Industrial Production Forecasting with Kalman Filtering
Maria Filomena Teodoro, Theodore E. Simos, George Psihoyios, Ch. Tsitouras · AIP conference proceedings · 2009
Under certains conditions commonly used forecast methods reveal themselves inappropriate for the desired result. Thus it is therefore required to resort to other methods deemed more adequate to the desired results.The main purpose of this work is to investigate the suitability of the Kalman filtering as short term forecast method. The work data set was made of qualitative surveys of conjunture and the index of industrial production. Along the work it is studied the application of this method to a set of qualitative and quantitative information. The ultimate objective is the attainment of short term forecast models for the industrial production index of the transforming industry.After the previous treatment of the data, a number of models are estimated and validated. At the end of this process two models are selected: the two better adjusted in accordance with criteria previously established. The next step in process, following the choice of the models, is the repetition of the sellected models estimation using all sample, after which the actual forecast is done. The results obtained with the described methods have shown sufficiently coincident with the estimates supplied by Portuguese National Institute of Statistics for the same period. The application of Kalman filtering to a set of qualitative and quantitative information in the attainment of models of short term forecasts appears quit promising thus.