Aura® implements the fractal computational network based on the stochastic propagators. The individual knots of this network may be any forecasting algorithms.

The adjective fractal does not mean that uses Aura® fractal algorithms for forecasting purposes (although it is fairly possible). It points in current context that Aura® itself is fractal. In fact, fractal is defined as the selfsimilar structure produced by a set of invariant derivation rules.

Aura® network is exactly such structure. It can hold infinite number of modeling levels. Each modeling level can hold the arbitrary amount of forecasting models. Each model, in turn, obeys the derivation rules of the homomorphic hierarchy of interfaces and can hold itself infinite layers of such interfaces in itself. Aura® as such derives exactly from such interface, so it can hold itself in recursion to infinite order.

Individual forecasting models with the unified interfaces may be interpreted as the knots of the Aura® fractal network. To connect them into the optimal forecasting strategy Aura® uses the concept of stochastic propagators. Each propagator represents the series of stochastic points with the certain variance estimated for each point. The network uses a number of optimization techniques to ensure the serial convergence of propagators to the minimum level of variance. In this way the optimal forecast is achieved.

In contrast to the most other statistical packages, Aura® operates not on traditional time series. It operates on individual observation points. So it can combine in a single model the data with very different time steps, from fraction of seconds to many years. The user should be very carefully with input observation dates for each series. They must exactly coincide, or, else, algorithm will synchronize them on its own.