What to Predict and Which Inputs to Use Video

If you’re building a neural network model, it’s a better idea to select a volatile issue rather than an issue that is simply trending. That is because neural nets are most effective learning patterns in volatile issues where there are more varied patterns in the relationships between inputs and the predicted output value.

The next question is where to start when selecting inputs. The Trader comes with over 800 indicators and it’s a daunting task to sift through that many indicators. Based on our experience in building models, we generally suggest users start with indicators from the Price Momentum and Volume categories and also look at the Linear Time Regression Slope indicator in the Regression category. We recommend using only a handful of indicators to prevent overfitting in the model.

Refer to the help topic Troubleshooting Your Model – What to Do if You Feel You Haven’t Been Successful for more details on creating successful models.

Was this article helpful?

Related Articles