Classification of Mushrooms to Detect their Edibility Based on Key Attributes

Abstract

Mushroom is found to be one of the best nutritional foods with high proteins, vitamins and minerals. It contains antioxidants that prevent people from heart disease and cancer. Around 45000 species of mushroom are found to be existing in the world-wide. Among these, only some of the mushroom varieties were found to be edible. Some of them are really dangerous to consume. In order to distinguish between the edible and poisonous mushrooms in the mushroom dataset which was obtained from UCI Machine Learning Repository, some data mining techniques are used. Weka is a data mining tool that has various machine learning algorithms which can be used to pre-process, analyse, classify, visualise and predict the given data. Thus in order to select the attributes that helps in the better classification of mushrooms, Wrapper method and Filter method in Weka are used to identify the best attributes for the classification. The attributes ‘odor’ and ‘spore_print_color’ were chosen to be the best ones that contributed to the better classification of edible and poisonous mushrooms. After the identification of the key attributes, classification is performed and decision tree is constructed based on those attributes and its Precision, Recall and F-Measure values are analysed.