Intelligent data analysis applied to debug complex software systems

摘要

The emergent behavior of complex systems, which arises from the interaction of multiple entities, can be difficult to validate, especially when the number of entities or their relationships grows. This validation requires understanding of what happens inside the system. In the case of multi-agent systems, which are complex systems as well, this understanding requires analyzing and interpreting execution traces containing agent specific information, deducing how the entities relate to each other, guessing which acquaintances are being built, and how the total amount of data can be interpreted. The paper introduces some techniques which have been applied in developments made with an agent oriented methodology, 1NGENIAS, which provides a framework for modeling complex agent oriented systems. These techniques can be regarded as intelligent data analysis techniques, all of which are oriented towards providing simplified representations of the system. These techniques range from raw data visualization to clustering and extraction of association rules.