Mining Fuzzy Association Rules in a Database Containing Relational and Transactional Data
Keith C. C. Chan, Wai-Ho Au · Studies in fuzziness and soft computing · 2001
Many effective algorithms have been developed to mine association rules in relational and transactional data separately. In this paper, we present a technique for the mining of such rules in databases containing both types of data. This technique, which we call Fuzzy Miner, performs its tasks by the use of fuzzy logic, a set of transformation functions, and by residual analysis. With the transformation functions, new attributes and new item types can be derived for either relational or transactional data. They also make it possible for association rules relating the two types of data to be discovered, e.g., the buying patterns related to the demographics of a group of customers. With fuzzy logic, Fuzzy Miner is not only able to discover Boolean and quantitative but also fuzzy association rules. This makes the patterns discovered more easily understandable by human users and more resilient to noise and missing data values. With residual analysis, Fuzzy Minder does not require any user-supplied thresholds that are often hard to determine. The Fuzzy Miner also discovers relationship between fuzzy and quantitative values and allows quantitative values to be inferred by the rules. With these features, Fuzzy Miner can be applied to real-life databases containing relational and transactional data.