| provides methods and corresponding tools which assist the developer of a KDD application in building up a high quality KDD process with as less effort as possible. Our approach is on the one hand based on the notion of task analysis and reusable problem-solving methods. On the other hand we exploit and integrate techniques to describe the characteristics of the available data to further guide the selection of applicable methods and algorithms. We use statistical measures, measures from the field of machine learning, e.g. missing values or noise, as well as information available in the data dictionary, e.g. attribute types or the size of relations. | |