Soft Set Approach for Mining Quantitative Association Patterns in Databases

Abstract

Association rule mining is an active data mining research area. In recent years, association rules from large databases have received considerable attention and have been applied to various areas such as marketing, retail and finance, etc. The traditional algorithms for mining frequent association patterns suffer from the problems of under prediction and over prediction of patterns. The main aim of the present paper is to develop a soft set approach for mining fuzzy quantitative association patterns in order to address the issues of under prediction and over prediction of these patterns. The proposed approach is illustrated with the help of a suitable example and experiment on a real world data set of air pollution. The transactional dataset is represented as soft set(fuzzy) using the concept of parameter co-occurrences in the transaction. The quantitative attributes are dealt with by fine -partitioning the value of each attributes and then creating new tables which represent each (fine) partition as a field. The results obtained by soft set approach, soft set fuzzy approach are compared with those obtained by Apriori algorithm without soft set approach. The significant differences have been observed in support and confidence levels of patterns obtained by three approaches.