COMBINATION FORECASTING USING CLUSTERIZATION

Combination forecasts are generated using predictions obtained from a group of forecasters. The forecasters are first divided into a number of pre-defined clusters, which typically will have been formed using statistical clustering techniques. In particular, clusters of forecasters can be formed based on similarities of the forecasters' predictions. Then, statistical data are calculated for each pre-defined cluster (e.g., measures of central tendency and dispersion). Finally, the statistical data for all the pre-defined clusters are combined so as to obtain a combination forecast.