Comparison of Wood Classification using Machine Learning

Abstract

Wood makes an essential contribution to human life and is used for various needs such as building materials, furniture materials, restorative materials, and so on. The economic value of each wood often varies significantly. Unfortunately, some woods have relatively high similarities and are difficult to distinguish for beginners. In this research, several models were built to detect several types of wood that have a fairly high similarity, namely: jeungjing wood (Paraserianthes falcataria), puspa wood (Schima wallichii) and suren wood (Toona sureni). Several classification algorithms are used to build classifiers, namely: the k−nearest neighbours, support vector machine, decision tree, random forest, and Inception-v3. The experimental results show that the accuracy level for the k−nearest neighbours, support vector machine, decision tree, random forest, and Inception-v3 are 82.00%, 90.00%, 69.33%, 90.67%, and 90,33%, respectively. The model produced by the Inception-v3 gives quite promising results and is suitable for practical use.