Machine learning
Ferra use machine learning techniques such as neural networks and support vector machines especially for short term projections where accuracy rather than theoretical model properties are importance.
These projections require large amounts of processing power and Ferra uses its own in-house compute cloud or, when its capacity is depleted, that of Amazon.
These models tend to be big, with a simple binary pattern input file easily taking up hundreds of megabytes of space and modelled using networks containing tens of thousands of weights.