Yes, but I’m currently testing few ideas based on the following message which I have found on the Eureqa group:
“Here are a couple ideas I had. I think you can do what you want, but it
requires some initial setup outside of Eureqa:
(1) Do binary classification:
Perhaps you can preprocess on the data, calculating when it would have been
a good idea to buy or sell the stock. In other words, create a new variable
of zeros and ones, with ones for rows where it would be optimal to have this
stock, and zero otherwise. Then try to model this value as a binary value,
like "buy = logistic( f(x,y) )". See:
http://blog.eureqa.com/2010/06/modeling-binary-yesno-output.html
The f( ) that Eureqa looks for then becomes a strength or probability toward
"buy" being one, which implements a strategy based on your preprocessing.
You could also do the same for detecting other qualitative events, as along
as you can setup a variable of 0's and 1's for it.
(2) Find a scoring function:
Another idea is to search for a scoring function, where the search relation
takes inputs for two stocks (or the same stock at different times) and
returns which one is "better". For example, a search relation like
"buy_x = logistic( f(x1,x2,x3) - f(y1,y2,y3) )"
Here, each row in the data contains variables for two stocks, x:{x1,x2,x3}
and y:{y1,y2,y3}, plus a variable "buy_x" that you've created that equals 1
when the stock x would be preferred over y, and 0 otherwise. Then the
"f(...)" that Eureqa searches for would effectively be a scoring function
that has high value when the input stock is good. You could then use a
solution f(...) to implement a trading strategy by ranking stocks based on
their f(...) value.
This method also requires preprocessing to create a "buy_x" variable and
setup of the data into pairs of stocks. Again, you could replace "buy_x"
with any other trait, such as stock x showing some trait more or stronger
than stock y.
“