Probabilistic Binary Classification with Use of 2D Cellular Automata

Abstract

In this paper are presented wide known classification methods modified from almost deterministic into probabilistic forms. The rule for the classification problem designed by Fawcett, known as $n4\_V1\_nonstable$ is modified into two proposed forms partially ($n4\_V1\_nonstable\_PP$) and fully probabilistic ($n4\_V1\_nonstable\_FP$). The effectiveness of classifications of these three methods is analysed and compared. The classification methods are used as the rules in the two-dimensional three-state cellular automaton with the von Neumann and Moore neighbourhood. Preliminary experiments show that probabilistic modification of Fawcett’s method can give better results in the process of reconstruction (classification) than the original algorithm.