A Novel Multivariate Discretization Method for Mining Association Rules
Hantian Wei · 2009
Data mining aims at discovering useful patterns in large datasets. In this paper, we present a novel multivariate discretization method for finding association patterns based on clustering and genetic algorithm. This method consists of two steps. Firstly we adopt a density-based clustering technique to identify the regions that possibly hide the interesting patterns from data space. Confined to the data in these regions, we then develop a genetic algorithm to simultaneously discretize multi-attributes according to entropy criterion. The effectiveness of the proposed method is demonstrated with the experiment on a real data set.