1 - Denis Toyoshima Escola de Engenharia de São Carlos - EESC - EESC
2 - FERNANDO FREIRE VASCONCELOS Faculdade de Economia, Administração e Contabilidade da Universidade de São Paulo - FEA - Administração
Reumo
Intralogistics plays an important role in the supply chain and through the automation of operations and optimization of processes it is possible to increase the efficiency of the internal movements of a warehouse.
The development of intralogistics projects aims to size the warehouse and its departments through the analysis of operational data for subsequent selection of equipment and level of automation. One of the main definitions of the project refers to the choice of storage that includes decisions about the grouping of SKUs and assigning these groups in the various departments.
In this context, the general objective of the work was to investigate different unsupervised machine learning techniques to identify which ones can assist in the problem of grouping SKUs and select the best group for allocation in an automatic storage system. The problem of grouping SKUs was examined through the application of conventional techniques based on univariate analysis such as ABC, COI and XYZ.
The research methodology was quantitative, exploratory and descriptive and the data were obtained from the logistics operation of a company in the retail sector. Unsupervised machine learning techniques such as PCA, Clustering and Autoencoder were also applied. These techniques proved to be more efficient in the grouping of SKUs since they consider all the variables of the base such as demand, demand variation and dimensional characteristics of the products.
As a result, it was found that the factor analysis of principal components, the analysis of clusters and the neural network autoencoder were efficient in performing the grouping of SKUs. To select the best groups, the Gaussian AHP method was used, which proved effective in identifying the most suitable groups for automation.
In the selection of the best groups, the Gaussian AHP method was used, which proved effective in identifying the most appropriate clusters for the proposed automation system. The present research proved the potential application of unsupervised machine learning techniques in the development of intralogistics automation projects and initiated some studies that can be deepened in future research.