[Swarm-Modelling] [Fwd: Re: [ABMs in finance]

[Steve Railsback]

-------- Original Message --------
Subject: Re: [Swarmfest2006] Keynote speakers
Date: Fri, 21 Apr 2006 05:09:18 -0600
From: Marcus G. Daniels <address@hidden>

Pietro Terna wrote:
>         I disagree totally: abm in finance is about behavior
Well, let me clarify.. On NYSE or LSE, for example, the transaction
history against the order books for stocks are recorded.  On these
transaction histories are the member and clearing (bank) codes of the
traders.  So the behavior of individuals can be observed directly, and
then some function to describe them inferred.   While the inference
process can be done using ABM, ABM models are expensive computationally,
and there is a lot of data to fit.   So of course it is about behavior,
but compared to many other areas of research, there is an unusually good
means for evaluating models (and in turn whether their complexity is
justified).   These are datasets with subsecond resolution that go on
for years across equities on multiple exchanges -- terabytes of data.
Dealing with all this is a big practical job, so given the choice
between making a complex agent model and effectively measuring the data,
I'd push for the latter.