Action Rules Discovery: Leveraging Attributes Correlation Based Vertical Partitioning

This paper tackles the challenge of extracting actionable insights from large datasets to enhance the knowledge bases of recommendation systems. We introduce a novel vertical dataset partitioning method utilizing attribute correlation clustering, enabling efficient parallel action rule discovery and significant processing time reduction. Our method’s effectiveness is demonstrated by comparing it with traditional random-based partitioning, focusing on precision, coverage, lightness, rule yield, and efficiency. Employing a rule-based generation approach with the RSES tool, we analyze rule quality, quantity, and computational demand. The results reveal promising strategies for action rule discovery with large-scale datasets, with potential applications across various domains like e-commerce and healthcare, offering valuable insights for large-scale dataset analysis.