Forecasting of time series with trend and seasonal cycle using the airline model and artificial neural networks

Many time series with trend and seasonal cycles are successfully modeled and predicted using the airline model of Box and Jenkins; However, the presence of nonlinearities in the data is neglected by this model. In this article, a new non-linear version of the airline model is proposed; for this, the linear component of moving averages is replaced by a multilayer perceptron. The proposed model is used to forecast two benchmark time series; It was found that the proposed model is capable of forecasting time series more accurately than other traditional approaches.

• Autores: Velásquez, J D; Franco, C J
• Fecha de publicación: 2012
• Institución: Universidad EAFIT
• Repositorio: Repositorio EAFIT
• Idioma: eng
• OAI Identifier: oai:repository.eafit.edu.co:10784/14444
• Acceso en línea: http://hdl.handle.net/10784/14444
• Palabra clave: Prediction; Nonlinear Models; Sarima; Multilayer Perceptron; Predicción; Modelos No Lineales; Sarima; Perceptrón Multicapa
• License: Copyright (c) 2012 J D Velásquez, C J Franco