Knowledge constrained evolutionary algorithms: a case study for financial investing

Abstract

The purpose of this paper is to examine the role of domain knowledge in guiding evolution. The hypothesis examined in this paper is that the use of knowledge represented as a semantic network will bias mutation so that changes in structure measured by the semantic net correspond to changes in function. This hypothesis is tested utilising information and data from the finance domain. In this paper relevant literature is reviewed and an experimental framework is proposed which incorporates knowledge in evolution. An empirical investigation is presented to demonstrate the role of knowledge and gradualness in evolution. Future work will involve investigating methods to identify or construct a semantic network which is 'meaningful' to humans as well as machines.