International Journal of Management Science and Engineering Management

Abstract. An overview of statistical forecasting methodology is given, focusing on techniques appropriate to short- and medium-term forecasts. Topics include basic definitions and terminology, smoothing methods, ARIMA models, regression methods, dynamic re

ISSN1746-7233,England,UK

International Journal of Management Science

and Engineering Management

V ol.1(2006)No.1,pp.17-36

An overview of short-term statistical forecasting methods

Russell J.Elias,Douglas C.Montgomery∗,Murat Kulahci

Department of Industrial Engineering,Arizona State University

(Received June132006,Accepted July102006)

Abstract.An overview of statistical forecasting methodology is given,focusing on techniques appropriate to short-and medium-term forecasts.Topics include basic definitions and terminology,smoothing methods, ARIMA models,regression methods,dynamic regression models,and transfer functions.Techniques for evaluating and monitoring forecast performance are also summarized.

1Introduction

Achieving accurate forecasts of future market demand is crucially important for today’s global man-ufacturing enterprises.Increasingly competitive pressures force manufacturers to closely match production capacities and mix-dependent run volumes to highly dynamic market conditions;excessive capacity results in a depressed return on assets,while inadequate capacity leads to lost market share.Resulting from the lean operating margins typical in today’s highly competitive global business environment,excessive forecast error on a continuing basis is an unacceptable mistake that gobbles profits and can push otherwise viable businesses into the negative returns category.

The critical importance of accurate short-to medium-term business forecasts is amplified today as a result of several revolutionary changes in the nature of international commerce.For example,consider the relatively recent emergence and proliferation of non-captive manufacturing entities.A fundamental shift is occurring throughout many industry structures,away from in-house manufacturing by the original equipment manu-facturers(OEM’s)and towards the use of second tier,merchant market manufacturing concerns known as foundries.For example,in the semiconductor industry,the cost of a new state-of-the-art fabrication facility now approaches two billion dollars,as cited by[21].Many semiconductor OEM’s are responding to this incredible capital requirement by going“fabless”;outsourcing much if not all of their wafer fabrication to non-captive,merchant foundaries that are able to service multiple OEM clients from high volume integrated manufacturing centers,as described by[1].This shift away from captive manufacturing resources towards merchant foundaries implies that the OEM’s will now be competing with each other for limitedfinite capacity at these foundries.The competitor that has superior market forecasting can contractually lock-in the required manufacturing capacity at these foundries,excluding their competitors(who potentially underforecasted true demand)from utilizing this resource against them in the battle for market share.Conversely,competitors who overestimate market demand would suffer either increases in non-performing inventory or the payment of cancellation charges to the foundries,both of which represent costly errors.

Directly related to the rise of merchant manufacturing is the emergence of a global economy.Goods and services are now able toflow more freely across international boundaries than at any point in history, permitting both customers and OEM’s to fulfill sourcing requirements anywhere in the world.While this new business model permits the utilization of the most efficient manufacturing resources,it also inherently in-troduces greater volatility for all market participants.More so than ever,a manufacturer has greater upside ∗E-mail address:doug.montgomery@asu.edu.

Published by World Academic Press,World Academic Union

第1页下一页