Application Research Of Spatio-textual Publish/Subscribe System Based On Topic Model

With the rapid development of the mobile Internet and the commercialization of5 G technology,location based services(LBS)are becoming more and more popular.Map navigation,community e-commerce and mobile social networking,these LBS services not only facilitate people's life,but also produce massive spatial data.As one of the most important technologies used in LBS services,spatio-textual publish/subscribe technology plays a vital role in the push,distribution,and query of massive spatial data.This thesis proposes an efficient spatial index structure and top-k matching algorithm by studying spatial index and text index in spatio-textual publish/subscribe technology,and designs and implements a spatio-textual publish/subscribe system based on the research content.The main work of this article and innovations include the following:(1)In order to solve the problem that most of the researches in the clustering construction of R-tree spatial index only randomly specify or calculate the Euclidean distance between spatial data;the connection between spatial data is only the association of Euclidean distance,this thesis proposes a dynamic k-value clustering algorithm based on k-means++.This algorithm introduces Latent Dirichlet Allocation(LDA)into R-tree construction,and optimizes the selection of clustering objective function and clustering center based on LDA.Compared with traditional clustering R-tree construction algorithm,this algorithm improves the index efficiency of R-tree and reduces the overlapping degree between nodes.(2)In order to solve the problem that most of the researches in the spatio-textual publish/subscribe technology,most researches only use keywords as text query conditions,ignoring the topic relevance of spatial text objects;the research on spatio-textual publish/subscribe is limited to the construction of spatial text object text index,and fails to consider the optimization of spatial index and efficient pruning strategy.This thesis proposes a spatial text publish / subscribe top-k query algorithm based on topic model.Firstly,the algorithm uses the topic R-tree after construction,constructs the rated topic R-tree through the topic probability of node reservation,and optimizes the spatial index;secondly,it reduces the redundant calculation by topic pruning and upper bound function pruning,and improves the efficiency of spatio-textual publish/subscribe top-k query.(3)In the way of back-end separation,the spatio-textual publish/subscribe system is designed and implemented,and the system development environment configuration,architecture design,function module design and specific function implementation are comprehensively described.The system uses the popular front-end and back-end open source framework.The system functions mainly include system statistics,new subscriptions and events,management of subscriptions and events,and visualization of top-k matching information.