Application Research Of Mushroom Toxicity Prediction Based On Data Mining Technology

Mushrooms are one of the most common foods in our daily life,including many kinds.Different kinds of mushrooms play a positive role in soil improvement and drug research.Edible mushrooms are loved by many people because of their delicious taste.In addition to the edible type,there are many different kinds of mushrooms,and most of them are toxic.Eating poisonous mushrooms by mistake will cause physical diseases,threaten our health and life safety,and seriously endanger our life safety.At present,the research on identifying whether mushrooms are toxic is not deep enough.Most people rely on people’s experience and subjective judgment,which is not scientific and accurate;Or through component research,but this method needs to rely on professional experimental personnel and laboratories,with long cycle and high cost.It is obviously unrealistic to judge the toxicity of each kind of mushroom sample through chemical component research.In order to predict mushroom toxicity conveniently in daily life and with high accuracy of prediction results,this paper constructs a mushroom toxicity prediction model based on data mining technology,and develops a prediction application applet through software programming.Firstly,obtain the characteristics of mushrooms collected in life and laboratory,form sample data,and process the data set through data analysis,data preprocessing and characteristic engineering standard process to make the original data available to the computer;Then the logistic regression model in machine learning,decision tree classification model and neural network model in deep learning are introduced for modeling,and the dimensionality is reduced based on principal component analysis and linear discriminant analysis.Through a large number of experiments,repeated segmentation and scrambling of the data set and super parameter debugging,it is concluded that the data set selected in this paper is not suitable for dimensionality reduction,and then the grid search algorithm is used for parameter optimization to select the best parameter + penalty item.In addition,based on the idea of random forest out of bag prediction,this paper optimizes the decision tree algorithm and proposes DTOOB algorithm to eliminate the risk of over fitting;This paper attempts the most commonly used method in the field of data mining-association rules.Finally,the DTOOB algorithm which meets the research goal of this paper and performs well is selected.Finally,the program architecture is designed.The steps include program architecture design,database design,applet code writing,software testing and software deployment to the server.Put the program into daily life.Users can enter the program interface through wechat search function,and present the prediction results after completing information input.