Análise de técnicas de aprendizagem de máquina para ranqueamento para produtos

Abstract

This work proposes the analysis of ranking learning techniques using as a basis the Shopping Queries Dataset provided by Amazon, in the KDD Cup 2022 competition. For this, an experiment was carried out with four ranking learning models that were chosen based on the articles written after the competition, and were evaluated based on. The experiments were evaluated using the cumulative gain with normalized discounting (NDCG), a multi-label classification metric used in machine learning problems. Finally, the main challenges for implementing the models in a real production environment are discussed, using the online sales company Americanas SA as a reference.