A method of classifying an integrated data base consisting of data from different sources is presented. The data consist of Landsat multispectral scanner imagery, structural information, geochemical measurements, radiometric measurements, and areomagnetic readings. The data are interpolated to the same grid. This, however, gives an immense amount of data, making data reduction necessary. Information was used to preserve statistics to describe the data within squares of 5 X 5 km. Some of the squares are known to have a high or a low potential for mineralization. These squares were then used as training areas in a multivariate classification, making it possible to classify the rest of the squares as having a high or a low potential for mineralization. The method is evaluated in southern Greenland, and a geologic evaluation of the classifications shows a good agreement between the classified results and the geologic evaluation.