From Market Club and INO TV
New neural network users often follow a predictable pattern. They feed the network one year of data and a haphazard roster of technical studies including the RSI, MACD, ADX, three moving averages, Stochastics and any other handy data. These new users then expect the network to generate accurate predictions thirty days into the future. They are dismayed and perplexed when they discover that the forecasts are worthless.
Ed’s workshop covers the human brain and artificial neural networks. You will learn the proper steps you must take to generate meaningful results. He shows you how to select appropriate inputs and how to prepare or preprocess and otherwise manipulate your data prior to training your software. He teaches you how to extract the test data and how to train the neural net to the correct accuracy level. He describes network architectures and activation functions and shows you what to do if the network will not train to the desired accuracy level. Finally, Ed discusses some of the classic traps which await the beginning neural network user such as colinearity, price shocks caused by such incidents as the Persian Gulf War, and the onset of option trading in a specific item. He also describes the effects of other non-typical market events. Anyone who has contemplated the use of neural net software will benefit from this workshop.
PDF Workbook
Ed Gately received his degree in physics and started his professional career as an electronics engineer. He worked his way up the corporate ladder to become the president of a successful electronics company. He began actively trading the stock market upon his retirement from that firm. At that time, Ed began to research neural networks extensively as a complement to technical analysis. Ed developed a neural network for forecasting the S&P 500 ten days into the future that, when tested with fresh data, generated amazingly accurate results. When forecasting ten days out, 38 percent of the predicted points were within one point of the actual S&P figure, 76 percent were within two points, and 93 percent were within four points. When forecasting five days into the future, the results were even better: 56 percent of the predicted points fell within one point of the actual S&P, 77 percent were within two points, and 99 percent within four points. Ed’s book, Neural Networks for Financial Forecasting, builds upon this research. Ed spends most of his time developing, writing about, and lecturing on neural networks. In addition to these efforts, he is currently workingSaturday Seminars are just a taste of the power of INO TV. The web’s only online video and audio library for trading education. So watch four videos in our free version of INO TV click here.
