Decision Guidance Analytics Language

This paper proposes a unified syntax and formal semantics for DGAL(X)—a Decision Guidance Analytics Language—for easy iterative development of decision guidance systems, where X is an external functional language, e.g., a noSQL query language such as JSONiq. It demonstrates the composability of analytic models using a case study of modeling and operational optimization of hierarchical service networks, described as recursive analytic models. Analytic models formally describe feasibility constraints and metrics of interest as deterministic/stochastic functions of fixed and control parameters. DGAL(X) provides a library of operators, including compute, predict, calibrate, and optimize. The uniqueness of DGAL(X) lies in the modularity and composability of simulation-like analytic models without manually crafting mathematical programming (MP), constraint programming (CP), and machine learning (ML) models. This results in the quality of optimization and machine learning results as well as the computational efficiency of the best underlying MP, CP, and ML algorithms, which outperform black-box-based simulations.