Weighting variables for binary clustering

This paper deals with binary clustering and the authors are particularly concerned with the ponderation of the variables. They place the binary clustering problem in the framework defined by Govaert. He defined a specific algorithm for this type of data, called MNDBIN, that the authors briefly describe in section 2. Next, they show that choosing a ponderation vector for the variables can be judicious to classify the data. Then the authors present and analyze some adaptable ponderation methods for clustering. They end by a comparative study of this approach with a probabilistic model using two samples of simulated and real data.