In this report, a novel qualitative model learning (QML) framework namedQML-Morven is presented. QML-Morven is an extensible framework and currently includes three QML subsystems, which employ either symbolic or evolutionary approachesas their learning strategies. QML-Morven uses the formalism of Morven, a fuzzy qualitative simulator, to represent and reason about qualitative models, and it also utilisesMorven to verify candidate models. Based on this framework, a series of experiments weredesigned and carried out to: (1) verify the results obtained by the previous QML systemILP-QSI; (2) investigate factors that in¿uence the learning precision and minimum datarequirement for successful learning; (3) address the scalability issue of QML systems.
|Name||Technical Report ABDN–CS–12–03|
- qualitative reasoning
- qualitative model learning
- artificial immune systems
- backtracking with forward checking