Discovering Condition-Combined Functional Dependency Rules

Yuefeng Du,Derong Shen,Tiezheng Nie,Yue Kou,Ge Yu
DOI: https://doi.org/10.1007/978-3-319-11116-2_22
2014-01-01
Abstract:Conditional functional dependency (CFD) on a relation schema is an important technique of data consistency analysis. However, huge number of CFD rules will lead to lower the efficiency of data cleaning. Thus, we hope to reduce the number of rules by raising support degree of CFD. As a result, some crucial rules may be discarded and the accuracy of data cleaning will be decreasesd. Hence, in this paper, we present a new type of rules which combines the condition values. Using the rules, we can reduce the number of CFD rules and maintain the accuracy of data cleaning. We also propose 1) a 2-process search strategy to discover the combined condition rules, 2) the method of combining the CFD rules by combining the inconflict values and 3) pruning method to improve efficiency of the search. Finally, Our experiments show the efficiency and effectiveness of our solution.
What problem does this paper attempt to address?