Research On Multi-Label Classification Algorithms And Their Applications

In recent years,we have entered the era of data explosion,with the data growth and data storage capacity enhancement,so that we can get different forms of data sources and stored in the information base.By analyzing and digging the data stored in the information base,it can effectively extract the valuable information and help the decision-making of business,scientific research and other activities.As a form of data analysis and mining,it can extract a model that can describe important data sets for predicting the discrete categories of data objects.The classification problem can be divided into single-label classification and multi-label classification according to the number of different types of labels.Most of the problems we face in traditional supervised learning tasks are single-label classification issues.However,in many classification tasks each sample needs to be associated with multiple category labels,such as in text categorization(associated with multiple types Joint book)and medical diagnosis(eg,diagnosis of multiple patients with disease).And these problems are single-label classification technology can not be solved,therefore,in recent years,the study of multi-label classification has been widely concerned by scholars at home and abroad.At present,the algorithm to solve the multi-label classification problem does not achieve satisfactory results,and the researchers also try to improve the classification performance by considering the label relevance and the integration through the classifier.Through the research and analysis of the existing multi-label classification algorithm,the RAkEL multi-tag classification algorithm[14]is a more efficient multi-label classification algorithm using classifier integration technology.However,because of the sub-There are still room for improvement in the classification effect,such as randomness and lack of relevance information of the label.In this thesis,a new multi-tag classification algorithm based on RAkEL algorithm is proposed by applying label correlation and classifier integration technology to the unified framework.Compared with other multi-label classification algorithms,the proposed method is superior to other multi-label classification algorithms in comparison with RAkEL multi-label classification algorithm.In addition,this thesis also explores the application of multi-label classification algorithm in the field of recommendation system.In the context of recommended systems,context-aware recommendation systems use context context information to further improve recommended accuracy and customer satisfaction,but the question of how context-aware systems are still recommended is to recommend project collections to target users.In this article,we will examine another recommended scenario in real life:when a user selects a project,we recommend the most appropriate application context,the context,for example,a user has decided to see a movie,Where he needs the advice(home or theater),and who(family or friends)to watch will get a better viewing experience.The scenario recommendation not only recommends the most suitable scenario for a user to consume a project to improve the consumer experience,but also to assist the user in making a project selection decision.In this thesis,we solve the problem of multi-label classification,and then we improve the multi-label classification algorithm and get the method applicable to the situation recommendation problem.Experiments were performed on two field datasets.Experimental results show that this algorithm can give personalized suggestions,and in a number of indicators better than the original algorithm.