Pre-processing of Meteorological Data in Knowledge Discovery

Abstract

As an application of Knowledge Discovery from Meteorological Databases, we attempt to relate recorded precipitation data from the Mikra Meteorological Station in Thessaloniki, Greece, to the recorded data of the closest to the station ERA-40 node (i.e., node with latitude 400N and longitude 22.50 E). We use the daily values of the ERA-40 meteorological data from the European Centre for Medium-Range Weather Forecasts (four times a day at 00h, 06h, 12h and 18h) for a period of 42 years (1960- 2001). We examine different scenarios in the pre-processing phase of the input data and we study their impact on the performance of data mining algorithms designed to predict the occurrence of precipitation in Mikra Meteorological Station. The goal is to determine the appropriate pre-processing on the input file that can ensure effective application of data mining techniques.