|  AUG 1991

A Hybrid System For Market Timing by Mark B. Fishman and Dean S. Barr

A Hybrid System For Market Timing by Mark B. Fishman and Dean S. Barr Hybrid technologies are perhaps the most-investigated topic in artificial intelligence today and here's how to apply the subject to trading. In two recent articles in STOCKS & COMMODITIES, we have demonstrated the application of advanced artificial intelligence technologies to the problem of market forecasting. That artificial intelligence works at all in predicting market movements compels us to rethink the absolutist position that markets are always random. What the evidence more plausibly suggests is that Standard & Poor's 500 index fluctuations are chaotic and their randomness is real—but it is intermittent. Between the periods of chaos, intervals of discernible non-randomness occur, and in principle, these are predictable, making market behavior recognizable often enough to be exploited (and making the periods of randomness identifiable, so that no adverse effects need accrue from decisions made on the basis of non-meaningful perturbations). What in the past has led us down the paths of doubt and disbelief (at least, those of us not analytically crippled by incurable optimism) has been a fault, not in ourselves, but in our software. ""Nonlinear dynamic systems"" (which is what markets are, mathematically speaking) simply do not happily succumb to analysis by classical statistical techniques. This is not to say that they can't be analyzed. The new arsenal of analytical weaponry that artificial intelligence technology has introduced may stand to break through that irksome harrier of complexity.

by Mark B. Fishman and Dean S. Barr