Clustering Analysis And Rules Educing In ARMRDB

In this paper, a novel data-mining model ARMRDB (Association Rules Mining in Relational DataBase) based on fussy set and rough set is presented for mining association rules in relational database. The difference from other model is that ARMRDB doesn't rely on special domain data, it's a all-purpose data mining model. The theory, structure and realizing technique of the model are introduced in this paper, and the process of clustering-analysis and rule-educing is described in detail. To improve the efficiency and robusting of the algorithm, we utilize the incremental method and take the space-time consumption into consider. Clustering analysis is an important phase, we make use of fuzzy analogue method to realize clustering. Through the clustering, the noise data can be eliminated to a certain extent, which can improve the correctness of result-rules. Another phase is rule-educing, owing to the attributes reduction, the redundant attributes can be omitted, and the association rules can be obtained in the key attributes set. Sometimes the format of rules can be adjusted according to the professional demand, and besides, it is necessary to evaluate the result-rules by objective measurement, at the same time, correlation analysis can remove the false rules. As to the research of the subjective measurement of rules and fuzzy evaluation algorithm, it's our next working target.