Research On Multi-label Classification Algorithm Based On Label-Specific Features And Label Correlation

With the development of machine learning,single label learning has been unable to meet the needs of application in the real world.More and more researchers pay attention to the multi-label learning problem.Multi-label learning is a learning framework for studying the problem that each sample is associated with a set of class labels.Multi-label learning tasks can be seen everywhere in the real world.In multi-label learning,each sample is associated with multiple class labels and these class labels may also be associated with each other.With the development of multi-label learning,a series of multi-label learning algorithms have been proposed by researchers and applied to various research fields.The challenge of multi-label learning is to learn an effective classification model,which can predict a set of class label that may be associated with a new instance.The common strategy adopted by these multi-label classification algorithms is to predict all class labels using the same set of features.It is not the best choice as each label may be most relevant to its own characteristics.Based on these problems,this paper focuses on the two aspects of "constructing the label-specific features" and“mining the label correlation".The main research contents of this paper are as follows:1.We study the theories of multi-label learning problem,multi-label k-nearest neighbor algorithm,multi-label learning with label-specific features algorithm and label correlation problem,analyze the differences between different algorithms.Research on how to construct effective mechanism of label-specific features,a new knn multi-label classification algorithm based on label-specific features is proposed.Firstly,the algorithm preprocess the features vector of multi-label data and construct the most discriminative feature for each class.Then,the adjusted ML-KNN algorithm is used to classify on the acquired characteristics.In the experiment,we analyze the influence of the parameters on the accuracy of data classification,compare the difference performance of the algorithm with different values of k and analyze the difference of the accuracy between different algorithms.The experimental results show that the proposed mechanism of label-specific construction is effective.It plays a positive role in improving the accuracy of classification.2.The problem of how to make full use of the complex correlation between labels to improve the performance of multi-label classification algorithm is studied in this paper.Propose a new knn multi-label classification algorithm based on local positive and negative labeling correlation.Firstly,preprocess the features vector of multi-label data and construct the most discriminative feature for each class.Then,in the training stage,the PNLC algorithm constructs the positive and negative label correlation matrix by using the truth label of each k-nearest neighbor for all the training samples.Finally,in the test phase,the k-nearest neighbors and corresponding positive and negative pairwise label correlations for each test example are first identified,then we make prediction through maximizing the posterior probability.The performance of the algorithm on different datasets is analyzed and the classification accuracy is compared with the different algorithms.