We plan to research impacts on EURUSD currency with neural network.

1. Select series to use them in model
2. Develop a model. Data representation.
3. Select input and iutput vectors.
4. Build an training set
5. Build a theural network
6. Train and evolve network
7. Get a NN results
8. Analyze NN results

17.05.2009

So, today we start first tests. 4 euro crosses has been selected to make training set of logarythmic price changes of pairs. Input data contained 6 hours of every pair changes. First testing results shows cross validation set minimal mean squared error is equals 0.3.