Data Mining is one of the most significant tools for discovering as association rules. Yet, there are some drawbacks in conventional mining techniques. Since most of them perform the plain mining based on pre-defined schemata through the data warehouse as a whole, a rescan must be done whenever new attributes are added. Secondly, an association rule may be true on a certain granularity but fail on a smaller one. Last but not least, they are usually designed specifically to find either frequent or infrequent rules. This article aims at providing a novel data schema and an algorithm to solve aforementioned problems. A forest of concept taxonomies is used as the data structure for discovery of associations rules that consist of concepts picked up from various taxonomies. Then, the mining process is formulated as a combination of finding the large itemsets, generating, updating and output the association patterns. Crucial mechanisms in each step will be clarified. At last, this paper presents experimental results regarding efficiency, scalability etc. of the proposed approach to prove the advents of the approach.
展开▼
