Dynamic Forecasting Using a Modified Least Mean Square Algorithm

摘要

In this paper, a dynamic forecasting method based on an adaptive time-series model is proposed for shorttermrnforecasting. The method employs an autoregressive (AR) time-series model to carry out the forecasting process.rnA modified least mean square (MLMS) algorithm is used for adjusting the AR model coefficients so as to minimisernthe sum of square of forecasting errors. A prototype dynamic forecasting system has been built and experimentallyrnverified. The results indicate that the forecasting values agree well with the actual values except for some pointsrnwhich are volatile in nature. Potential applications of the system can be found in sales demand forecast, inventoryrnmanagement as well as collaborative planning, forecast and replenishment (CPFR) in production logistics.