Automated Determination of Mushroom Edibility Using an Augmented Dataset

Abstract

This paper studies methods and datasets for automated classification of mushrooms as edible or poisonous based on easily observable properties such as colors, textures, and dimensions of mushroom parts. The focus is on data-intensive methods that build upon recent work that has led to an augmented database of mushroom features. This dataset is studied in detail with the goal of explicating properties and easing further use of the dataset by others. The merit of the database features for the classification task is quantified using several metrics. Results quantify the accuracy and efficiency of classification using all and only a few of the features.