Multi-label classifier chain image recognition method based on PageRank and mutual information

The invention relates to a multi-label classifier chain image recognition method based on PageRank and mutual information, and belongs to the technical field of multi-label classification in data mining. According to the method, firstly, mutual information is used for calculating correlation between tags and between tags and features, then, the features are subjected to analogy to form webpages, the link relation of the webpages is replaced with the correlation relation of the tags, finally, by means of the thought that the webpages are sequenced according to the webpage access probability in the PageRank algorithm, a tag training sequence is generated, a classifier chain is generated, and the webpages are classified. Therefore, the image is identified. According to the method, the tag correlation is fully considered, and a proper tag sequence is sought, so that better image classification performance is obtained. 本发明涉及一种基于PageRank和互信息的多标签分类器链图像识别方法，属于数据挖掘中多标签分类技术领域。本发明首先利用互信息来计算标签与标签之间和标签与特征之间的相关性，然后，将特征类比成网