A compound-Poisson Bayesian approach for spare parts inventory forecasting

Spare parts are often associated with intermittent demand patterns that render their forecasting a challenging task. Forecasting of spare has been researched through both parametric and non-parametric approaches. However, little contributed in this area from Bayesian perspective, most such research is built around the Poisson distributional assumption. distribution known to have certain limitations and, further, empirical evidence on inventory performance methods lacking. In paper, we propose new method based compound distributions. The proposed compared Poisson-based Gamma prior as well frequentist one. A numerical investigation (on 7400 theoretically generated series) complemented by an assessment data about 3000 stock keeping units automotive sector analyse four methods. We find outperform other higher efficiency reported for prior. This outperformance increases variability. From practical some added complexity. also improves longer lead-times variability when