Prediction of Thermally Modified Wood Color Change after Artificial Weathering Based on IPSO-SVM Model

Abstract

The support vector machine (SVM) model was applied to predict the color change of heat-modified wood after artificial weathering. In order to improve the prediction performance, the improved particle swarm optimization (IPSO) algorithm was used to optimize the parameters of the SVM model, and an improved particle swarm optimized support vector machine (IPSO-SVM) model was established on the basis of the nonlinear descending weight strategy to improve the particle swarm optimization. To verify the performance of the established model, the MAE, RMSE, and R2 of the test set and training set were compared with the PSO-SVM model and the SVM model. Analysis of the results showed that compared to the PSO-SVM model and the SVM model, the IPSO-SVM model reduced the RMSE of the training set data by 49% and 72%, the MAE by 52% and 78%, the STD by 14% and 68%, the test set data by 6% and 24%, the MAE by 2% and 25%, and the STD by 22% and 29%, respectively. The results show that modeling studies using the IPSO-SVM model provide results showing that color changes in heat-modified wood after artificial weathering can be successfully predicted without expensive and time-consuming experimental studies.