長崎大学経済学部

研究活動

No.2002-04 (September 2002)
Adaptive Demand-Forecasting Approach based on Principal Components Time-series
(an application of data-mining technique to detection of market movement)
Toshio sugihara (Professor, Faculty of Economics, Nagasaki University)

Abstract:
In this study, an adaptive demand-forecasting approach adopting the data-mining technique which detects the correlation between the target variable and other related elements, is proposed. With the inclusion of the scheme of the time-series analysis based on the state-space approach, this approach has two characteristic points. One is the state-space which is formed by principal components composed of various market variables. Another is the self-organization of the state-space using a neural network. This approach is applied to the two cases of the demand-forecasting. Some comparisons of forecasting accuracy (extrapolation test) with this approach and non-self-organizing models (AR, etc) are used to evaluate the effectiveness of the proposed approach. Consequently, we achieved significantly higher accuracy using this approach compared to the other approaches.

Keywords:
Adaptive demand-forecasting; Data-mining Approach, (Extended) Kalman Filter; Principal Component Analysis; Neural Networks.