A comparison of time series forecasting between artificial neural networks and box and jenkins methods

• Joanna Collantes Duarte Universidad de Los Andes-Venezuela
• Gerardo Colmenares La Cruz Universidad de Los Andes-Venezuela
• Giampaolo Orlandoni Merli Universidad de Los Andes-Venezuela
• Franklin Rivas Echeverría Universidad de Los Andes-Venezuela

Palabras clave: forecasting time series, Box and Jenkins methodology, ARIMA model, transfer function model, artificial neural network, neo-fuzzy neuron

Resumen

This paper deals with a comparison between Box and Jenkins methodologies and Artificial Neural Networks on time series forecasting. ARIMA and Transfer Function Models are compared with Neural Network Models. Performance oft he building models are analysed using comparative criteria during the prediction and fitness stage.