Unity: A NoSQL-based Platform for Building Decision Guidance Systems from Reusable Analytics Models

Enterprises across all industries increasingly depend on decision guidance systems (DGSs) to facilitate decisionmaking across all lines of business. Despite significant technological advances, current DGS development paradigms lead to a tight-integration of the analytics models, methods and underlying tools that comprise these systems, often inhibiting extensibility, reusability and interoperability. To address these limitations, this paper focuses on the development of the first NoSQL decision guidance management system (NoSQL-DGMS), called Unity, which enables decision-makers to build DGSs from a repository of analytics models that can be automatically reused for different analytics methods, such as simulation, optimization and machine learning. In this paper, we provide the Unity NoSQL-DGMS reference architecture, and develop the first implementation, which is centered around a modular analytics engine that symbolically executes and automatically reduces analytics models, expressed in JSONiq, into lower-level, tool-specific representations. We conduct a preliminary experimental study on the overhead of OPL optimization models automatically generated from JSONiq using Unity, compared with manually-crafted OPL models. Preliminary results indicate that the execution time of OPL models that are automatically reduced from JSONiq is within a small constant factor of corresponding, manuallycrafted OPL models.