Neural Network with Embedding Layer for Assistance of Selecting Boxes in Packing Task

Abstract

There is a shortage of skilled workers in packing process of the logistics site. In particular, selecting packing boxes needs high-skilled workers because this task is complex combinatorial problems owing to many kinds of items and packing boxes. Therefore, the purpose of this study is to assist workers in selecting appropriate packing boxes, regardless of the skill level. To achieve this purpose, we proposed the neural network model for selecting packing boxes. To adapt the neural network to multidimensional combinatorial problems, we used an embedding layer that is commonly used for natural language processing. For evaluation of the proposed model, we used on-site data to compare between the selection by the neural network and that by skilled workers. As a result, we achieved 70% accuracy in this comparison.