MODEL SELECTION OF ENSEMBLE FORECASTING USING WEIGHTED SIMILARITY OF TIME SERIES

Several methods have been proposed to combine the forecasting results into single forecast namely the simple averaging, weighted average on validation performance, or non-parametric combination schemas. These methods use fixed combination of individual forecast to get the final forecast result. In this paper, quite different approach is employed to select the forecasting methods, in which every point to forecast is calculated by using the best methods used by similar training dataset. Thus, the selected methods may differ at each point to forecast. The similarity measures used to compare the time series for testing and validation are Euclidean and Dynamic Time Warping (DTW), where each point to compare is weighted according to its recentness. The dataset used in the experiment is the time series data designated for NN3 Competition and time series generated from the frequency of USPTO’s patents and PubMed’s scientific publications on the field of health, namely on Apnea, Arrhythmia, and Sleep Stages. The experimental result shows that the weighted combination of methods selected based on the similarity between training and testing data may perform better compared to either the unweighted combination of methods selected based on the similarity measure or the fixed combination of best individual forecast.