Partial least squares (PLS) regression and its application to coal analysis

Abstract

Instrumental chemical analysis methods use the relationships between a signal obtained and a property (generally a concentration) of the system under study. The study and applications of these relations is known as chemometrics, a discipline of intense development, with ample applications in chemical and process industry and in environmental studies. The method of partial least squares (PLS) is one of the most used in chemometrics. This method is closely related to principal components regression (PCR). PLS have theoretical and computational advantages that have led to a great number of applications. The numbers of Internet sites referring to PLS are hundreds of thousands. Here, we give the fundamentals and show an application to prediction of coal properties from mid-infrared data, with the purpose of developing fast, non-destructive methods of analysis for these materials.

Keywords

• PLS
• multivariate regression
• chemometrics
• coal analysis