Enhancing Interestingness Evaluation in Ontology-Based Association Rules: A Case Study on US Birth Data

In the domain of association rule mining, evaluating the interestingness of discovered rules plays a crucial role in extracting meaningful patterns. However, the context of interestingness poses challenges that call for improvements in rule evaluation. This study focuses on addressing this problem by enhancing the evaluation of interestingness in ontology-based association rules. In this study, we present the effective rule evaluation using the ontology (EREO) model, which aims to evaluate the interestingness of ontology-based association rules in the context of US birth data. The EREO model incorporates three levels of rule evaluation: the utilization of proposed effective measures, consultation with domain experts, and the utilization of AI-based methods. To indirectly evaluate the interestingness of ontology-based rules, we propose two effective measures: Ontology-based rule specificity (ORS) and ontology-based rule complexity (ORC). Rule evaluation is further facilitated by domain experts and AI-based methods, employing an interestingness measurement scale (IMS). Furthermore, we compare the average interestingness scores obtained from ORS, ORC, and the EREO model with those derived from traditional interestingness measures. Our findings demonstrate that the proposed interestingness measures consistently outperform the traditional ones, as indicated by higher average scores. Additionally, we observe a positive relationship between the interestingness scores obtained using the three levels of the EREO model. Overall, this study effectively showcases the efficacy of ontology-based association rule evaluation in improving the quality of discovered rules and supporting informed decision-making processes.