Research on Mushroom Classification Based on XGB Technology

Abstract

The mushroom classification problem, as a typical binary classification problem, has become a widely studied object in the field of machine learning. Traditional mushroom classification methods typically rely on manual feature extraction and rule-based criteria establishment, which are often susceptible to human factors, resulting in relatively low classification accuracy. With the development of machine learning technology, especially the emergence of ensemble learning methods, based on various machine learning models, particularly tree-based models, it is possible to efficiently distinguish edible mushrooms from poisonous ones. This article focuses on discussing the application of eXtreme Gradient Boosting (XGBoost) classifier in mushroom classification. This paper compares the performances of different classification models, including Random Forest (RF), Gradient Boosting Machine (GBM), and XGB classifier. It demonstrates the advantages of XGB-based classification methods in mushroom classification, using the Matthews correlation coefficient (Matthews Correlation Coefficient, MCC) as the primary evaluation metric. Additionally, we further explored in depth the crucial roles of data preprocessing, feature selection, and hyperparameter tuning strategies in model optimization, and ultimately determined the best application practices of XGB in mushroom classification problems.