Machine learning in trading: theory, models, practice and algo-trading - page 517
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
If anyone wants to play around, the scaffold is trained on increments and gives a prediction for 1 bar forward. The depth of training, lag for increments and number of entries are set in the settings (each new entry is a shift of 1 bar back). Then the forecasted value is subtracted from the current prices. The histogram is drawn only for each new bar.
Please tell me at once what percentage of error it represents?
Tell me at once - what percentage of error?
I did not look at the errors, watched the histogram and changed periods. According to my senses - large, and with increasing the training sample decreases not much... but I later dopilu z-score for it and try to immediately through the bot to check. In fact, in this framework you can shove anything you want, not just increments.
If I set small lag for increments (here it's 1) it even seems to predict often. Each bar in the histogram is a forecast for the next bar. Below zero it is a Buy position, above it is a Sell position. And the value is how many points it is expected to grow/decrease
By the way, if anyone has any suggestions on how to add to the model also trend component + to the increments, I would be grateful, it should slightly improve the predictions on the strong long-term trends (in the flat it predicts well)
By the way, if anyone has any suggestions how to add to the model the trend component + to the increments, I would be grateful, it should slightly improve the predictions on the strong long trends (it predicts well in the flat)
Add a derivative of a suitable MA, and you'll be happy).
Or better the derivatives from a pair of MAs with different periods.
Add a derivative of a suitable MA, and you'll be happy).
Or better derivatives from a pair of MACs with different periods.
MACD or something? )
MACD of some kind? )
No, exactly two derivatives of different MACDs and their inputs to the NS. MACD is less informative.
I'll try later, yeah.
I'll try that later, yeah.
If you do not spare the inputs, you can add the delta between MAchas. The delta can be preliminary passed through the sigmoid and subtracted by 0.5 to bring it to zero.
No pity, I did up to 500 entries :) yes, indeed, tie the increments to something trendy, for example, 50 consecutive increments for 50 entries and 50 momentums or deltas of some kind, and repeat this 5000-10000 times. It will calculate in 10 minutes at most.
No pity, I did up to 500 entries :) yes, indeed, tie the increments to something trendy, for example, 50 consecutive increments for 50 entries and 50 momentums or deltas of some kind, and repeat this 5000-10000 times. It will calculate in 10 minutes at most.