Dimensionamento de estoques na cadeia de suprimentos a partir da clusterização

Abstract

The development of models using data available to companies is a necessity in order to remain competitive in the global market. This work addresses the sizing and positioning of inventories in the supply chain, using the clustering of products by their categorical characteristics and sales history as a basis for projecting the allocation of new collections. Experts were consulted to validate the available variables, assigned to products related to past collections and future collections. Products from past collections were then clustered using the k-modes algorithm and their performance was measured for each preferred origin according to the destination of the order, as indicated by the logistics partner. Using the same criteria, the products that compose the future collection were associated with the clusters created, and based on the measured past performance, their volumes were projected for each warehouse. The study compares the results of three different scenarios, executed, potential and model. The research achieved its objectives by creating a product clustering model for future collections, identifying relevant variables in the process and assessing the model's adherence to inventory sizing. The proposed model defines criteria for allocating products and suggests reducing operating costs and improving the logistics service, allowing hidden factors to be identified, such as a possible increase in the level of service and customer satisfaction.