ADynamicForecasting SystemwithApplications inProductionLogisticsCHEUNG Chi-fail, LEE Wing-bunl, LO Victor2, WANG Wai-mingl(1. Department of Industrial and Systems Engineering, The Hong KongPolytechnic University Hung Hom, Kowloon, Hong Kong, China,2. Kaz (Fast East) Ltd, Room 1705-07, Kodak House II, 39 Healthy Street EastNorth Point, Hong Kong, China)Abstract: Production logistics involve the co-ordination of activities such as production and materialscontrol (PMC), inventory management, product life cycle management, etc. Those activities demandfor an accurate forecasting model However, the conventional methods of making sell and buydecision based on human forecast or conventional moving average and exponential smoothingmethods is no longer be sufficient to meet the future need. Furthermore, the underlying statistics ofthemarket information change from time to time due to a number of reasons such as change of globaleconomic environment, government policies and business risks. This dema nds for highly adaptiveforecasting model which is robust enough to response and adapt well to the fast changes in the datacharacteristics, in other words, the trajectory oftheOdynamic characteristicsOof the data.In this paper, an adaptive time-series modelling method was proposed for short-term dynamicforecasting, The method employs an autoregressive (AR) time-series model to carry out thefore-casting process. A modified least mean square (MLMS) adaptive filter algorithm was establishedfor adjusting the AR model coefficients so as to minimise the sum ofsquared of forecasting errors. Aprototype dynamic forecasting system was built based on the adaptive time-series modelling method.Basically, the dynamic forecasting system can be divided into two phases, ie. the Learning Phase andthe Application Phase. The learning procedures start with the determination of upper limit of theadaptation gain based on the convergence in the mean square criterion. Hence, the optimum ELMSfilter parameters are determined using an iteration algorithm which changes each filter parameteriethe order, the adaptation gain and? the valuesinitial coefficient vector one by one inside apredetermined iteration range. The set ofparameters which gives the minimum value for sum ofsquared errors within the iteration range is selected as the optimum set of filter parameters. In theApplication Phase, the system is operated under a real-time environment. The sampled data is
A Dynamic Forecasting System with Applications in Production Logistics CHEUNG Chi-fai1, LEE Wing-bun1, LO Victor2, WANG Wai-ming1 (1. Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University Hung Hom, Kowloon, Hong Kong, China; 2. Kaz (Fast East) Ltd, Room 1705-07, Kodak House II, 39 Healthy Street East, North Point, Hong Kong, China) Abstract: Production logistics involve the co-ordination of activities such as production and materials control (PMC), inventory management, product life cycle management, etc. Those activities demand for an accurate forecasting model. However, the conventional methods of making sell and buy decision based on human forecast or conventional moving average and exponential smoothing methods is no longer be sufficient to meet the future need. Furthermore, the underlying statistics of the market information change from time to time due to a number of reasons such as change of global economic environment, government policies and business risks. This demands for highly adaptive forecasting model which is robust enough to response and adapt well to the fast changes in the data characteristics, in other words, the trajectory of the0dynamic characteristics0of the data. In this paper, an adaptive time-series modelling method was proposed for short-term dynamic forecasting. The method employs an autoregressive (AR) time-series model to carry out the fore-casting process. A modified least mean square (MLMS) adaptive filter algorithm was established for adjusting the AR model coefficients so as to minimise the sum of squared of forecasting errors. A prototype dynamic forecasting system was built based on the adaptive time-series modelling method. Basically, the dynamic forecasting system can be divided into two phases, i.e. the Learning Phase and the Application Phase. The learning procedures start with the determination of upper limit of the adaptation gain based on the convergence in the mean square criterion. Hence, the optimum ELMS filter parameters are determined using an iteration algorithm which changes each filter parameteri.e. the order, the adaptation gain and? the valuesinitial coefficient vector one by one inside a predetermined iteration range. The set of parameters which gives the minimum value for sum of squared errors within the iteration range is selected as the optimum set of filter parameters. In the Application Phase, the system is operated under a real-time environment. The sampled data is
processed by the optimised ELMS filter and the forecasted data are calculated based on the adaptivetime-series model The error of forecasting is continuously monitored within the predefined toleranceWhen the systemdetectsexcessiveforecastingerror,afeedback alarm signal was issuedfor systemre-calibration.Experimental results indicated that the convergence rate and sum of squared errors during initialadaptation could be significantly improved using the MLMS algorithm. The performance of thesystem was verified through a series ofexperiments conducted on the forecast ofmaterials demandandcostinginproductionlogistics.Satisfactoryresultswereachievedwiththeforecast errorsconfining within in most instances. Further applications of the system can be found in sales demandforecast,inventory management as well as collaborative planning,forecast and replenishment (CPFR)inlogisticsengineeringKeywords:adaptivetime-seriesmodel, dynamicforecasting.productionlogistics,modifiedleastmean squarealgorithm
processed by the optimised ELMS filter and the forecasted data are calculated based on the adaptive time-series model. The error of forecasting is continuously monitored within the predefined tolerance. When the system detects excessive forecasting error, a feedback alarm signal was issued for system re-calibration. Experimental results indicated that the convergence rate and sum of squared errors during initial adaptation could be significantly improved using the MLMS algorithm. The performance of the system was verified through a series of experiments conducted on the forecast of materials demand and costing in production logistics. Satisfactory results were achieved with the forecast errors confining within in most instances. Further applications of the system can be found in sales demand forecast, inventory management as well as collaborative planning, forecast and replenishment (CPFR) in logistics engineering. Key words: adaptive time-series model, dynamic forecasting, production logistics, modified least mean square algorithm