Design of {Low} {Volume} {eCommerce} {Picker}-to-{Parts} {Fulfillment} {Sections} using {Model}-{Based} {Supervised} {Machine} {Learning}

Picker-to-parts e-Commerce fulfillment sections are still quite common for lowvolume picking activities. This paper presents a design method to size such sections with the view to estimating their performance to help in bid design. A machine learning algorithm is trained to understand the impact of design, planning, and operational parameters on total pick distance. Numerical experiments with different machine learning algorithms are illustrated. The Random Forest, Decision Tree, Gradient Boost, XG Boost, LightGMB, CatBoost and a tuned Artificial Neural Network show the best performance in terms of error and fit. SHAP analysis shows that the picklist size, layout dimensions, seasonality, and the slotting algorithm are the features of the experimental study in descending order of importance. While this result may be specific to the data parameters chosen, it is important to use SHAP analysis to understand machine learning output.