Fungal Intelligence: Machine Learning Models Reveal Optimal Substrate Formulations for Maximizing Mushroom Yield

This research focused on applying a list of machine learning models to identify the best Substrate to increase the growth of a mushroom. In this study, three (3) distinct types of substrates were applied and analyzed. Substrate A (79 % sawdust, 1 % lime, 5 % molasses, 15 % rice bran), Substrate B (63 % rice straw, 10 % vermicast, 27 % decomposed sawdust), and Substrate C (50% rice straw, 30% sawdust, 20% forest topsoil). A total of 90 samples were cultivated and evaluated during the first harvest, detailing the growth performance of the mushroom in terms of weight, Height, and the number of stems. The growth outcomes from the collected samples were categorized using classification models, including decision trees, logistic regression, random forest, k-nearest neighbors, and support vector machines. The findings show that most models achieved high accuracy, ranging from 89 % to 100 %, which effectively distinguished them based on growth patterns. Among regression models, both yielded the best predictive performance with an${R}^{2}$value of 0.1. Results revealed that Substrate B produced the highest mushroom yield, attributed to its balanced composition and nutrient-rich properties from vermicast and decomposed organic matter. The study demonstrates that applying machine learning models with mushroom cultivation can enhance substrate optimization and yield forecasting. Lastly, this study would contribute to the agribusiness and agricultural sectors by developing new techniques for evaluating different substrate formulations, showcasing how machine learning can support sustainable, evidence-based approaches in agricultural biotechnology.