Decision Mining with User Preference

Abstract

We are drowning in information, but starving for knowledge. It is necessary to extract interesting knowledge from large and raw data. Traditional data mining approaches discover knowledge based on the statistical significance such as frequency of occurrence, which leads a large number of highly frequent rules are generated. It is a tough work for users to manually pick the rules that they are really interested. It is necessary to prune and summarize discovered rules. Most importantly, different users have different interestingness in the same knowledge. It is difficult to measure and explain the significance of discovered knowledge without user preference. For example, two rules Perfume → Lipstick and Perfume → Diamond may suggest different potential profits to a sales manager, although both are frequent rules.