From: John Mason Subject: Re: New website: AI in finance and investment Date: Fri, 29 Dec 2000 13:51:44 GMT Newsgroups: alt.math.recreational,comp.ai.genetic,comp.ai.neural-nets,sci.math,comp.ai.genetic Looks like a tremendous resource. My interest is in using such tools not for portfolio optimization, but in optimal 'harvesting' of a portfolio while in retirement. Believe T.Rowe Price charges several hundreds dollars to accomplish this, on an ongoing basis, and claims to use Monte Carlo techniques. WOnder if anyone has published detailed information/software on this narrow aspect of finance optimization. Anyone? Thanks. -- John > Franco Busetti wrote: > > While developing a thesis on portfolio optimisation I spent a lot of > time gathering information on heuristics and artificial intelligence, > specifically focused on investment and finance. > > I've put them together as a website of resources for the fields of > general heuristics, neural networks, genetic algorithms, simulated > annealing and tabu search; with databases, articles, references, > journals, software, links, conferences and education for each field. > > See www.geocities.com/francorbusetti/. > > Hope it adds value, > Franco ============================================================================== From: jmb184@frontiernet.net (John Bailey) Subject: Re: New website: AI in finance and investment Date: Fri, 29 Dec 2000 21:46:49 GMT Newsgroups: alt.math.recreational,comp.ai.genetic,comp.ai.neural-nets,sci.math,comp.ai.genetic Summary: [missing] On Fri, 29 Dec 2000 13:51:44 GMT, John Mason wrote: >Looks like a tremendous resource. My interest is in using such tools not >for portfolio optimization, but in optimal 'harvesting' of a portfolio >while in retirement. Believe T.Rowe Price charges several hundreds >dollars to accomplish this, on an ongoing basis, and claims to use Monte >Carlo techniques. I am in a position similar to yours and have found a number of interesting leads on this subject. To my delight I finally found a use for the math I learned in graduate school. As an example: http://www.ggw.org/donorware/kelly/ addresses the optimum mix of growth vs safe securities. A sampling of useful articles in my reprint collection: http://xyz.lanl.gov/abs/cond-mat/0009437 http://xyz.lanl.gov/abs/cond-mat/9910212 Drop me a line without your anti-spam email address and we may find useful things to discuss. John ============================================================================== From: Lester Ingber Subject: Re: New website: AI in finance and investment Date: 30 Dec 2000 15:03:01 GMT Newsgroups: alt.math.recreational,comp.ai.genetic,comp.ai.neural-nets,sci.math,comp.ai.genetic The website is a great addition to the e-literature. The Kelly page is interesting. I think the point should be made that you can do better by taking into account different distributions for different components in the portfolio. For example, even neglecting correlations (which can be easily included in this scheme), consider 2 components x and y, dx = f_x dt + g_x dw_x dy = f_y dt + g_y dw_y where dw's are Wiener noise. The use of g's constants reduces to Gaussian distributions, and using g_x=x and g_y=y reduces to lognormal distributions. (Both distributions typically not very realistic as can be seen by fitting actual market data.) After fitting each market in the protfolio, e.g., fitting each g to each market assuming some polynomial form, then StdDevs (the g's) can be aggregated if one wishes to follow the approach given in the referenced Kelly page, or they can be used for other approaches, etc. Lester [quote of previous message deleted --djr] -- Prof. Lester Ingber http://www.ingber.com/ PO Box 06440 Sears Tower Chicago IL 60606-0440 http://www.alumni.caltech.edu/~ingber/