Theory and Implementation of Anticipatory Data Mining

Analysing data with data mining techniques needs much time – especially, if the data set that is analysed is very large. Data mining is an important phase in the knowledge discovery in databases process (KDD process). Yet, it is only a part of the KDD process. Improving data mining also improve the KDD process but the improvement can be minor because improving data mining affects only a single phase of a set of phases. Other phases such as the pre-processing phase contribute much to the total time of a KDD project. Commonly, it is necessary to iterate the phases pre-preprocessing and data mining before the result of the data mining phase satisfy the analyst's requirements. Again, repeating phases also worsens the performance of total project time. This dissertation presents a new method to improve performance and quality of the KDD process. The idea is to pre-compute intermediate results which depend on no specific setting of any analysis. When the specific setting of an analysis becomes clear, the data mining system can compute the final result of that analysis using the intermediate results.