QGEN Project
QGEN Project was a project for the task of finding new profitable trading strategies in a financial market. During my work on it, I developed the Competing Genetic Algorithm, which can be used in many other areas of machine-learning.
The main idea is to represent in the genetic code a sequence of mathematical operations of arbitrary length, which would compute predicted price movement of some asset. And then let genetic evolution do the job of evolving this code.
I have made some improvement to the classical genetic algorithm. As it fails to develop more than one “specie” at a time.
My idea is that in most cases there could be more than one good solution. Why not search for them simultaneously?
I trained a bunch of trading agents and even tested them on a real broker’s account.