Summary: | 碩士 === 國立東華大學 === 資訊工程學系 === 96 === Mining of frequent patterns in databases has been studied for several years. However, real-world data tends to be dirty and frequent pattern mining which extracts patterns that are absolutely matched is not enough. An approach, called fault-tolerant frequent pattern (FT-pattern) mining, is more suitable for extracting interesting information from real-world data that may be polluted by noise. The proposed approach considers the problems of mining both proportional and fixed FT-patterns. The number of faults tolerable in a pattern is proportional to the length of the pattern in proportional FT-pattern mining, and is fixed in fixed FT-pattern mining. A new graph structure, FT-association graph, is presented to help filter out impossible candidates with high efficiency. This study proposes two algorithms. The first algorithm considers that the dataset can be stored in memory, and the second deals with the situation that the dataset cannot fit into memory. Both algorithms can be easily modified to adapt to fixed FT-pattern mining. Experimental results indicate that the proposed algorithms are highly efficient for mining both proportional and fixed FT-patterns.
|