A fuzzy case-based system for weather prediction

Case-based reasoning is emerging as a leading methodology for the application of artificial intelligence. We describe an investigation into the application of case-based reasoning in airport weather forecasting. Knowledge about temporal features that human forecasters use to construct analogous climatological scenarios is encoded in a fuzzy similarity measure. The fuzzy similarity measure is used to locate the k-nearest neighbours from the historical database. These nearest neighbours are in turn adapted to produce values for the forecast parameters. Five sets of experiments show inter alia that the proposed WIND-1 system produces highly accurate forecasts based on real climatological data, using a standard technique for assessing the accuracy of forecasts produced by human forecasters.