| Warehouse activities play a key role in the final customer service level. From the warehouse processes, order picking is the major contributor to this category overall expenses. Order batching is commonly employed to improve the resources efficiency. Several heuristics have been proposed for the order batching problem, most of them developed for static batching, although scarce research has been focused on dynamic batching via stochastic modeling.;We present an a novel approach to the problem developing a framework based on machine learning application directly to historical order batches data; gaining valuable knowledge regarding how are the batches formed and what attributes are the most meaningful in this process. This knowledge is then translated into simple batching decision rules capable of batch orders in a real-time scenario (dynamically). The framework was compared to FCFS heuristics and single picking; the results indicate higher performance. |
