Kalman Forecasting In A Decision Support System

Abstract

This paper addresses the question of providing a forecasting system with the following properties: (1) user-friendly input avoiding leaming a new syntax or menu structure, (2) user-friendly graphical output, (3) user-friendly response times, (4) sophisticated forecasting: specifically ARIMA modeling, (5) flexible operation, allowing analysis of data other than forecasting. These aims are achieved by implementing ARIMA estimation within a spreadsheet-based decision support system on a microcomputer, Kaiman Filtering is used rather than Box-Jenkins estimation, since it is noniterative.

Résumé

Cet article traite d’un modèle de prévisions ayant ies propriétés suivantes: (1) entree de donnees simple de façon à éviter l’apprentissage d’une nouvelle syntaxe ou structure de menu; (2) production simple de graphiques; (3) temps de reponse rapide; (4) prévisions sophistiquées: spécifiquement les modèles ARIMA (5) opérations flexibles permettant diverses analyses de données en sus des prévisions, Ces objectifs sont réalisés par la mise en place d’estimés ARIMA à l’intérieur d’un système tahleur d’aide à la décision sur micro ordinateur. La méthode de filtre de Kaiman est utilisée plutôt que la méthode Box Jenkins parce qu’elle est non-itérative.

Additional information

Notes on contributors

David Wright

DAVID J, WRIGHT is Associate Professor of Management Science at the University of Ottawa, Faculty of Administration, Ottawa, KIN 6N5, Canada. He received his undergraduate and PH D degrees from Cambridge University, England and has since held government and university positions on both sides of the Atlantic. He holds two govemment research grants and has published in Management Science, Operations Research, Journal ofthe Operational Research Society, IEEE Transactions on Automatic Control, European Journal of Operations Research, Intemational Joumal of Forecasting, Computers and Industiral Engineering, and several other joumals.