Ontology-based data mining model management for self-service knowledge discovery

Data mining (DM) models are knowledge-intensive information products that enable knowledge creation and discovery. As large volume of data is generated with high velocity from a variety of sources, there is a pressing need to place DM model selection and self-service knowledge discovery in the hands of the business users. However, existing knowledge discovery and data mining (KDDM) approaches do not sufficiently address key elements of data mining model management (DMMM) such as model sharing, selection and reuse. Furthermore, they are mainly from a knowledge engineer’s perspective, while the business requirements from business users are often lost. To bridge these semantic gaps, we propose an ontology-based DMMM approach for self-service model selection and knowledge discovery. We develop a DM³ ontology to translate the business requirements into model selection criteria and measurements, provide a detailed deployment architecture for its integration within an organization’s KDDM application, and use the example of a student loan company to demonstrate the utility of the DM^3.