Research On Knowledge Automatic Classification Algorithm For Mobile Health Applications

Nowadays,the development of mobile Internet and intelligent terminal is more and more fast,and it has penetrated into every corner of society.The new medical health information service model came into being.In the social background of the population is aging with chronic diseases complicated and the demand for health management is increasing,more and more mobile health applications have been spawned through combining mobile Internet technology and intelligent terminal technology.Whenever and wherever possible,people can obtain medical information they need from mobile devices.It can be more convenient and user-friendly to help people to manage their health through information services from these applications.However,mobile health information services must be located in the users how to obtain the knowledge immediately,effectively and accurately.In this regard,classification can largely solve the disorder of Internet information.But at present,the mobile health applications,especially the information service,the users don't trust enough about it due to the lack of certain industry standards and professionalism.However,the classification algorithms based on machine learning,which has become the mainstream algorithm,lack sensitivity to information and tend to overlook the relevance of information in terms of specialization and hierarchy of classification.In addition,the limitations of these algorithms in the development and application of mobile health are the long development cycle,low efficiency and so on.Therefore,this paper studies on the text classification algorithm in mobile health to improve the quality of information service module.In this paper,we construct a thesaurus in the single field of word co-occurrence with language model.Through the network of semantic relation thesaurus,we mining the semantic similarity between vocabulary and thesaurus to implement automatic text classification in the single field.The focus of this paper is how to extract effective information from the network resources to achieve classification,and how to set up the establishment and visualization of the semantic network.The main contributions are as follows:To extract domain keywords from a corpus in a domain.First,we analyzed the effects of various factors subject word text words become;then after the word segmentation splicing pretreatment,and three stage filtering mechanism,we finally mark and sort the chosen words to extract topic terms according to the weights.To establish a thesaurus of the field,then to reveal the semantic relation network between words from the thesaurus.We mainly researched on the semantic relationship between the thematic terms: first,set these terms as constituent elements of the thesaurus;after that,respectively analyze the inline relationship among the thesaurus and among the modifiers,and the correlativity between the terms and the modifiers.At last,the semantic complex network and its visualization are realized according to the above elements and relationships.A text classification algorithm based on word co-occurrence model is proposed.The main idea of the algorithm is to combine the thesaurus to analyze the similarity between the document characteristic words and the subject words,so as to map the similarity between the document and the category,and realize the classification of the text.The similarity is divided into two parts: the main part of calculating the co-occurrence degree between terms of thesaurus and text representation models was evaluated;and the other part which similarity was used the smoothing method to construct between modifiers and text representation models,as an auxiliary correction parameter,makes the classification have more semantic relevance to a certain extent.Finally,the performance of classification is kept at a high level.Compared with the SVM algorithm,the average accuracy is improved by 2%,especially in the professional category,the accuracy rate is obviously better.The classification algorithm is feasible in mobile health applications,which is more effective,more specialized and more intelligentized.