Research On Spatio-Temporal Keywords Query Algorithm For Massive Short Texts

With the development of mobile social networks and location-based services,time information,location information and short text information are getting closer connection.The comprehensive processing and analysis of information with time and location attributes has important application requirements.The increasing number of smart phones and applications has led to an expanding number of users.The short text with time and location attributes is in an explosive growth.How to efficiently find valuable information that meets the diversity needs from massive data has attracted more and more experts and scholars' attention.Massive short texts spatio-temporal keywords query came into being,which holds has important theoretical and profound significance and application value.The importance of the time dimension of traditional spatial keywords query algorithms is ignored.Moreover,it doesn't work well on large-scale data,unable to meet the performance requirements of users' query.On the other hand,the existing query algorithms are not suitable for the social network application environment,the ignorance of social relationships between users leads to a poor quality of query result.To solve the above problems,improvement and innovation applied to the short text spatio-temporal keywords query algorithms.The main research works are as follows?(1)According to the practical application demands,this thesis extends the classical spatial keywords query problem by adding the time dimension.The short text spatiotemporal keywords query model is designed,which combines time,space and short text information tags.At the same time,efficiency of short text space-time keyword query in massive data environment is improved,and designs a query framework based on Map Reduce computing model.The query framework processes include: 1)The MultiVersions Spatio-temporal R-tree(MVSTR-tree)in the data flow environment by using the sliding window model is established;2)The index on the basis of the MVSTR-tree structure is built and can be updated and maintained dynamically;3)Based on the wellmaintained MVSTR-tree structure,a Top-k spatio-temporal keywords query algorithm for massive short texts is proposed,the Map Reduce computing model to parallelize query to find the most relevant Top-k short texts data objects which are closest to the users' query time,location and the content of the text information.(2)Aiming at the application requirements of social network environment,massive short texts spatio-temporal keywords query algorithm for social network applications is proposed.Applying the users' social relationship and the popularity of short texts to the spatio-temporal keywords query,the k data objects with high text relevance and the highest sum of,time and location relevance and social relevance to the users are returned.In this thesis,an index structure(Social IR-tree,SIR-tree)and pruning strategy suitable for social network application query is proposed,and combine MVSTR-tree index with SIR-tree to achieve the goal of efficient query.In this thesis,the proposed algorithm is compared experimentally on different data sets.The experimental results show that the proposed algorithm of Top-k spatiotemporal keywords query for massive short texts and the spatio-temporal keywords query algorithm for massive short texts oriented to social network applications have better applicability and high efficiency.