Sarcouncil Journal of Engineering and Computer Sciences

Abstract: The increasing interest in the topic of resilient, time-varying data storage systems aims to improve the reliability and scalability of forecasting systems. There are now many incredibly broad aspects of financial data that can be characterized as complex, non-linear, and/or unstable due to various factors occurring in the global marketplace. Thus, systems with time-varying data must be flexible, secure, and able to ingest large quantities of temporally dynamic data. This review draws on multiple disciplines to examine recent advances in modeling dynamic systems, resilient architecture, tolerable load uncertainty, vulnerability to managed cyberattacks, and intelligent information allocation that contribute to resiliency associated with forecasting systems in the energy, hydrology, and behavioral sciences. The amalgamation of multiple-application approaches begins to inform a conceptual framework for resilient forecasting architecture within the financial sector. A resilient state-space network that utilizes distributed machine storage and infrastructure, as well as machine-learning models for forecasting, is reviewed to demonstrate when it may be reliable to use time-varying data to provide resilient operational capacity supported by accumulated data or high-fidelity forecasting. The paper also provides a summary comparison table, an architectural diagram, and a graphical summary of proposed performance that characterizes the resilient forecasting architecture.