Predicting Porosity in Grain Compression Experiments Using Random Forest and Metaheuristic Optimization Algorithms.

Grain stored for long periods is highly susceptible to localized condensation, mold growth, and insect infestations, leading to significant storage losses. These issues are particularly acute in large-capacity bungalow warehouses, where food security concerns are even more pronounced. The porosity of grain piles is a critical parameter that influences heat and moisture transfer within the grain mass, as well as the ventilation of grain storage. To investigate the distribution pattern of bulk grain pile porosity in bungalow warehouses, this study employs machine learning (ML) techniques to predict grain pile porosity based on compression experiments. Four metaheuristic optimization algorithms-particle swarm optimization (PSO), gray wolf optimizer (GWO), sine cosine algorithm (SCA), and tunicate swarm algorithm (TSA)-were introduced to enhance the random forest (RF) algorithm, and five ML-based models (RF, PSO-RF, GWO-RF, SCA-RF, and TSA-RF) for predicting grain porosity were developed. The predictive performance of the five models was analyzed using error analysis, Taylor diagrams, evaluation metrics, and multi-criteria assessments to identify the optimal ML prediction model. The results indicate that the predictive performance of the four RF-based hybrid models surpasses that of the single RF model. Among these hybrid models, the TSA-RF model demonstrated the best predictive performance, achieving