Average Distance Based Method for Association Rules Discovery in Relational Databases

The paper describes the new method for association rule discovery inrelational databases, which contain quantitative and categoricalattributes. Many methods have been developed yet. But most of thesemethods are based on initial equi-depth discretization of quantitativeattributes. These approaches bring the loss of information. The basicidea of the new method is to separate processing of categorical andquantitative attributes. The first step finds frequent k-itemsetscontaining only values of categorical attributes and then quantitativeattributes are processed one by one. Discretization of values duringquantitative attributes processing is based on a new measure calledaverage distance.