Ontology Learning Similarity Algorithm In Sparse Vector And Multi-dividing Setting

Under the current background of big data,the data model and the corresponding algorithm can better express the link between data in its storage,management and calculation,and thus it can meet structured storage and management of big data needs.Therefore,as a kind of concept structured storage,sharing and management model,ontology has drawn more and more attention from the researchers in various disciplines,and has become one of the key problems in the study of big data and information field in recent years.Ontology graph can be used to represent the concept structured storage of ontology,where each vertex represents a concept,and the edge between vertices represents some relationship between concepts(such as implicit affiliation,etc.).In addition to structured storage and representation,ontology model will help scholars calculate,analyze the statistics,and reason on the concept information and obtain some relevant conclusions.The algorithm on ontology focuses on the relationship between data,and thus the similarity computation between concepts(ontology graph vertices)is the core algorithm of ontology engineering applications.Under the background of big data,the concept of ontology graph contains a lot of information,and graph structure will also be extremely complicated,so that the traditional heuristic method to design the ontology concept formula is difficult to do data processing under the background of big data.Hence,more and more researchers pay great attention to ontology concept similarity calculation method through machine learning tricks,and it has become the mainstream in the field of ontology.The structure of most ontology graph is similar to the tree structure,the concept of the corresponding vertex in each branch often represents a large class,and thus multi-dividing algorithm is suitable for ontology learning.This paper mainly used mathematical tools to analyze the theoretical part of ontology algorithm.We get the mathematical characteristics of optimal ontology function in the setting of multi-dividing.The results show under the condition that the ontology graph structure previously determined,the optimal ontology function is decided by the ontology classification of vertices in each branch,which implies that it's directly determined by the graph structure.Furthermore,we analyze the multi-dividing ontology algorithm in the setting that loss function is convex function,and similar conclusion is yielded.In addition,by means of mathematical analysis and derivation,we design an algorithm for computing ontology similarity based on ontology sparse vector learning.It is characterized in high dimensional data representation of ontology concept information to extract the key information and effectively reduce the dimension.