Advanced Analytic Modeling Flow using Bayesian & Classical Tools to Inform Army Fleet Readiness

Music City Center 

The U.S Army Materiel Command (AMC) develops and delivers materiel readiness solutions to ensure globally dominant land force capabilities. AMC manages the global supply chain, synchronizing logistics and sustainment activities across the Army, including using advanced analytics to ensure Army readiness. The AMC Analysis Group released the AMC Innovation Competition: Predicting and Influencing Readiness to crowdsource technical and creative approaches to identify key influential features and develop a robust modeling approach to predict unit-based readiness for the fleet of tactical vehicles. The proposed advanced analytic modeling flow presented in this work uses vetted statistical tools to identify influential markers for fleet readiness, they are then used to construct various time series and machine learning models under both Bayesian and Classical assumptions. The models are compared on model accuracy metrics, the best model is chosen and a global hyperparameter sensitivity analysis conducted for feature validation. Model drift detection methods are implemented to ensure model precision & accuracy remain at acceptable levels for model persistence under Army requirements.

Bayesian Time Series

Machine Learning

Army

Model Drift Detection

Advanced Analytics Modeling Flow

Feature Selection 

Government Statistics Section