Algorithms On Answering Why-not Questions On Convoy Queries

In recent years,why-not questions,aiming to seek clarification on the missing tuples for query results,have captured wide academic attention among database communities.Answering why-not questions can help improve the usability of trajectory database remarkably if reasonable explanations about the why-not questions are made,and this may bring a series of important practical applications.Besides,the study of trajectory data mining of moving objects,more recently,is also becoming hot issues and tough topics,and convoy query tends to be an essential application in the filed of trajectory data mining.It will refine convoy queries and improve availability of database tremendously that if the database could give a reasonable explaination when the expected convoy is missing from the convoy query result.However,none of related works are conducted to solve the problem above.Motivated by this,we develop efficient algorithms to answer why-not questions on convoy query in this paper.In this thesis,the research works of trajectory data mining technology,convoy query technology and why-not questions are summarized firstly.Next,the concept of why-not questions on convoy query was proposed and defined.Then,we study and analyze the effect of clustering order,density threshold MinPts and distance threshold Eps on clustering results of Density-Based Spatial Clustering of Applications with Noise(DBSCAN),and present a framework WQCQ(Why-not Questions on Convoy Queries)based on query refinement strategy to answer why-not questions on convoy query.Convoy queries need three parameters:an integer lifetime k,a density threshold MinPts and a distance threshold Eps,and our framework offers three solutions to answer why-not questions on convoy queries with minimum penalty.In the first approach,we will modify the query parameter-lifetime k.And the query parameter density threshold MinPts and distance threshold Eps will also be refined respectively in the second and third methods.The rest of this paper reports experimental results and effectiveness and efficiency of our proposed algorithms by conducting extensive experiments with two real datasets and one synthetic dataset.And our experimental results show that all the algorithms can answer why-not questions effectively and have good extendibility.Finally,we conclude the thesis with some directions for future work.