Depending on how well the hypothesis (NMPC-Graph) fits to the historic data the use of CYNERELO™ software results in non-stochastic, falsifiable models of particular precision. If the model is able to produce historic output data well from historic input data the hypothesis proves util.
Since our approach allows to incorporate expert knowledge via NMPC-Graph into AI models, it stays ahead in comparison to classic AI approaches whose modelling is limited by either deriving almost only from pure data without human scenario knowledge or by over-simplified quasi-linearity.
Micro- and macroeconomic systems often are determined by convoluted feedbacks and dependencies of unknown delays. For classic AI regression algorithms which hardly can be guided by a lifelike scenario description there spans an immense problem search space where today’s computing power quickly becomes overstrained with increasing problem complexity.
Our approach is innovative: By supplying a hypothesis about scenario’s convoluted feedbacks and dependencies of unknown delays in a simple and compact way there is a helpful guidance for AI. Thus, search space is diminished by magnitudes and model precision can be reached with less computing power and less data.
If this way the analyst achieves to create a precise model, the benefit is not only a good forecasting instrument. In addition, it so turned the initial hypothesis into an empiric, quantifiable, and valuable scientific theory.
|previous page||next page|