Again, consider the monthly log returns of GE stock from January 1926 to December 1999. Reserve the returns in 1998 and 1999 for forecasting evaluation.
- Fit a 3-2-1 feed-forward neural network to the return series and calculate the mean squared error of the 1-step ahead forecasts in the forecasting subsample. Write down the biases and weights of the network in the estimation subsample.
- Suppose that we are interested in forecasting the direction of the 1-month ahead stock movement. Fit a 6-5-1 feed-forward neural network to the return series using a Heaviside function for the output node. Compute the 1-step ahead forecasts in the forecasting subsample and compare them with the actual movements.