KNOWLEDGEABLE LEARNING USING MOBAL - A MEDICAL CASE-STUDY

Building up a knowledge base is a complex task in which theoretical kn owledge needs to be integrated with practical experience. This integra tion can be supported by a system that can manage linking between rule s, representing experts or textbook or theoretical knowledge and facts (or data), representing cases from practice. Conflicts between rules and real-world cases can have diverse causes. Case data can be noisy o r inconsistent or both. We use rules to filter cases and confine noise or inconsistency. However, rules can be overly general and can classi fy more cases than intended. Then machine learning can be used to find additional restrictions for the rules. If, however, no significant si milarities can be determined between the misclassified cases, we seek additional expert input to support effective implementation of learnin g tools. Using a case subset, we capture expert input and formulate a set of ''enriched'' cases. We then use the enriched cases to learn add itional rules and to introduce practical features to validate and refi ne the original rule base. In this paper, we discuss the role of machi ne learning as a vehicle for supporting rule and data validation. We d emonstrate our results by an application of the MOBAL system to a real -world medical domain.