Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Strategies & Market Trends : Neural Nets - A tool for the 90's -- Ignore unavailable to you. Want to Upgrade?

To: CatLady who wrote (46)	1/8/1998 3:27:00 PM
From: Optim	Respond to of 871

>Until I get a much better level of confidence in a neural net, I'm >considering it as the third level factor, after fundamental >and conventional technical analysis. This is probably the best way to apply an ANN. If you have a system that works currently, using a net may better the odds on a trade. No one trading technique should be used alone, unless you have build an extremely robust system and have tested it extensively on out of sample testing. Even then it is up to the trader to monitor all recommendations and use his or her instinct to take or deny the trade. No system eve work 100% all of the time. But a 60% accurate system can be very profitable! >That's the theory, have you personally seen it work out yet? I have on some issues. For example I am working on a TIPS (Toronto Index Participation Units - like an index fund) trading model in Neuroshell Trader that uses intermarket relationships to forecast and indicator similar to the percent change in price 3 days in the future. I found that by adding preprocessed data such as bond, tbill, oil and gold prices, the model has increased in accuracy. Again this is a bit of an art though, and I am a long way off from a truly profitable, robust system. A future release of NS Trader is supposed to have a Genetic Algorithm feature that will help optimize inputs and preprocessing. But until then it's sort of trial-and-error. Luckily I find this kind of stuff fun! >Once someone demonstrates repeatable success with a neural net, >won't the random walk theorists finally have to admit defeat? Are >we at that point yet? I haven't done enough reading on the subject. I don't think that will ever happen. The markets are always adjusting as new techniques and technologies are incorporated, which is why a model that works great today is useless tomorrow. So if an individual picks a random basket of stocks, and has luck on their side, they may just walk away ahead of the rest. There is no such thing as a highly profitable, guaranteed investment. You take your risks and hope for the best. >Are there any AI/Neural Net books that you'd recommend? There are a few, but I will have to post them later on, as I am at work right now. Optim