Customer Reviews

Neural Networks for Financial Forecasting (Wiley Trader's Exchange)

5 star:		(1)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(3)

 

 

About the only positive thing I can say is that he has done a good job at capturing the feel of the trial and error process that is so necessary in developing a neural network model. That said, I could write a book about what's wrong with the book. There is so much poor advice and misinformation in the book, I would scarcely know where to begin. As I read it, I literally cringed at the so-called "information" being passed on to the unaware.

His errors includes poor sampling of test data, no validation data, poor data preparation, highly correlated inputs, small sample size, basing design changes in on insignificant changes in error, etc.

The extensive bibliography may be useful to some, but I would bet my net worth that it is not a list of books he has read. If you have a copy of Azoff's book on forecasting, you will find that Gately essentially cribbed the bulk of his bibliography from Azoff. The few remaining entries were cribbed from the bibliography found in the Neuroshell literature. The book is fairly heavy with hype for the Neuroshell product. (Not that it's a bad product.)

If you're thinking of buying this, I would recommend instead Master's Practical Neural Networks in C++. Doesn't matter if you don't program, it has plenty of good text, and most importantly, sound advice.

The reviewer that wrote the poor and misleading review described the book perfectly.

In a 4 page passage of the book I found at least 10 errors involving contradictions between text, tables and figures. This becomes very frustrating if you are trying to reproduce the results Gately achieved--a task I have found impossible.

Gately also claims great success however he does not test his neural nets on a validation set of data. This is a fatal error in real trading. It is therefor impossible to determine if he was successful at all.

It is clear that Gately does not have a background in quantitative sciences, as he has little skill in accurately and completely reporting procedures and results.

The idea of the book and the trial and error approach is good but another author should write the second edition.

This book offers a lot and delivers very little. For one thing, it gives no theory of NN, instead it wastes one full chapter on the brain and NN's. The front page offers to show "Top techniques for designing and applying the latest.....", but this is false advertizing!. In short, VERY DISAPPOINTING!.

What does this book promise?

"Succinctly explains how neural networks function, what they can accomplish as well as how to use, construct and apply them for maximum profit."

Ok, so perhaps it should say .... An introductory text with basic definitions and a example of how to use a neural network to predict the S & P 500. But even on the grander promise this book delivers. The book is too basic for those previously familiar with NNs, but perfect for someone who merely would like to begin and fears the prospect of becoming too frustrated to finish. Gately points out potential snares and discusses how to overcome them and walks the reader through a project. He describes how to preprocess the data to ensure meaningful results.

I have followed his steps and have been very please with my project, this book and the value I received when purchasing it.

Great for intermediate level investors who want to explore neural networks as they apply to financial markets. Author sticks to the overview process. One case study is used to illustrate procedures and concepts. Work is reasonably paced. I recommend it as a good starting point for neural network analysis. Requires some background in financial markets and Excel. Could benefit from including a CD of examples and concept elaboration.