Thank you for your questions.
To work with NN in Tradecision you need to read some information first. I would recommend you to watch the Tutorial “How to create a neural model and model-based strategy” at www.tradecision.com/tutorials.htm
Also it can be useful to read Tradecision help or User manual www.tradecision.com/download_manual.htm
Also I am emailing you a PDF of How to Create a Profitable Neural Strategy
Yes, a model specific only to the stock for which it was created. You open a symbol in Tradecision and only after that you can create a model.
It does make sense to use a model in a neural strategy. For example you create a model on GOOG. Then go to the Strategy Builder and create a neural strategy with this model as a condition of the strategy rules. Say entry long GOOG_Model_1@GOOG()> 4;
Model need some data to be trained on. This data is stock itself and model inputs. You cannot create a model for all stock at once.
On more advanced level you can use Committees to work with multiple models. In many cases using several models together in your neural strategy is more efficient than using one model along.
The simplest approach is to use a committee of several models. Each model can be responsible for a certain aspect of the market. You can create one model based on momentum indicators, another one based on volatility indicators and the third one based on fundamental factors. Then, you can enter long positions based on momentum and fundamental model and short positions
based on volatility and fundamental model. Or you can enter a position when two of three models forecasts the same.
For example, to enter a long position only when 2 of 3 models predicted a rise you need to write
he following formula in the Improvian Editor for long entry expression:
( Momentum_Model() > 0 and Volatility_Model() > 0 ) or ( Momentum_Model() > 0 and
Fundamental_Model() > 0 ) or ( Volatility_Model() > 0 and Fundamental_Model() > 0 )
Tradecision Support Team