|
PureBytes Links
Trading Reference Links
|
You can develop a real-time neural net for TS this way...
1. Create all the indicator signals and target value signal that you plan to feed your neural net in TradeStation.
2. Plot all of them on the same chart (multiple subgraphs OK)
3. Export the chart to a .CSV file.
4. Install Braincel, ( a perceptron neural net builder ) as an add-in to MS Excel.
5. Import into Excel the CSV file.
6. Train the neural neta and save it as a NET file.
7. Write a function in EL that can call the NET file and feed it the same indicators use sued to make the CSV file.
8. You now have a real-time DLL based neural net whose output can be used in the EL function just as any other value.
For example, I used Braincel to train a net to learn the 12x12 times table. I then used it in TradeStation to scale data whose range was within 0-12. I then compared the result to another plot that was simply the series multiplied by some constant, say 3. The two plots were almost identical. Of course, using a neural net to simply scale data is not what I would consider optimal usage, but the test validates the above sequence of steps.
For details on Braincel, see http://www.jurikres.com/catalog/ms_bcel.htm#top
>>A good indication is that institutionals with all their money have not embarked
on a widespread trading effort on such premises. If it were so good, you would
see lots more money invested in such ventures.<<
That assumes the user knew what he/she was doing with this technology. Nonlinear modeling is more difficult than linear modeling as it is very easy to think you have something when you really don't. Also, most users try to force the net into predicting future prices, a very difficult target. Finally, correct data preprocessing is crucial for successful modeling and newbies prefer to completely ignore this process because it is totally uninteresting. For these reasons, most users fail. However, those who study proper methodology, assign small tasks for the neural net, and persist in their experimentation are rewarded.
"Genius is 1% inspiration and 99% perspiration." - Albert Einstein
Regards,
Mark Jurik
http://www.jurikres.com
|