Abstract
This paper investigates the convergency of the
probability of genetic regression in data mining based
on Gene Expression Programming (GEP) and the proposed
optimised algorithm based on GEP Minimised Residual Sum
of Square Genetic Algorithm (MRSSGA). By extensive
experiments on Genetic Programming (GP), GEP and MRSSGA
show: (1) that all algorithms could find the target
function from the data with low noise; (2) by comparing
the convergency speeds, new algorithms in GEP are 20
times faster than GP and MRSSGA and 60 times faster
than GP for simple data; (3) for very complex data with
an unknown function type, GEP and MRSSGA are
respectively 900 and 1800 times faster than GP at
finding ideal functions; and (4) aimed at the actual
data, the precision of models created by using genetic
regression methods is much more exact than traditional
methods.
Users
Please
log in to take part in the discussion (add own reviews or comments).