Classification of Strength Properties of Commercially Important Wood Types Grown in the United States by Machine Learning

Authors : Kenan Kılıç

Pages : 66-80

Doi:10.17474/artvinofd.1717771

View : 65 | Download : 212

Publication Date : 2026-02-24

Article Type : Research Paper

Abstract :Commercially important wood species grown in the United States are divided into hardwood and softwood based on their mechanical properties. Comparative analysis is conducted by optimising six different machine learning algorithms: SVM, XGBoost, Random Forest, Logistic Regression, KNN, and Decision Tree. Preliminary processes such as completing missing data, coding categorical data, and standardisation are applied to the dataset to make it suitable for machine learning algorithms. Experiments were conducted using the stratified 10-fold cross-validation method. Hyperparameter optimisation was performed with GridSearchCV. The SVM algorithm provides the best accuracy with 96.90%. This model is followed by XGBoost with 95.13% accuracy and an AUC of 0.9891, followed by Random Forest with 94.25% accuracy. Logistic Regression performs with 90.27% accuracy, Decision Tree with 90.71% accuracy, and KNN with 88.05% accuracy. Results show that kernel-based (SVM) and ensemble-based (XGBoost, RF) models provide higher classification performance than linear and instance-based models. These models have the potential to improve wood quality control processes, increase resource efficiency, and contribute to sustainable forestry practices.
Keywords : Makine öğrenme, Odun sınıflandırma, Odun özellikleri, Geniş yapraklı ağaç odunu, İğne yapraklı ağaç odunu, Mekanik özellikler

ORIGINAL ARTICLE URL