Classification of wooden wastes with machine learning approaches

Abstract

In this study, 200 wood waste samples from different origins were analysed by Inductive coupled plasma optical emission spectrometry (ICP-OES) and Inductively coupled plasma mass spectrometry (ICP-MS) for 11 elements (lead, cadmium, aluminium, iron, zinc, copper, chrome, arsenic, nickel, mercury and sulphur) that are likely to present in wood waste. In the study, the data as non-hazardous and hazardous was evaluated based on the standard (TS EN ISO 17225-1, 2021). Artificial neural network (ANN) and random forest (RF) analyses were then applied to better analyze and interpret the data. In this way, statistical separation of wood wastes as non-hazardous and hazardous was realized. Accordingly, it was shown that random forest analysis with an accuracy rate of 100% was better than artificial neural network analysis with an accuracy rate of 99%. Results suggested that wood wastes could be recycled and entered the production cycle in a way to contribute to the national economy or be incinerated with appropriate methods in bioenergy production in an environmentally friendly way which would be possible with the accurate classification of these wastes. In this study, the classification of wood wastes as hazardous and non-hazardous with 100% accuracy rate using ICP data with machine learning approaches, which is not encountered in the literature review.