Application And Improvement Of Partial Least Squares Regression Algorithm

The main research direction of this paper is to use the traditional regression method represented by partial least squares regression model and stepwise regression to establish a mathematical model.Because the amount of data cannot reach an excessive quantity and the information contained is complex,the traditional method is trapped.The limitations of the algorithm lead to the accuracy of the prediction results being affected by the multiple correlations between the independent variables in the modeling process.A new statistical method of multivariate data analysis-partial least squares regression is proposed.Partial least-squares regression(hereinafter referred to as PLSR),after verification,the method can obviously solve the problem of model prediction accuracy caused by the small number of experimental samples and the complexity of sample data parameter data.In the first chapter,the paper first introduces the difficulty of the data model with least squares regression as an example when the amount of data is small and the parameters are too many.The second chapter mainly introduces the modeling method of traditional partial least squares regression model and the corresponding formula derivation,and explains the multiple correlation problems encountered by other algorithms,as well as the main multi-correlation checking methods.An important factor in the least squares regression model.The third chapter proposes the problems that may exist when applying the traditional PLSR method,and proposes its own improvement strategy.The parameters are processed by orthogonal projection.This method can effectively process related parameters in the initial stage of modeling to remove.Redundant information among them,thus effectively improving the accuracy of the model.The fourth chapter is based on the background of the BOXA X-ray fluorescent taste analyzer in predicting the elemental content of the pulp.It is proposed to apply the improved PLSR method,and use Matlab to program the experimental model,and according to the spectrum analyzer.Obtain the band diagram corresponding to the Pb element in the zinc concentrate slurry,and the corresponding laboratory real value,establish a mathematical model corresponding to the Pb content prediction,and at the same time substitute the data into other comparison models,mainly to evaluate and compare the absolute error of the model.Relative error,according to this demonstration,the improved PLSR does play a role in reducing the error in the application of this problem.The fifth chapter combines and summarizes its own improved algorithm and proposes corresponding development prospects for the algorithm.There are two main innovations in this paper: 1.The application of orthogonal projection is used to process the parameter data needed in the modeling process to reduce the redundant information caused by mutual influence.2.Apply the proposed method to the actual production environment to improve production efficiency and save the corresponding labor cost.