Time-constrained fashion sales forecasting by extended random vector functional link model

Forecasting is about providing estimation of the future that cannot be observed at the moment. In this chapter, the random vector functional link (RVFL), which is a variation of the artificial neural networks (ANN) model, is used in establishing a fashion sales forecasting model. It is well-known that the RVFL inherits the learning and approximation capability of ANN, while running much faster than the traditional ANN. In order to develop a real world forecasting application, we propose a time-constrained forecasting model (TCFM), implemented by an extended RVFL, in which the user can define the time limit and a precision threshold for yielding the forecasting result. Real datasets collected from a fashion retail company are employed for the analysis. Our experiment has shown that the proposed TCFM can produce quality forecasting within the given time constraint. Future research directions are outlined.