Statistical Characterization and Optimization of Artificial Neural Networks in Time Series Forecasting: The One-Period Forecast Case