Mining quantitative associations in large database

Association Rule Mining algorithms operate on a data matrix to derive association rule, discarding the quantities of the items, which contains valuable information. In order to make full use of the knowledge inherent in the quantities of the items, an extension named Ratio Rules [6] is proposed to capture the quantitative association. However, the approach, which is addressed in [6], is mainly based on Principle Component Analysis (PCA) and as a result, it cannot guarantee that the ratio coefficient is non-negative. This may lead to serious problems in the association rules' application. In this paper, a new method, called Principal Non-negative Sparse Coding (PNSC), is provided for learning the associations between itemsets in the form of Ratio Rules. Experiments on several datasets illustrate that the proposed method performs well for the purpose of discovering latent associations between itemsets in large datasets. © Springer-Verlag Berlin Heidelberg 2005.