Mineração de dados para comparação dos padrões de compras em uma rede de supermercados do oeste do Paraná antes e durante a pandemia de Covid-19

Abstract

In Brazil, in addition to the tragic consequences that the COVID-19 pandemic brought to the health area, other serious problems emerged, such as the increase in unemployment, successive increases in the prices of various products and the decrease in per capita household income. Some recent researches indicate that these problems provoked changes in the consumption profile of the families, however they are very specific researches and that they used in a very restricted way data mining techniques. In this sense, this work aimed to apply the association rule mining technique to: a) identify shopping patterns in a supermarket chain in the western region of Paraná during the previous period and during the COVID-19 pandemic; and b) compare the patterns obtained for the two periods, in order to enable the company's specialists to analyze whether the differences were unexpected and relevant from the point of view of the management of the chain's stores. As a result, the Apriori algorithm was applied using different support and confidence values for each of the analyzed municipalities in order to obtain the 30 best association rules in each of them. Through these results, it was possible to identify that during the period of the pandemic there was a change in the pattern of customer purchases in relation to the years 2018 and 2019, which for the specialists of the supermarket chain was already expected. However, according to the experts, there were important discoveries for the company with the completion of this work, such as changes in the specific purchasing patterns of each municipality during the years 2020 and 2021.