|
Kiran
2016-05-20 18:45:23
|
|
I ran AI->Prediction model with 5 inputs on SPY (following the various How-To tutorials and blogs) and Output is Boolean: perf(close,1)>0
1) How do i interpret if the prediction is accurate ? e.g. i get
- Directional Accuracy, Signal Accuracy, RMS error, Absolute error in Training and Validation periods. What's a threshold for good predictive power?
- Also, not sure how to interpret Output/Bar (a double value) - my Output function is Boolean, so this doesn
2) Is there a way to specify Optimization parameters in Prediction model Inputs, so it can learn which parameters are effective and drop the rest?
e.g. can i input something like ... Optimize("period", 4, 32, 4); RSI(per); instead of adding 8 RSI inputs?
- need a way to specify "bulk inputs"
3) How can i see the details of the model to determine which inputs are effective drivers of the results and which ones aren't?
I realize i can use Predict("name") function in a strategy, but it doesn't give insight into which factors are working
|
|