Research On Usability Analysis Of Typical Preference Query Results

In many database application domains,the volume of data has hit a record high and continues to grow exponentially.In decision support applications,one of the challenging issues is to retrieve the most important(top-k)answers which end-users are interested in from massive data given user-specified preference queries.If users are unsatisfied with the returned results,Why questions may be raised to filter out unwanted objects from the result list.On the contrary,the corresponding Why-not questions may be asked to get expected objects not returned by the system.This thesis focuses on the usability of query results,especially in two typical types of preference queries,i.e.,Top-k query and Skyline query.In Top-k queries,it is not easy to specify preferences explicitly.In this context,Why questions are often asked when the unsatisfying query results are retrieved.To deal with the Why questions,a sample-based BTW approach is proposed.By adjusting either the weights of preferences,the k value or both these parameters,this approach can get a new query at a lower cost and thus exclude Why tuples results.However,the BTW approach requires to perform Top-k query processing on each object in the dataset.It results in high time complexity as the number of data increases.To improve the efficiency of the BTW approach,an improved BTW approach is proposed by leveraging two optimization techniques that integrate appropriate pruning strategies.To verify the effectiveness of the solution,a variety of cases are analyzed.Experimental studies using real and synthetic datasets are also conducted to examine the effectiveness and efficiency of the algorithm.Regarding the Why-not questions of Top-k query in a specific area,one important problem is that some expected query results are not retrieved.In this context,users may want to know the reasons so as to take corresponding measures to obtain the Why-not tuples.To help users to get the desired results,a general processing scheme ATWN algorithm is proposed.Its basic idea is to start from a general case of a Why-not element in a twodimensional region and then gradually extend to multiple Why-not elements in a multidimensional region.The algorithm is locally optimized to improve its performance.The effectiveness and efficiency of the proposed algorithm are tested through several experiments.The skyline operator is important for business applications involving multi-criteria decision making in practice.By leveraging existing processing schemes of Skyline query on the orthogonal region,four measures are proposed to help include Why-not points in the results based on the specific location of the Why-not object.They include the MWP algorithm to modify the Why-not point,the MRN algorithm to narrow the orthogonal region,the MWR algorithm to modify both together,and the MRE algorithm to expand the orthogonal region.The experimental evaluation on real and synthetic datasets shows that the algorithm can provide higher-quality solutions to the Why-not questions in Skyline queries on orthogonal regions.This research can provide feasible solutions for solving Why or Why-not questions in other related preference queries.It has academic and practical value for advancing the theory and practicability of database usability.