on the complexity of this network is similar to the prediction of color next bar.
results are not the best but the expectation is positive. I think we must add some additional conditions in trade rules and stop level
10 years on the EURUSD 1440. three-layer perceptron nonlinear configuration 100-5-1. 100 bars are used during the training. Test sample no. 500 epochs.
output AC (1,2,2) for 1 value ahead. Simple trading rule: OUT[0] > 0 = BUY POSITION and OUT[0] < 0 SELL POSITION.
Above I pointed out the equity in AUDUSD H4. And now EURUSD H4 equity Out-of-Sample period. Slightly worse but the good movement took (including news of withdrawal of Greece from the European Union).
Based on the network, the image that I posted earlier, I wrote a simple EA. The results of EA for the last month quite stable. Network predicts the indicator values of AO at 1-3 ahead. Post the schedule for the period of the equity Out-of-sample.
Forecast indicator AO AUDUSD H4 (05/10/2011). 5 different from the input array of neural networks. Each forecast is targeted as the average of 5 successive generations.