Normalizing Network Outputs

This network normalization idea is from Murray Ruggiero, which he used in building a model in his latest articles (Futures Magazine, Jan, Feb, and March 2000). You normalize the output of a net by feeding it into a simple stochastic %k (Time Series category). Then find the optimum cutoff points in the Trading strategy wizard between 0 and 100. It seems to improve some models.

Both we and Murray have been using this idea with Adaptive Net Indicators, but it works with Turboprop 2 as well. Just make a net. Then make a rule like “buy when the simple stochastic of the net is > 50”. Now optimize that in the Trading Strategy Wizard.

Was this article helpful?

Related Articles