Extracting Informative Rules from High Dimensional Data Using a Numerical Approach

This paper presents a new approach whose aim is to extent the scope of numerical models by providing them with accurate knowledge extraction capabilities. These capabilities are especially useful for the management of high dimensional data. The basic model which is considered in this paper is a multi-topographic neural network model. The powerful features of this model are its generalization mechanism and its mechanism of communication between topographies. These two mechanisms allow rule extraction to be performed whenever a single viewpoint or multiple viewpoints on the same data are considered. The presented approach aims at extracting the most informative rules. This approach relies on an original algorithm that can be used for incrementally extracting the generators and the closed itemsets of the dataset with a prior access to the information provided by the numerical clustering model