Improvement And Research Of Multi-label Learning Algorithm

The problem of multi-label classification is an important subject of data classification research in the field of data mining.In the era of big data,the explosion of data volume and the structure of data labeling are becoming more and more complex.Multi-label classification problems are widespread in the real world.How to find a fast and effective multi-label classification algorithm in the field of data mining has become a hot issue,and the rise of research on multi-label data mining is increasingly highlighting its value.This article focuses on the problem of multi-label classification,the main work is as follows:First,it summarizes and analyzes the current algorithms applied to multi-label classification learning: methods based on problem conversion strategies and methods based on algorithm conversion strategies.The classification principle,classification steps,advantages and disadvantages and adaptation conditions of each type of algorithm are elaborated in detail,and several of these algorithms are simulated and compared on the data set.Secondly,I deeply studied the multi-label classification algorithm ML-KNN based on KNN(K Nearest Neighbors).To increase the value of the posterior probability,the ML-KNN algorithm is likely to cause misjudgment of labels in the case of uneven data distribution.An improved new multi-label classification algorithm is proposed: an improved ML-KNN algorithm with increased posterior probability to improve the accuracy of the classification algorithm.Then,a penalty function is introduced based on the ML-KNN multi-label classification algorithm with increased posterior probability to further improve the performance of the classifier.Through the classification experiment on the data set,the simulation shows that the prediction performance of the improved algorithm in this chapter is better than the ML-KNN algorithm as a whole.Finally,a new multi-label lazy learning algorithm-IMLLA(An Improved Multi-Label Lazy Learning Approach)algorithm is deeply researched.In this paper,based on the IMLLA algorithm,different size parameters are used to determine the neighbor samples corresponding to each class,and more effective information than the marker count vector is extracted from the neighbor samples to assist the classification process.The improved algorithm was used for classification experiments on the data set,and compared with the IMLLA algorithm in detail.The results show that the improved IMLLA algorithm is of great benefit in improving the performance of the multi-label lazy learning algorithm.