Avaliação de desempenho de algoritmos de mineração de dados e simulação de Monte Carlo na descoberta de tendências no Hambre Delivery

Abstract

The food service sector streamlines transactions and contributes to the improvement of product and service quality, resulting in continuous growth and increased value of purchases made through marketplaces. With the growing adoption of food service digitalization, new information and insights can be derived from trend analyses based on databases generated from commercial transactions. To collect such data, the sector has implemented specialized applications that have proven effective for users seeking services through digital platforms. Furthermore, the integration of Artificial Intelligence (AI) with these applications has reshaped business operations, representing an emerging trend among companies offering products and services online. In this context, there is an increasing need to consider software-based solutions that leverage AI to systematize the analysis of trends within collected data. This study, therefore, proposes an investigation through simulations designed to evaluate computational performance by combining the Monte Carlo method with various data mining algorithms, aiming to identify the most suitable model to support decision-making in managing the food service sector via apps. To validate the effectiveness of the simulations, real-world data was collected from partner stores on the Hambre Delivery platform. The simulations assessed the FP-Growth, FP-Max, Apriori, and Eclat algorithms, taking into account scalability, execution time, and memory usage as performance metrics. The results indicate that the Eclat algorithm is more appropriate for small, low-complexity datasets, whereas FP-Growth and FP-Max scale well to large data volumes and demonstrate superior efficiency in both execution time and memory consumption. Additionally, the 27 association rules generated revealed significant trends, demonstrating that the application of the Monte Carlo method produces more accurate and reliable patterns.