A decision support system for DM algorithm selection based on module extraction
Man Tianxing, Nataly Alexandrovna Zhukova, Aung Myo Thaw, Saddam Abbas · Procedia Computer Science · 2021
Data mining techniques are needed in various fields. However, most data researchers do not have sufficient knowledge and experience. Due to the significant number of algorithms and parameters for data analysis, intuition-based decisions can’t lead to optimal solutions. Ontology is widely adopted to build knowledge-driven decision support systems since it is suited to encapsulate the concepts and relationships of terms associated with various domains. It is suitable for capturing knowledge by computers, allowing sharing and reusing it whenever necessary. All concepts and relationships about data analysis can be described using Web Ontology Language (OWL). Its reasoning mechanism is vital in any knowledge-based system. Ontology can be reasoned to recommend suitable solutions for a specific analysis task by considering the data characteristics and task requirements. In the previous work, we have developed a system that describes comprehensive knowledge and logical internal relationships about data mining. We found that the reasoning and query complexity is high in the large size ontology, so this paper focuses on the modeling and implementation of an ontology-based decision support system for data mining algorithm selection that contains a new sub-ontology extraction method. Extracting effective content can significantly reduce the size of the ontology. This extraction method can improve query efficiency on the premise of ensuring the quality of information.