Inproceedings,
Using Particle Swarm Optimization and Genetic Programming to Evolve Classification Rules
The Sixth World Congress on Intelligent Control and Automation, WCICA 2006, 1, page 3415--3419. Dalian, IEEE, (2006)
DOI: doi:10.1109/WCICA.2006.1713002
Abstract
According to analysing particle swarm optimisation
(PSO), the structure of genetic programming (GP) and
classifier model, PSO algorithm and GP were made to
combine to evolve classification rules. Rules were
described as binary tree which non-leaf node denoted
rule structure and leaf-node was correspond to rule
value. Leaf node and non-leaf node employed different
evolutionary strategy. First, PSO was applied to evolve
leaf node in order to obtain the optimum rule of
certain structure, then GP was adopted to optimise rule
structure. The best rules were obtained after the twice
optimisation. Finally, the new method indicated
efficiency through experiments on several datasets of
UCI
