Predictive Modeling of Agricultural Biomass Paper Production using Ma¬chine Learning Approaches

Abstract

The transformation of biomass into high-value materials presents a promising solution for sustainable development. Accurate prediction of material properties based on biomass composition and processing parameters is critical for optimizing production efficiency and product quality. This study explores the application of multiple machine learning algorithms— Linear Regression, Decision Tree, Random Forest, and Support Vector Regression (SVR)—to predict material characteristics derived from biomass inputs such as cellulose, lignin, hemicellulose content, pulping time, and energy consumption. A dataset based on experimental values was used to train and evaluate the models. Among the tested approaches, the decision tree model showed the highest prediction performance (R² = 1.000); however, the limited dataset size may have contributed to possible overfitting followed by Linear Regression (R²= 0.987) and Random Forest (R² = 0.976), while SVR showed limited performance (R² = 0.081) due to the small dataset size. The results highlight the effectiveness of tree-based and linear models in accurately modeling the complex interactions between biomass composition and processing parameters. This research underscores the potential of machine learning techniques in advancing biomass valorization strategies and optimizing sustainable material production. The study demonstrates the feasibility of machine learning-assisted prediction for biomass-based paper production, while highlighting the need for larger datasets and industrial-scale validation.