A Bayesian Hierarchical Model for the Cost-Effectiveness of Industrial Machinery Fleets in Rwanda: A Methodological Evaluation

{ "background": "The assessment of cost-effectiveness for industrial machinery fleets in developing economies is often hampered by sparse, heterogeneous data and complex operational interdependencies. Traditional deterministic models fail to adequately quantify uncertainty, limiting robust decision-making for asset management and capital investment.", "purpose and objectives": "This study presents a novel Bayesian hierarchical modelling framework designed to evaluate the cost-effectiveness of heavy machinery fleets. Its objective is to provide a robust, probabilistic methodology that integrates multiple data sources and explicitly accounts for operational variability and uncertainty in the Rwandan context.", "methodology": "A three-level Bayesian hierarchical model was developed, with machinery units nested within fleet types and sites. The core model for the log-cost-effectiveness ratio of unit $i$ is specified as $\\text{log}(\\text{CER}i) = \\alpha{j[i]} + \\beta Xi + \\epsiloni$, where $\\alphaj \\sim \\text{Normal}(\\mu{\\alpha}, \\sigma_{\\alpha}^2)$ represents fleet-type varying intercepts. Inference was performed using Hamiltonian Monte Carlo sampling, with model fit assessed via posterior predictive checks.", "findings": "The model successfully quantified substantial heterogeneity in cost-effectiveness across different fleet types, with posterior distributions revealing that excavators and haul trucks were the most cost-effective asset classes. A key finding was that for haul trucks, the 95% credible interval for the cost-effectiveness ratio was [1.4, 2.1], indicating a significantly higher return relative to other assets. Operational downtime was identified as the most influential predictor of poor cost-effectiveness.", "conclusion": "The proposed Bayesian hierarchical model offers a statistically rigorous framework for cost-effectiveness analysis under data constraints, providing a superior alternative to deterministic evaluations. It effectively captures uncertainty and variability inherent in industrial machinery operations.", "recommendations": "Adoption of this probabilistic modelling approach is recommended for infrastructure agencies and private contractors to inform fleet procurement and maintenance strategies. Future work should integrate real-time sensor data to enhance model granularity and predictive capability