Advanced Forecasting Methods and Applications with RFID Data Inputs
This research project is to investigate advanced forecasting algorithms for accurately forecasting irregular demand pattern with external factors, explore their applications with Radio Frequency Identification (RFID) real-time data inputs, and enhance SIMForecaster.
Contact PersonYuan Xue Ming()
The project focuses on investigating three classes of advanced forecasting methods, namely Croston’s methods, non-parametric methods and support vector regression methods, which cater for irregular demand patterns and external factors influencing demand patterns. The purpose is to develop new optimal forecasting algorithms based on these three classes of advanced forecasting methods to accurately forecast irregular demand patterns with external factors. The project also explores the possibilities for using real-time RFID data to improve forecasting accuracy.
The applications will be for all the parties in a global supply chain including suppliers, manufacturers, distribution centres, 3rd party logistics services providers (3PL), retailers, shops and customers. SIMForecaster is able to be customised for integration with companies’ existing enterprise systems such as Manufacturing Resource Planning (MRP), Enterprise Resource Planning (ERP), SAP, etc.
The two major research issues which the project addresses are as follows:
- Many companies which we are working with are facing difficulties in predicting irregular demand patterns. There is lack of proper algorithms to predict these demand patterns well.
- Some external factors such as promotion scheme, pricing, contingent events, may affect outcomes of demand forecasting much. Very often, there is a lack of data to describe these external factors. It is difficult to specify the distribution check form for impacts of these external factors.