基于改进随机森林回归模型的木材染色配方预测研究  

Research on Predicting Wood Dye Formulation Based on an Improved Random Forest Regression Model

作　　者：管雪梅[1] 何中生 杨渠三 GUAN Xuemei;HE Zhongsheng;YANG Qusan(College of Machinery Electricity,Northeast Forestry University,Harbin 150040,Heilongjiang,China)

机构地区：[1]东北林业大学机电学院,黑龙江哈尔滨150040

出　　处：《西北林学院学报》2024年第6期231-237,共7页Journal of Northwest Forestry University

基　　金：国家自然科学基金(32171691);哈尔滨市科技局制造业创新人才项目(CXRC20221110393);中央高校项目(257202BF02)。

摘　　要：为提高木材染色计算机智能配色的准确性和实用性,选择(樟子松)单板为染色材,以随机森林回归模型为基础,先用随机搜索算法(RS)寻找模型最优参数,再引入主成分分析法(PCA)进行特征提取,并以配方的均方误差作为评价指标。与固定参数的随机森林回归模型(RFR)模型及其他模型相比,优化后的模型(PCA-RS-RFR)配方均方误差为0.00491,相较于固定参数RFR(0.00695)、RS-RFR(0.00618)等具有明显优势,运行效率也较出色。In order to enhance the accuracy and practicality of computer-aided color matching for wood staining,Pinus sylvestris var.mongolica veneer was selected as the staining material.Building upon the random forest regression(RFR)model,the study first utilized the random search(RS)algorithm to identify the optimal parameters for the model,followed by the incorporation of principal component analysis(PCA)for feature extraction.The mean squared error of the formula was used as the evaluation metric.Compared with the fixed-parameter RFR model and other related models,the mean squared error of the formula for the optimized model(PCA-RS-RFR)reached 0.00491,which notably outperformed the fixed-parameter RFR with an error of 0.00695 and the RS-RFR with an error of 0.00618.The operational efficiency is also quite excellent.

关 键 词：木材染色配方 随机搜索算法 随机森林回归模型 反射率曲线 

分 类 号：TS652[轻工技术与工程]