Potential of machine learning algorithms for predicting the properties of medium-density fiberboard (MDF): preliminary results

Traditional quality control methods in the woodbased panel industry, especially for medium-density fiberboard, are insufficient to compete in the current market. In addition, due to the rapid growth of wood-based panel production and the need to provide competitive products in the market, there is an unprecedented need to explore new methods of quality control throughout the production process. Therefore, it seems necessary to use new quality control methods based on artificial intelligence and machine learning algorithms, because they have high predictive and optimization capabilities. The aim of this research is to develop suitable model to identify the most important and effective variables in the production process of industrial fiberboards and finally to predict the properties of the final product such as the bending strength (MOR) based on industrial data. For this purpose, the R software environment was used to implement the random forest algorithm to identify important variables. The performance of the model was evaluated using the coefficient of determination (R²) and the root mean square error (RMSE). The results showed moderate accuracy with an R² value of 0.49, which means that the model explained 49% of the variance of the dependent variable. The RMSE was 1.565, indicating a low prediction error. These metrics demonstrate the robustness and reliability of the random forest algorithm in managing complex data sets and producing accurate predictions.