Mission-Driven Logistics Optimization using Machine Learning

GCSS-A is of critical importance in supporting Army supply, maintenance, and logistics operations, in addition to serving as the financial system of record. The current GCSS-A implementation has a cumbersome user experience and user interface and does not provide machine learning/artificial intelligence capabilities to support modern day command, maintenance, and supply operations. We propose to build on our prior experience with data pipelines, machine learning, procurement/inventory management, and modern user application design to deliver a functional extension to GCSS-A by the completion of Phase I. Capabilities include, but are not limited to, improved data validation and user input, enhanced data visualization, and machine learning methods to support Mission-Based Forecasting (MBF). The proposed Phase I work includes: 1) Integration of representative data into the data pipeline for visualization, exploration, and model development. The data pipeline provides for automated data cleaning and reconciliation of disparate data sources, 2) Conducting user workshops with stakeholders via SPARTN-referenced soldier touchpoints. 3) Iterating on user workflows (including data entry) and application functionality through regular deployments of new prototype functionality, and 4) Training and validating machine learning models to support data quality and to provide decision support for mission/exercise planning (including asset selection, preparation, and mobilization). The proposed work is aligned with our existing efforts with the US Marine Corps. Specifically, we are working to extend the functionality of GCSS-MC using machine learning and with an improved application interface. Prior experience includes the application of machine learning to predict work requests, equipment operational status, and unit readiness.