First task any system acting on the base of real world data meets is the task of pattern recognition. Before the system can produce an action corresponding to a request or an outer situation it should recognize type of the situation and activate required invariants. It means that the categories that compose the knowledge system should have the feature of self-identification.

We need a reliable classification algorithm to support pattern recognition in our developments. To cover this need we developed the algorithm of Evolutionary Classification and corresponding software. The algorithm has shown good convergence and ability to build classification models even when the number of dimensions is quite big.

You can get familiar with the algorithm on the Evolutionary Classification site:  We put there an online application that you can use to build a classifier either from data of you own or from the data artificially generated for you.

As with the Evolutionary Regression we strongly believe that Evolutionary Classification can and should greatly benefit from self-learning. Use of the knowledge collected and organized during the algorithm processing can save a lot of extraneous steps in future classification sessions.