Selection of an appropriate time series model and estimation of its parameters may become very challenging tasks for short series of observations. The state-of-the-art information criteria often fail to adequately identify the predictive model for the small sample sizes, for which real-life applications are routinely made. Within this research, we propose a forecasting approach for averaging across multiple predictive models and demonstrate its usefulness especially for small samples. The proposed method incorporates selected data-mining techniques and similarity measures to find most appropriate weights. The performance of the proposed method is illustrated with simulation study for stationary processes and the experimental study for the benchmark datasets.
CITATION STYLE
Kaczmarek-Majer, K., & Hryniewicz, O. (2018). Data-mining approach to finding weights in the model averaging for forecasting of short time series. In Advances in Intelligent Systems and Computing (Vol. 642, pp. 314–327). Springer Verlag. https://doi.org/10.1007/978-3-319-66824-6_28
Mendeley helps you to discover research relevant for your work.