Abstract
Here a new statistical pattern classifying system is proposed. This system essentially solves the problem of the "peaking phenomenon". That is, the phenomenon whereby the accuracy of a pattern classifier peaks as the features increase under a fixed size of training samples. Instead of estimating the distribution of class-objects, the system generates a region on the feature space, in which a certain rate of class-objects are included. The pattern-classifier identifies the class if the object belongs to only one class of coverage region, but answers "unable to detect" if the object belongs to the coverage region of more than one class or belongs to none. Here, the coverage region is simply produced from coverage regions of each feature, and then it is extended if necessary. Unlike the Naive-Bayes classifier, the independence of each feature is not assumed in this classifier. Testing of the system on classification of characters shows that the performance does not significantly decrease as the features increase unless apparently useless features are added.
