Analytical Query Patterns: Domain Independent and Domain Specific Cores of Analytical Queries
Sprache des Vortragstitels:
Dagstuhl-Seminar on Next Generation Domain Specific Conceptual Modeling, 18. ? 23. November 2018, Dagstuhl-Seminar 18471
Sprache des Tagungstitel:
Analytical query patterns capture the reusable core of analytical queries. They are in business analytics the counterpart to design patterns in software engineering.
An analytical pattern is defined by (1) a set of pattern elements, (2) a set of constraints over pattern elements, (3) an a pattern expression with pattern elements embedded in some analytical query language such as SQL. Each pattern element is a named placeholder of one or a list of element(s) of the dimensional-fact-model (DFM) of a data warehouse, such as dimension, level, fact, or measure. Pattern elements may be input-parameters, result-parameters, or local pattern elements. Analytical patterns are best exploited by using an enriched the DFM-model that comes with ontologies of predicates (over facts or dimensions) and of calculated measures that can also constitute elements of an analytical pattern.
An analytical pattern is practically/fully instantiated by binding some/each formal input parameter element either to the name for a DFM-element (to be bound later during application) or to the identity, e.g., URI, of a DFM-element (static binding). A fully instantiated analytical pattern is applied to a specific data warehouse (DWH) by identifying the data warehouse context and by dynamically binding the name of actual pattern elements to the identity of a DFM-element in the indicated DWH-context. A pattern application is valid if the bindings of the pattern elements satisfy the pattern constraints.
We present a set of domain-independent analytical patterns identified by generalizing similar analytical queries frequently accounted in various domains (such as medicine, farming, and production). We present selected domain-independent analytical patterns that are defined by partially instantiating domain-independent patterns with facts, dimensions, levels, predicates, and calculated measure of the domain ontology, whereby bound input elements become local elements.
Sprache der Kurzfassung:
Hauptvortrag / Eingeladener Vortrag auf einer Tagung