安全与环境学报》
CAS
CSCD
北大核心
2012年第4期170-173,共4页
Journal of Safety and Environment
基金
国家自然科学基金青年基金项目(71101104)
关键词
水文学
用水量预测
逐步回归分析
PLS模型
R软件
hydrology
water consumption prediction
stepwise regression analysis
partial least-square model
softwareR