Discovery of Field Functional Dependencies

Integrity constrains in relational database were proposed for logical database designation. Recently, the data quality problem is getting more and more attentions, and integrity constrains are used for detecting and repairing inconsistent data. In the purpose of detecting inconsistent data more comprehensively, previous research has proposed more type of integrity constrains, including conditional functional dependencies, editing rules and fixing rules, etc. In this paper, a new type of constrain was proposed: field functional dependencies (FFDs). In case that a database is logically well designed, it becomes difficult to detect and repair inconsistent data according to normal functional dependencies, while FFDs can be a complementary to this case. To make well use of FFDs, this paper has concentrated on the problem of discovering FFDs, i.e., given a sample database instance, how to find out such kind of constrain rules between attributes. Usually the input data could be so large that it becomes expensive and inefficient for manually discovering of FFDs. According to the properties of FFDs, an efficient algorithm was designed to discover these rules. Finally, the efficiency of the discovering algorithm, and the coverage of FFDs when detecting data errors compared with previous works were experimentally evaluated.