Rule Confirmation Measures: Properties, Visual Analysis and Applications
[ 1 ] Instytut Informatyki, Wydział Informatyki i Telekomunikacji, Politechnika Poznańska | [ P ] pracownik
2022
rozdział w monografii naukowej
angielski
- rules
- evaluation measures
- confirmation
EN According to Bayesian confirmation theory, for a E -> H rule, evidence E confirms hypothesis H when E and H are positively probabilistically correlated. Surprisingly, this leads to a plethora of non-equivalent quantitative measures that attempt to measure the degree in which E confirms H. This observation has triggered research on the differentiating characteristics of confirmation measures --- their analytical properties, tools for visual inspection of those properties, and the applications of confirmation measures in rule-based systems. This paper constitutes an extensive overview that covers the analysis and development of rule confirmation measures and their properties. It moves from research on desirable properties of confirmation measures, through visualization methods that support this process, to current applications of rule confirmation measures and lines of future research in the field.
09.02.2022
401 - 423
20