| Wood microfibril angle (MFA) has a very important relationship with the physical and the chemical nature of wood, at the same time, it has a strong connection with the anatomical character of wood, as well as the modulus and the bending strength. In this research, different parameter optimization methods were used for forecasting and analyzing the MFA of Chinese Larch based on near-infrared spectroscopy (NIR) technique and support vector regression (SVR).The major results of this dissertation are as follows:(1) SVR based on the non-heuristic parameter optimization algorithm-cross validation (CV) parameter optimization was used to predict the MFA of Chinese Larch, for the calibration set, the correlation coefficient (R) and the mean square error (MSE) were0.91and0.1834, while the R and MSE were0.93and0.1048, respectively, for the validation set. Results showed that wood MFA of Chinese Larch could be well predicted by using SVR in near-infrared spectroscopy based on CV algorithm.(2) SVR based on the heuristic parameter optimization algorithm-genetic algorithm (GA) was used to predict the MFA of Chinese Larch, for the calibration set, the R was0.95and the MSE was0.0380, for the validation set, the R was0.94and the MSE was0.0813. Results showed that GA parameter optimization method could be well used in the SVR prediction.(3) SVR based on the heuristic parameter optimization algorithm-particle swarm optimization (PSO) was used to predict the MFA of Chinese Larch, for the calibration set, the R was0.96and the MSE was0.0216, for the validation set, the R was0.96and the MSE was0.0288. Results showed that PSO parameter optimization method could be well used in the SVR prediction.(4) Three parameter optimization algorithms-CV, GA, PSO and their corresponding SVR modeling results were compared. Results showed that the parameter optimization algorithms based on CV, GA and PSO could be well used in the SVR prediction as well as satisfactory result of NIR models of MFA. However, models based on the GA and PSO were better than the model based on the CV. |
