The best out of sample normalized root mean squared error was 1.0096 which was achieved by the
neural network using the eight data series after the neural network structure had been reduced by
using the alternative objective function for the full training run