Abstract:

This paper presents gene expression programming for attribution reduction in rough set (GEP-ARRS), which designs a new gep code to convert attribution reduction into expression tree and a new fitness function. Meanwhile, to solve optimal reduction quickly, GEP-ARRS implements dynamic population creation strategy to reduce gene length of GEP constantly in order to accelerate solution efficiency of GEP. By extensive experiments, it is showed that for mass or high dimension data sets, in the matter of speed and quality which solves attribution reduction, GEP-ARRS has more apparent advantage by contrast with traditional attribution reduction algorithm on intelligence computing.

Key words:

intelligence computing