Forums  > Trading  > Agent Based Models  
     
Page 1 of 1
Display using:  

leone


Total Posts: 7
Joined: Feb 2007
 
Posted: 2007-02-27 09:24
Hello,

I would like to add to my automated trading systems a framework of Agent Based Models.

I have already experience in forecasting and trading stocks and futures, and did in the past some work with pattern definition/recognition and more classical statistical methods (multi factor models, spline based regressions etc).

Does anybody have experience with Agent Based Models within the trading framework? Are there some good links, some good starting literature?

I appreciate any help!!

leone

urnash


Total Posts: 561
Joined: Sep 2006
 
Posted: 2007-02-27 10:56

Hi leone,

The search function is rather hidden on NP, have you used it? I came up with the two following links, but I'm not sure if that's what you want (my idea to get rich is only using one economic agent: me). See here and here. But I'm sure that other NP-ers know much more than me...

And welcome to NP! Party


Life can only be understood backwards; but it must be lived forwards - Kierkegaard

leone


Total Posts: 7
Joined: Feb 2007
 
Posted: 2007-02-27 16:53
Hi urnash,
thanks for the hint and the links! I am now looking through the threads and google through some of the therein listed links.
leone

leone


Total Posts: 7
Joined: Feb 2007
 
Posted: 2007-03-06 12:04
Does anybody have experience with special software for agent based models?
Are there standaloan programs or libraries/packages?
I found something on the web but I was not fully satisfied ...
Maybe some hint?
leone

ZeroBeta


Total Posts: 84
Joined: Oct 2006
 
Posted: 2007-03-06 19:38
from a purely simulation point of view, it might be interesting to look at this

"The Santa Fe Artificial Stock Market consists of a central computational market and a number of artificially-intelligent agents. The agents choose between investing in a stock and leaving their money in the bank, which pays a fixed interest rate. The stock pays a stochastic dividend and has a price which fluctuates according to agent demand. The agents make their investment decisions by attempting to forecast the future return on the stock, using genetic algorithms to generate, test, and evolve predictive rules."

there is a swarm implementation of it here


doctorwes


Total Posts: 576
Joined: May 2005
 
Posted: 2007-03-06 20:02
For those of you interested in macroeconomics rather than finance, the "Post Walrasian" school has been exploring the use of models in which agents don't aggregate nicely because they don't satisfy enough of the (unrealistic) neoclassical rationality assumptions. Since there are no representative agents, standard techniques don't apply, and you need to do a lot of computationally intensive agent-based modeling. There's a nice overview here. The research program is still in its infancy, but seems quite promising.



leone


Total Posts: 7
Joined: Feb 2007
 
Posted: 2007-03-07 15:38
thanks to zerobeta and doctorwes for the useful links. i will read now some materials and come back after ...

leone


Total Posts: 7
Joined: Feb 2007
 
Posted: 2007-03-08 11:51
What software would be best to start for Multi Agent Models? I read now some part of the SWARM tutorial, but I am not sure if it is worth. Because on the other side it may be possible to produce some own code in C++ and/or R/Splus.
Effectively I just have code in R and C++ to handle stock and futures data, including all the things around of that, i.e. indicators, models, tested (and seldom satisfactory) systems

@zerobeta: the artificial stock market seems nice. but for an own trading system it does not seem to help a lot. my agents must learn on real data, not on simulated... but the idea of constructing/watching at artificial markets is nice for sure

Tradenator


Total Posts: 1587
Joined: Sep 2006
 
Posted: 2007-03-08 12:00
Try Mathematica?  (just a guess)

leone


Total Posts: 7
Joined: Feb 2007
 
Posted: 2007-03-08 12:06
@Tradenator: your link describes the connection between Mathematika and REPAST (http://repast.sourceforge.net/). In a similar way R/Splus or other languages could be connected to REPAST, I assume. Do you have some experience with agent based trading models and REPAST?

leone

Tradenator


Total Posts: 1587
Joined: Sep 2006
 
Posted: 2007-03-08 12:18

Nope, I just recall people crapping on about interacting agents and Mathematica a few years ago.  Sorry I can't give you better answers.

Edit: I also see that Python and REPAST can talk via Repast Py.  I have had good success with Python and R, so I would think that this might be a good start for you.  (I am not an uber-hacker so if I can get python to talk to R it must be easy...)  Maybe try the tutorial in the link.

Previous Thread :: Next Thread 
Page 1 of 1