Enhancing Association Rule Mining for Solving the Storage Location Assignment Problem

Abstract

In the era of online retailing, reducing the picking time of orders is of great importance. One way of driving the picking time down is to optimise the locations of articles within the warehouse, a problem which is referred to as the Storage Location Assignment Problem (SLAP). The SLAP is an NP-hard problem, and it is therefore desirable to find a relaxation of the problem. In this thesis a rule based approach to the problem is proposed, focusing on association rule mining and rules created from a neural network utilising distance metric learning. These rules are then used by a greedy and a genetic algorithm, to optimise the article placements. The data used to find rules and evaluate the algorithms come from an online retailer of electronic spare parts. When evaluating the genetic algorithm on this dataset, it performs worse than the baseline of storing the most frequently purchased articles closest to the picking depot. However, the greedy algorithm outperforms this baseline by up to 11%, showing that there is a lot of promise for this rule based approach.