|  MAR 1991

Artificial Intelligence And Market Analysis by Mark B. Fishman, Dean S. Barr and Walter J. Loick

Artificial Intelligence And Market Analysis by Mark B. Fishman, Dean S. Barr and Walter J. Loick Markets are chaotic and their randomness is real but intermittent; we can approach the development of trading systems with that assumption. We can also assume that predictable intervals of non-randomness occur between periods of chaos, making market behavior recognizable enough to be exploited. Two forms of technology have been notably successful in the development of dynamic trading systems: expert systems and, more recently, neural networks. EXPERT SYSTEMS An expert system is simply a set of rules that express some things we can reliably believe about the world: EXPERT SYSTEM 1 Rule 1: IF the window is wet THEN it is (probably) raining (90%) Rule 2: IF it is raining AND I left my oar windows open THEN I should run outside and close them (99%) Rule 3: IF it was sunny this morning THEN I (probably) left my car windows open (95%) These are noncontroversial rules about which most of us are fairly ""expert."" Note that these are all, in some sense, rules of thumb, or what artificial intelligence (AI) people call ""heuristics"": none of them represents an absolute certainty. If the window is wet, my sprinkler may be on, or perhaps a firetruck has just arrived to put out the flames in my house. If I left my car parked in any major city, then rule 3 is probably crazy. Finally, even rule 2 is open to debate. If I am currently in bed with pneumonia or a broken leg, perhaps I should consider some other way of dealing with my car's open windows. Still, these rules have a certain reliability, and they correspond to the rules in our own lives.

by Mark B. Fishman, Dean S. Barr and Walter J. Loick