Extending the Margin Algorithm

The design of a classifier usually has the important step of attribute selection. A computationally tractable scheme almost always relies on a subset of attributes that optimize a certain criterion is chosen. The result is usually a good sub-optimal solution. Previously we showed how an all attributes approach can be used efficiently by the classifier to classify a given data point in a binary dataset. The resulting classifier is transparent, and the approach compares favorably with previous approaches in both accuracy and efficiency. This paper extends that work to multi-class data sets and the order of attribute use in the classification process