期刊名称：Case Studies in Business, Industry and Government Statistics
印刷版ISSN：2152-372X
出版年度：2007
卷号：1
期号：1
页码：54-60
出版社：Bentley University
摘要：This case study illustrates the prediction of energy savings at a chemical plant after energy saving improvements were implemented at the plant. The study shows the danger in fitting prediction equations by automatic proce-dures, and how over fitting an equation can actually increase the error of prediction. Errors of prediction esti-mated from the data used to find the equation are shown to be optimistic. Data splitting is demonstrated as a method to validate a prediction equation, and a more reliable prediction equation than the one found by the automatic procedure was developed along with a more realistic estimate of prediction error.
关键词：Least-Squares Regression; Overfitting; Data-Splitting; Mean Square Prediction Error.
