Our service allows to find out, if there are any significant attributes or attribute groups, even without finding decision rules for a training set beforehand. If the calculated criteria equals to 1, then a training set contains decision rule, which is guaranteed to divide the training set into classes, otherwise there are no such decision rules for the training set. We guarantee that. It’s up to our client to use or not to use this information.
The criteria is calculated for discretized training sets. The discretization process is controlled by a client by selecting a discretization type for each attribute on the stage of preprocessing of the data. The system supports such attribute types as float, interval, nominal.