Forecasting financial time series via adaptive model synthesis



Journal Title

Journal ISSN

Volume Title



Four techniques for time series forecasting are analyzed and combined in an artificial intelligence scheme In this thesis. The techniques are: the "three point reversal point and figure" method, the Ney model, the Zweig model, and the autoregressive-integrated-moving average model. The literature search which led to the selection of these techniques from the great number available Is discussed. The methodology of each forecaster Is described In detail and examples of each are shown. The computer programs produced are listed In the appendices. This forecasting technique Is unique in that It is the first time that several recognized forecasters have been combined In a dynamic learning algorithm. The results suggest that this technique Is accurate and has great practical value for three to six month ahead economic time series forecasting.