Type
Conference PaperKAUST Department
Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) DivisionComputer Science Program
Machine Intelligence & kNowledge Engineering Lab
Date
2014-04-16Online Publication Date
2014-04-16Print Publication Date
2014Permanent link to this record
http://hdl.handle.net/10754/564875
Metadata
Show full item recordAbstract
Web-based Data Imputation enables the completion of incomplete data sets by retrieving absent field values from the Web. In particular, complete fields can be used as keywords in imputation queries for absent fields. However, due to the ambiguity of these keywords and the data complexity on the Web, different queries may retrieve different answers to the same absent field value. To decide the most probable right answer to each absent filed value, existing method issues quite a few available imputation queries for each absent value, and then vote on deciding the most probable right answer. As a result, we have to issue a large number of imputation queries for filling all absent values in an incomplete data set, which brings a large overhead. In this paper, we work on reducing the cost of Web-based Data Imputation in two aspects: First, we propose a query execution scheme which can secure the most probable right answer to an absent field value by issuing as few imputation queries as possible. Second, we recognize and prune queries that probably will fail to return any answers a priori. Our extensive experimental evaluation shows that our proposed techniques substantially reduce the cost of Web-based Imputation without hurting its high imputation accuracy. © 2014 Springer International Publishing Switzerland.Citation
Li, Z., Shang, S., Xie, Q., & Zhang, X. (2014). Cost Reduction for Web-Based Data Imputation. Lecture Notes in Computer Science, 438–452. doi:10.1007/978-3-319-05813-9_29Publisher
Springer NatureConference/Event name
19th International Conference on Database Systems for Advanced Applications, DASFAA 2014ISBN
9783319058122ae974a485f413a2113503eed53cd6c53
10.1007/978-3-319-05813-9_29