Using Semantic Data Mining for Classification Improvement and Knowledge Extraction

The objective of this position paper is to show that the integration of semantic data mining into the DAMIART data mining system can help further improve classification performance and knowledge extraction. DAMIART performs multi-label classification in the presence of multiple class ontologies, hierarchy extraction from multi-labels and concept relation by association rule mining. Whereas DAMIART combines knowledge from multiple data sources and multiple class ontologies, the proposed extension should also explore available ontologies over attributes. This will allow the system to produce not only more accurate classification results but also improve their interpretability and overcome such problems as data sparseness.