Data Mining with Histograms and Domain Knowledge - Case Studies and Considerations

An approach to data mining with histograms is introduced. Several examples of applying data mining for various types of histogram are presented. The problem of filtering out uninteresting histograms is described. It is shown that no suitable, logically correct deduction rules to solve this problem exist. Expert deduction rules are introduced as deduction rules supported by indisputable facts, however, incorrect according to mathematical logic. A method for deciding whether a given expert deduction rule is correct according to a given indisputable fact is developed. Applied examples of correct expert deduction rules are described.