A numerical method of statistical pattern recognition

A numerical method of statistical pattern recognition is proposed, it needn't assume that the predictand meets the requirement of certain distribution. A statistical relationship between the predictand and the predictor was obtained directly by computer from its practical distribution, and then the type of pattern was recognized. Compared with the usual analytic discriminatory method, the Predictive effectiveness is improved.In this method, the influence of transparence is better. Therefore, it is better to be used in the very short-range forecast, in which the relationship between cause and effect is more obvious.

Project Supported:

Clc Number: