AN EXPERIMENTAL STUDY ON BANK FORECASTING USING REGRESSION DYNAMIC LINIER MODEL
DOI:
https://doi.org/10.33050/ccit.v5i1.486Keywords:
Forecasting, Regression variables, RDLM, BMA, AICAbstract
Nowadays, forecasting is developed more rapidly because of more systematicaly decision making process in companies. One of the good forecasting characteristics is accuration, that is obtaining error as small as possible. Many current forecasting methods use large historical data for obtaining minimal error. Besides, they do not pay attention to the influenced factors. In this final project, one of the forecasting methods will be proposed. This method is called Regression Dynamic Linear Model (RDLM). This method is an expansion from Dynamic Linear Model (DLM) method, which model a data based on variables that influence it. In RDLM, variables that influence a data is called regression variables. If a data has more than one regression variables, then there will be so many RDLM candidate models. This will make things difficult to determine the most optimal model. Because of that, one of the Bayesian Model Averaging (BMA) methods will be applied in order to determine the most optimal model from a set of RDLM candidate models. This method is called Akaike Information Criteria (AIC). Using this AIC method, model choosing process will be easier, and the optimal RDLM model can be used to forecast the data. BMA-Akaike Information Criteria (AIC) method is able to determine RDLM models optimally. The optimal RDLM model has high accuracy for forecasting. That can be concluded from the error estimation results, that MAPE value is 0.62897% and U value is 0.20262.
Downloads
References
Aplevich, J., (1999). The Essentials of Linear State-Space Systems. J. Wiley and Sons.
Grewal, M. S., Andrews, A. P., (2001). Kalman Filtering: Theory and Practice Using MATLAB (2nd ed.). J. Wiley and Sons.
Harvey, A., (1994). Forecasting, Structural Time-series Models and the Kalman Filter. Cambridge University Press.
Mubwandarikwa, E., Faria A.E. (2006) The Geometric Combination of Forecasting Models Department of Statistics, Faculty of Mathematics and Computing, The Open University.
Mubwandarikwa, E., Garthwaite, P.H., dan Faria, A.E., (2005). Bayesian Model Averaging of Dynamic Linear Models. Department of Statistics, Faculty of Mathematics and Computing, The Open University.
Turkheimer, E., Hinz, R. and Cunningham, V., (2003), On the undesirability among kinetic models: from model selection to model averaging. Journal of Cerebral Blood Flow & Metabolism, 23, 490-498.
Verrall R. J. (1983). Forecasting The Bayesian The City University, London
West , Mike. (1997). Bayesian Forecasting, Institute of Statistics & Decision Sciences Duke University.
World Primary Commodity Prices.(2002). Diambil pada tanggal 23 Mei 2008 dari http://www.economicswebinstitute.org.
Yelland, Phillip M. & Lee, Eunice. (2003), Forecasting Product Sales with Dynamic Linear Mixture Models. Sun Microsystem.
Zainun, N. Y., dan Majid, M. Z. A., (2003). Low Cost House Demand Predictor. Universitas Teknologi Malaysia.
