基于appgram的字符串近似搜索与连接的研究与改进-软件工程专业论文.docxVIP

AbstractAbstract Abstract Abstract When we do the querying in a traditional database，the selection opermion and the join operation may be regarded as one of the most important operations．But in real application，since the potential error or the different data representation by different format，if we use the exact selection or join operations，the querying may not return answer due to the mismatch in data field．In order to solve the problem of the above situation，by adopting a fault tolerance mechanism，the approximate selection and join are introduced into the query processing．HoweveL the string similarity search and join not only can be used for the database，but also in many other areas such as the DNA sequence analysis，time series processing，duplicated Web page detection，spell checking，data cleaning，data integration，search suggestions in search engine and SO on． According to the large time expense of the previous algorithms for string similarity search and join，in this thesis，we will focus on improving the algorithm performance and work on the framework of the Appgram．After briefly describing the problem definition of the string similarity search and join，related concepts around these problems，the widely-adopted algorithm framework and the performance measurement，the thesis has an in—depth study on the performance bottleneck of the algorithm，proposes some feasible solutions and conducts several experiments． (1)By making use of the string structure information，we design a new inveaed index and improve the strings’count filtering，SO we avoid counting the same 矗equency for several times in some cases，reduce the candidate size in the verify stage and improve the algorithm performance effectively with less time expense． (2)We extend the original filter to a multi-level approximate filter in the second filtering stage with the full use of the original boundary conditions，thus，we develop a more strict selection constraint to the candidate set，reduce the size of the candidat