Combined mining: analyzing object and pattern relations for discovering and constructing complex yet actionable patterns

Combined mining is a technique for analyzing object relations and pattern relations, and for extracting and constructing actionable knowledge (patterns or exceptions). Although combined patterns can be built within a single method, such as combined sequential patterns by aggregating relevant frequent sequences, this knowledge is composed of multiple constituent components (the left hand side) from multiple data sources, which are represented by different feature spaces, or identified by diverse modeling methods. In some cases, this knowledge is also associated with certain impacts (influence, action, or conclusion, on the right hand side). This paper presents an abstract high-level picture of combined mining and the combined patterns from the perspective of object and pattern relation analysis. Several fundamental aspects of combined pattern mining are discussed, including feature interaction, pattern interaction, pattern dynamics, pattern impact, pattern relation, pattern structure, pattern paradigm, pattern formation criteria, and pattern presentation (in terms of pattern ontology and pattern dynamic charts). We also briefly illustrate the concepts and discuss how they can be applied to mining complex data for complex knowledge in either a multifeature, multisource, or multimethod scenario.