SI
SI
discoversearch

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: Len Giammetta who wrote (111)6/16/1998 9:49:00 PM
From: Jay Hartzok  Read Replies (1) | Respond to of 871
 
Len, No luck at all with it. In fact, I'm almost ready to trash it. After examining their website, I have come to the conclusion that you need "Profit" to make this program work satisfactorily.

Jay



To: Len Giammetta who wrote (111)6/17/1998 10:26:00 AM
From: Optim  Read Replies (1) | Respond to of 871
 
>Optim's pretty quiet, must be burning himself out on the new "Profit" program.

That's pretty accurate... :)

Actually, there is so many parameters and gadgets in the Profit & NGO combination that it is taking me a bit of time to get up to speed with this package. So far I have designed a few nets (S&P 500, Newbridge Networks, Abitibi Consolidated) that have tested really well, but they stop working after a short period. I just figured out how to force NGO to cut down the number of inputs automatically for me, which has increased the out of sample testing significantly. Now I just need to work on my data transformations (modifying O,H,L,C,V to work well with the nets) a bit so that they are more helpful to the nets.

The Profit documentation also gave me a few new ideas for NS Trader. After modifying what I was predicting in one of my NS Trader nets, the out-of-sample testing imporved over 100%. My Newbridge net now captured over 200% in the test period Jan1/97 to Dec31/98. This assumes you are willing to short, though.

I'll keep you posted...

Optim