Utilização de regras de associação para identificação de combos presenteáveis

Abstract

With the evolution of the technology, the volume of data has been growing significantly over the years, and along with that, the necessity to extract information from large datasets has become increasing. The theme of Knowledge Discovery in Databases has the objective to explore and extract relevant information until then unknown by specialists, converting such perceptions into messages that are easy to interpret. This process is composed of several techniques that demand great computational efforts that would not be possible without the advent of programming. The Association Rules method is one of these procedures, where it performs the role of finding relationships between different pieces of information, being widely used to improve the performance of sales in supermarkets. The study local of this work is a retail commercial establishment that works with thousands of items, where its difficulty lies in reducing low-flow products stored in stock. The objective of this work is to use the Association Rules technique to generate statistically attractive combos and exhaust items accumulated in stock. To execute this study, bibliographic surveys were realized, where the Fp-Growth algorithm was chosen to perform the processing of the information, given its effectiveness in executing large data sets. Of the thirteen products selected, promising associations were found for two of them. Altogether, seven association rules, for promotion and non-promotion periods, were found.