A new nearest-neighbor rule in the pattern classification problem

摘要

A new nearest-neighbor (NN) rule is proposed. In this rule, the k-nearest neighbors of an input sample are obtained in each class. Two classification examples are presented to test the NN rule proposed. The number of samples misclassified Nm is evaluated. The minimum of Nm in the the NN rule proposed is found to be nearly equal to or less than those in the k-NN, distance-weighted k-NN and fuzzy k-NN rules. The NN rule proposed is shown to be very flexible. It will yield good classification results, if the parameters introduced in it are optimized.