Recognition And Research Of Poisonous Mushroom Based On Machine Learning

Because many types of mushrooms are poisonous,in recent years,it has become a matter of close concern to harvest wild mushrooms in the wild and forests for food poisoning.Correctly and quickly determining whether mushrooms are poisonous is the key safety issue of their picking activities,which has been widely spread in various countries.This paper designs an application based on mobile devices,which can determine the toxicity of mushrooms quickly and conveniently.This method supports the use of the mushroom toxicity detection recommendation system of mobile devices to determine the toxicity of mushrooms,and is easy to be used by users.This paper mainly has the following three parts:1.Mushroom data feature optimization selection.Each sample in the standard mushroom data set selected by this project has 1 class tag and 22 features,but these 22 features have different influences on the class tag obtained.In this paper,CART algorithm is used to optimize and select the five most critical attributes,thereby improving the accuracy and operating efficiency of the algorithm.2.Selection of optimal algorithm.In this paper,the mushroom data set respectively the naive bayes,decision tree and support vector machine(SVM)three kinds of commonly used machine learning algorithms,respectively the three kinds of algorithm accuracy,F value,10 fold cross index to evaluate aspects of analysis,the algorithm accuracy and efficiency under the premise of give attention to two or morethings,comparative analysis to select the decision tree algorithm was applied to the mobile terminal identification application;3.Design of mobile terminal.Subject to choose Android studio development environment and a Python program developed respectively mobile terminal and server application model prediction,the mushroom characteristic data and predicted results in the cloud storage,at the same time using the cloud API interface and the client,the server to send and receive data,eventually form a simple and easy to use mushroom toxicity identification proposal system.Through the optimization of mushroom data feature selection and three kinds of machine learning algorithm evaluation,comprehensive considering the accuracy and efficiency requirements,design the client-the cloud-poisonous mushroom identification application server architecture,that no matter when and where the user can quickly predict the mushroom is poisonous,has certain practical significance.