For illustration purposes take historical demands for an arbitrary product. If you know upcoming demands you can plan accordingly (for instance, stock amount and production capacity). The better the forecast the more accurate your planning.
From the background of features like seasonality and outliers regression analysis determines gradients of trend functions.Neural networks learn complex patterns from historical series. They approximate unknown functions instead of determining averaged trends.Are time series non-linear and functions not easily determined with traditional methods, neural networks provide an alternative. Deviation between forecast and actual value decreases. Quality of predictions rises.
