This Project is made of two research lines:
- First, a Genetic Programming engine has been built (GPPE), that projects datasets to spaces of higher or smaller dimension, where classification and regression is easier (cuasi-linear).
- In the second research line, we have used genetic techniques to evolve regression rules and Particle Swarm Optimisation (PSO). The rules evolved have the property that in addition to obtaining accurate rules, the subspace where each rule is appropriate, is also obtained. Also, the binary PSO for classification rules has been studied, by following an innovative approach where the solution is coded by the whole set of particles (Michigan approach) and not by each one of them (Pittsburg approach).