While other systems rely either on a single method or make their method selection based on a best fit criterion, the founders of Futurion, Dr. Robert Carbone and Dr. Spyros Makridakis, then leading academics in the field of forecasting, innovated by pioneering the first true AI forecasting engine. With each new iteration, the Auto Pilot machine learning engine identifies whether or not the data exhibit seasonality. It detects outliers, replaces them with most probable quantities, determines what beginning time series observations are to be discarded, and therefore extrapolates only on relevant filtered data. It automatically recognises whether sales are increasing or declining, whether they have plateaued or reached a floor, and accordingly, uses the appropriate statistical method, parameters and/or rule. With each iteration, it re-evaluates its previous actions as machine learning takes place. These core components of the AI engine have been continuously refined through over 30 years of experience with special rules having been integrated to handle erratic and intermittent sales behaviors linked, for example, to either tender businesses or items with small sales volume. The end result is a reliable Auto Pilot forecast that creates a realistic benchmark and reduces the number of user modifications.
