Applying Kalman Filter for Correlated Demand Forecasting

Product demands are known to be serially correlated. In this research, a first order autoregressive model, AR (1), is utilized to simulate product demand processes whose behavior are stationary. Since demand forecasting is important to the efficiency improvement of product supply chain system, different forecasting techniques are utilized to predict product demand. In this research, Kalman filter is deployed to forecast demand simulated by AR (1) model. Product demands are simulated at the different degrees of autoregressive coefficients. After the application of Kalman filter to the designated data, the forecasting errors are calculated and the results indicate that Kalman filter is an efficient technique to predict demands in the future.