The Study On Modeling And Regression Prediction Of Lumber Moisture Content Based On Dimensionality Reduction

Lumber Moisture Content is the important parameter to judge the dryness of lumber and the quality of wooden products. Nevertheless its data acquired are mostly redundantly and incomplete because of the complexion of the course of drying, interference factors that exist in dryness environment and the physical characteristics of lumber itself. Therefore, exploring the scientific and effective way is not only of research significance and practical value for analyzing the data of lumber moisture content’s change trend in the course of drying, but do help operators reasonably adjust the drying course and improve the level of drying process.This paper makes use of relative parameters from experiment data to establish multidimensional data collection according to the high correlation between lumber moisture content and the medium temperature and moisture in the course of drying. To solve the problem that the big difference of each feature quantity’s data level influences the regression analysis function, the method of data feature normalization is proposed to pretreat the data collection so as to unify each feature quantity’s data level of lumber dryness multidimensional data and improve the regression analysis function of the lumber moisture content. However, when making use of the data dealed with in that way to do multivariate regression analysis, there is still multicollinearity among prediction variables so that unstable factors exist in solution space. In order to avoid the problem above, this paper proposes principle components analysis (PCA) from dimensionality reduction to do the feature extraction for the normalized lumber dryness data. It results in the optimization for multidimensional data, the improvement of the quality of data and utilization rate regression and also the improvement of analysis efficiency of lumber moisture content.Support Vector Machine (SVM) has the characteristic of solving the problems of nonlinear, high dimension, and uncertainty. Because of lumber dryness experiment parameter collection is redundantly and uncertain, this paper establishes lumber moisture content regression model based on support vector machine for the experiment data of lumber moisture content, temperature, moisture to predict and analyze it. While doing regression prediction with SVM, the values of penalty parameter c and kernel function parameter g will directly influence the regression prediction function of support vector machine. To settle it, this paper utilizes three kinds of intelligent optimization algorithms, grid search, genetic algorithm, particle swarm optimization; and through comparison one of them can be chosen to apply in the value process of support vector machine parameter c and g of lumber moisture content.Through utilization of different data pretreatment and parameter optimization methods, model training and regression prediction based on lumber moisture content experiment support vector machine is operated. The result states that while doing the lumber moisture content analysis, the combination of feature normalization and PCA works as the data pretreatment method; genetic algorithm works as parameter optimization method of support vector machine data regression analysis. Moreover, established lumber moisture content model has great generalization ability and prediction accuracy so that the data analysis efficiency and its reliability in course of lumber moisture content detection are improved. This paper propose a new efficient research method for the establishment of lumber moisture content model and prediction, and promote the lumber scientific and technological research.