Prediction Comparison by using Transfer Function Models

Abstract

Prediction of time series is the most important and widest spread for researcher nowadays, for it's importance in different approaches ,especially study of natural phenomena .This study deals with the prediction of fuzzing pattern matching models which use three algorithms;(Singh,2001) algorithms ,(Altai,2010) algorithms ,while the third algorithm is an evolution algorithm which combined the two algorithms, in addition we make some correction which leads to give a better results than both algorithms and also the prediction of multiple time series models used which called transfer function models . These methods were applied on the monthly data of time series for Tigris river in Mosul city for the year 2010, we take five variables represented by Turbidity as an output variable while input variables were represented by PH value , Chemical oxygen demand , Total dissolved solids and Electrical conductivity, and by using prediction criteria of fixed or exact adjustment, The results show that the transfer function models give more exact results than evolution algorithm.