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
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Larry Livingston who wrote (824)9/7/2000 5:11:56 PM
From: mark_andrews  Read Replies (2) of 871
 
Hi Larry,

I have been doing extensive research on this and I have discovered a few hidden secrets which no one will mention in their sales material. Those who do mention model selection (for one) in their sales material do not have any evidence to substanciate their claims. I have not seen anything about model selection error rates. How do you know a model will continue to perform by using a given method of evaluation?

Even when you have a prefect net which does not make any mistakes in-sample or out of sample, it does not mean it will trade well in the future. Even if you have to right inputs, nets are not good for more than 6 months to a year because they tend to overfit the input data. In order to find models that will perform well moving forward, you need years of data the model has not seen. Even then, there are many other tests that need to be done before you can have the confidence the model will continue to perform in the future.

I have developed well over 2000 models in the past 2 1/2 years using automation running 24/7. I have use nets and standard optimization techniques. No matter how well a system performs using historical data, the model will probably not perform well unless the variable settings remain the same over a long period of time. I have found that nets tend to cause more noise than long term trading accuracy. It is truly mind boggling when you see what it takes to make a system trade well and then you are still plegued with the decision of which model to trade. Being able to select a model that will consistantly give you 100-200% AROR is difficult to do. Many get discouraged and invest in mutual funds.

Regards,

Mark
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  

HomeMailHotSubjectMarksPeopleMarksKeepersSettings
Terms Of UseContact UsCopyright/IP PolicyPrivacy PolicyAbout UsFAQAdvertise on SI
© 2025 Knight Sac Media.  Data provided by Twelve Data, Alpha Vantage, and CityFALCON News