The Axiomatic Fuzzy Sets Based Ensemble Learning Decision Tree

With the advent of the era of intelligence and the development of big data processing technology,data information is no longer just a message concept,but an intangible treasure of value.In the wave of the times,data mining technology has become an irreplaceable role in the social field,which playing an important role in financial analysis,medical diagnosis,military equipment and other fields.The principle of the decision tree algorithm is clear and simple,and the classification performance is stable,meanwhile the classification result is high in accuracy,and it also has good adaptability to different types of data.However,the general decision tree method cannot give a classification explanation that is easy for humans to understand,so it often leads to understand the classification result but cannot give a corresponding description;in addition,a single decision tree may produce accurate classification due to the randomness or imbalance of the data.The phenomenon of extremely low rate.The axiom fuzzy set theory can flexibly express the rules behind the data using fuzzy sets through algebraic operations and logical combinations;the integrated learning algorithm uses the respective classification results and scores given by multiple learners,and finally summarizes the results after calculations As a result,performance is more stable and accuracy is improved.Combining the Axiom Fuzzy Set theory and ensemble learning algorithm,this dissertation proposes an ensemble learning decision tree model based on axiomatic fuzzy sets,improves the node splitting index of the fuzzy decision tree classifier,and proposes a weighted fuzzy information gain rate.The uniform data classification effect has been greatly improved;the AFS structure of the Axiom Fuzzy Set theory is improved,and the bitmap storage structure is increased,which improves the efficiency of the algorithm;combined with the Axiom Fuzzy Set framework,an appropriate fuzzy set is given for each category description,so that the classifier has good interpretability,people can intuitively understand the meaning behind it;combined with the Adaboost integrated learning model,the data is reduced by attributes,and the integrated decision tree model is constructed to improve the accuracy of classification.The proposed algorithm model is cross-validated on the UCI data set with other classic classification algorithms.The experimental results show that the algorithm model in this dissertation has a higher accuracy rate on most data sets,which explains the model in this dissertation.Effectiveness.In addition,a network online classification system based on the Springboot and Vue framework is built through the frontend and back-end separation technology,and the system can be used to analyze data conveniently and efficiently through the network.