New article Self-optimization of EA: Evolutionary and genetic algorithms has been published:
This article covers the main principles set fourth in evolutionary algorithms, their variety and features. We will conduct an experiment with a simple Expert Advisor used as an example to show how our trading system benefits from optimization. We will consider software programs that implement genetic, evolutionary and other types of optimization, and provide examples of application when optimizing a predictor set and parameters of the trading system.
It is important to define the following for solving the optimization task:
parameters that will be optimized;
optimization criterion — scalar that needs to be maximized/minimized. There can be more than one criterion;
target (objective, fitness) function that will calculate the value of optimization criterion.
A fitness function in our case will consistently implement the following:
forming the initial data frame;
dividing it into train/test;
training the model;
testing the model;
calculating optimization criterion.
Author: Vladimir Perervenko