Ensemble Learning based Classification of Edible and Poisonous Agaricus Mushrooms

Agriculture has a crucial role in sustaining human populations in numerous countries globally, particularly in economies that are still in the process of development. Mushroom growing has emerged as a prominent agricultural practice in contemporary times. The primary and essential element in the practice of mushroom growing is the identification of mushroom damage. Mushrooms have the potential to be contaminated by various microorganisms such as bacteria, pests, and pathogenic molds. These contaminants can lead to significant crop losses and pose health risks to consumers. Mushrooms can be classified into two categories based on their health implications: those that are deemed safe for consumption, referred to as healthy or edible mushrooms, and those that pose a risk to human health if ingested, known as unhealthy or dangerous mushrooms. Various strategies for mushroom categorization exist; however, these methods are predominantly manual and can be time-consuming, particularly when dealing with large quantities of items that need to be categorised. The utilization of machine learning and deep learning technologies has grown prevalent in the classification of agricultural products. This study involved an analysis of 8124 data samples from the UCI repository Mushroom dataset for the purpose of classification. This dataset was originally contributed by The Audubon Society Field Guide to North American Mushrooms. This study employs various approaches, namely Random Forest, Bagging, Gradient Boosting, Adaboost, and XG Boost. Among the several approaches considered, Adaboost exhibited the highest performance, achieving an accuracy of 95.35% in a 10-fold cross-validation.