2006 Nov 01 - Wed

Fuzzy Logic

For an Automated Trading system I've been developing, I've come across the fact that Fuzzy Logic may assist in making decisions on how to trade at particular times of the day depending upon what conditions are predominant. Amazon has a bunch of theoretical books, but hardly any at all for the practical practitioner. I did purchase The Fuzzy Systems Handbook, 2nd Ed by Earl Cox. I'm about half way through it now. I've got through all the bits that make up the basic fuzzy sets. The sections are liberally sprinkled with C++ code. I'm not sure how much of it will compile in today's tools. The book was written back in the age of Windows 98. On the other hand, the code snippets are readable for one needs to understand what is happening in the commentary.

I had approached the subject from a different perspective though. I started by searching for code libraries. I came across FLUtE: Fuzzy Logic Ultimate Engine. The fellow has written a code library in C#. The code does compile in Microsoft Visual Studio with the v2.0 run time libraries. Coding new stuff in it may be somewhat of a challenge as the documentation is quite sparse. But then again, that is par for the course.

After taking a look at the modules, I came across something called 'hedging'. At the time, I didn't know what it was all about. And that prompted me to look for some good practioner's books. Hence the book I referenced above. Hedging, is obvious once you think about it. It is adding fuzziness to an existing fuzzy rule. The concept is well described in the book. The book doesn't exactly flow from front to back. For instance, during the beginning of the book, the author introduces a concept called alpha-cuts, and incorporates its use in to the development and discussion of fuzzy rules.. I can see what they do, but where and how they are applied, I'm still not exactly certain. And I'm up to page 344 now. There have been some hints, but no concrete usage criteria. I'm sure it will become clear as I move along in the book.

It was good that I did some prior reading, otherwise I think I would have been lost with the onslaught of information. I recall one of the first things I read was the document regarding the Mathworks Fuzzy Logic Toolbox. You can review the document in html or as a complete pdf document. In the pdf version, on page 56 (2-26), they have an excellent drawing summarizing how everything fits together.

Another book that helped fill in the gaps is An Introduction to Fuzzy Logic For Practical Applications by Kuzuo Tanaka. It is quite expensive for its 138 pages, but does have some useful background info. In the end, it is a good pocket guide for starters. It did leave me thinking that there were holes in the information presented. Earl Cox's book has filled in some of those gaps.

Having said all that, I'm just beginning to build the environment for using fuzzy logic in my trading solution. Although the FLUtE code looks like it could work, I'm going to try my hand at some basic C# code for fuzzy rules and rulesets to get a feel for what is needed. Once that is in place, I think I can then use a Genetic Programming engine I wrote to see if I can optimize some rule selections.