INTRODUCTION TO NEURAL NETWORKS
Chapter 5: Selecting A Concept That Works
Van K. Tharp: I looked all over for an expert to write about neural networks for this book. One of the problems with neural networks is that they are complex, they tend to border on curve fitting, and you can spend a lot of effort simply trying to predict whether one market will be higher or lower tomorrow, with about 55 percent accuracy. It’s quite frustrating, especially when my gut feeling was that more could be done with neural networks.
Finally, I happened upon Louis Mendelsohn through his web site and was quite impressed with what I saw there. Most of the articles he has written (over 50 of them) are presented in their entirety. Mendelsohn has gone way beyond predicting tomorrow’s price and actually uses neural networks in some very useful ways. As a result, I was delighted when he agreed to write this section of the concepts chapter. He’s an internationally acclaimed technical analyst, investment software developer, and financial author.
Louis Mendelsohn: Introduction to Neural Networks
The integration of intermarket analysis with traditional single-market technical analysis is necessary for profitable trading in the 1990s and beyond. Today’s limited single-market focus must yield to a broader analytic framework that addresses the nonlinear interdependence of today’s financial markets. In 1991 I first wrote about this framework, referring to it as “synergistic market analysis.” This approach allows traders to quantify complex intermarket relationships, assess the simultaneous impact of multiple related markets on a given market, and measure the leads and lags that exist within these relationships.
Neural networks are an excellent tool to implement synergistic analysis. They can be used to synthesize disparate data and find hidden patterns and complex relationships between markets. Neural networks are real, and they do work! In fact, they perform an outstanding job at processing extensive amounts of intermarket data. It is their ability to quantify subtle relationships and detect hidden patterns between numerous related markets that makes neural networks an important mathematical tool in the financial arena. How else could a trader examine the past 10 years of price data on 5, 10, or 15 related markets simultaneously to discern the effects that these markets have on a specific market?
Additionally, through the use of neural networks, financial forecasting becomes possible, so that traders can gain an anticipatory, not just a retrospective, vantagepoint on the financial markets. Anyone can tell you where a market has been in the past by simply looking at its price chart, but the real money is in correctly anticipating the future direction of that market! Through the use of neural networks applied to intermarket analysis, traders can actually forecast the financial markets, similar to the way meteorologists forecast the path that a hurricane is expected to take. Forecasting is never 100 percent accurate. It never will be. But from a decision-making standpoint under conditions of uncertainty, it’s a major step in the right direction.
To incorporate intermarket analysis into your trading plan, it isn’t necessary to change your trading style or stop using single-market indicators that work reasonably well. Intermarket analysis can be used to augment existing single-market approaches.
In order to appreciate the difference between single-market analysis and intermarket analysis, put one hand over one of your eyes. All of a sudden your peripheral vision is sharply restricted and your ability to grasp the entire environment is greatly reduced. That’s what single-market analysis is like in today’s financial environment. Now remove your hand and instantly your peripheral vision is restored. That’s what intermarket analysis is all about – broadening your perspective.