	US 7,562,062 B2
	Forecasting system tool
Cedric Ladde, London (United Kingdom); and Anargyros Garyfalos, Ipswich (United Kingdom)
Assigned to British Telecommunications plc, London (United Kingdom)
Filed on Mar. 23, 2006, as Appl. No. 11/387,046.
Claims priority of application No. 05252022 (EP), filed on Mar. 31, 2005.
Prior Publication US 2006/0247859 A1, Nov. 02, 2006
Int. Cl. G06F 17/00 (2006.01); G06N 5/02 (2006.01)

U.S. Cl. 706—47 [706/12; 706/13]

12 Claims

1. A method of generating a forecast system for generating forecasts of a set of time-series data stored in one or more data stores using a forecast system development tool implemented by one or more computer programs running on one or more devices, the method comprising:

using the tool to:

provide a plurality of functions for a casting heuristic algorithm component using one or more generically structured core forecast algorithm components and one or more user-selected forecast algorithm components, and

determine a tuning sequence for said core and user-selected forecast algorithm components by

associating said core and user-selected forecast algorithm components with a framework, wherein the structure of the framework determines the tuning sequence for said core and user-selected forecast algorithm components,

associating each said core and user-selected forecast algorithm components with a set of one or more rules provided by said forecast heuristic algorithm component,

tuning a first forecast algorithm component provided by a first one of said core and user-selected algorithm components in said predetermined tuning sequence determined by said framework using said time-series data to form conditioned time-series data, and

tuning subsequent forecast algorithm components in said predetermined tuning sequence using said conditioned time-series data comprising time-series data conditioned by all of the tuning processes previously performed in said predetermined tuning sequence,

wherein collectively the sequence of core and user-selected forecast algorithm components determined by the framework and said heuristic forecast component generates said forecast system for generating a forecast of said set of time-series data.