Population-Level Modelling for Truck Fleet Survival Analysis

Abstract

Population-level modelling is used to address issues of data sparsity in the survival analysis of a simulated truck fleet. Specifically, hierarchical Bayes with mixed effects improves the predictive capability of hazard models. A set of correlated functions is learnt over the vehicle population – in a combined model – to approximate fleet predictors. Model uncertainty is reduced when sub-fleets of vehicles are allowed to share correlated information. In turn, vehicle groups with incomplete data (automatically) borrow statistical strength from data-rich groups.