A COMPARATIVE-STUDY FOR FORECASTING USING NEURAL NETWORKS VS GENETICALLY IDENTIFIED BOX AND JENKINS MODELS

This paper aims to discuss the results and conclusions of an extensive comparative study on the forecasting performance between two differen t techniques: a genetic expert system in which a genetic algorithm car ries out the identification stage embraced in the three-phase Box&Jenk ins univariate methodology; and a connectionist approach. At the heart of the former, an expert system rules the identification-estimation-d iagnostic checking cyclical process to end up with the predictions pro vided by the SARIMA model which best fits the data. We will present th e connectionist approach as technically equivalent to the latter proce ss and due to its, alas, lack of any conclusive existent algorithm abl e to identify both the optimal model and architecture for a given prob lem, the three most common models presently at use and 20 different ar chitectures for each model will be examined. It seems natural that if a comparison is to be made in order to provide a straight answer as to whether or not a connectionist approach outperforms the univariate Bo x&Jenkins methodology, the benchmark should clearly be the set of time series analysed in the work 'Time Series Analysis. Forecasting and Co ntrol' by G. E. Box and G. M. Jenkins. Series BJA through to BJG give a total of 1200 plus measures to evaluate and compare the predictive p ower for different models, architectures, prediction horizons and pre- processing transformations.