Abstract: | This work looks into evolutionary techniques to explore numerical databases. The first part concerns an interactive genetic programming (GP) algorithm where, in cooperation with the user, a 2D graphical representation of the data gets evolved. The user supports the GP by selecting (aliging a fitness value to) interesting representations which are futher evolved.The second part is an extension of the int4eractive evolutionary data mining to non-interactive evolutionary data mining. the user as a fitness function is replaced by a computational fitness function. The performance of the two approaches is compared on several standard databases.
|