Clusterização como técnica de apoio à decisão para um marketplace eletrônico logístico

Abstract

Road transport is one of the most impactful modes in the global transportation matrix. Despite its advantages of flexibility and availability, the sector is characterized by high fragmentation. In the past, the intermediation process in this logistics chain was inefficiently handled by freight agents in terms of time and cost. An effective solution to address the need for agility and ease is through electronic logistics marketplaces, which are systems allowing carriers to advertise their loads to truck drivers searching for freight. However, the ease of automating load and capacity matching has resulted in technology providers dealing with an unprecedented volume of data. Valuable insights about user behavior can be derived from this diverse dataset. Despite the popularity of logistics marketplaces, scientific literature has not kept pace with their growth. Given these opportunities, this study aims to identify patterns in a cargo advertisement database of a logistics marketplace using clustering, which can assist in decision-making. Following the CRISP-DM procedure, data on load postings from 2019 to 2021 were collected, and the clustering trend of the database was confirmed using RStudio software. The CLARA algorithm was subsequently employed, and the quality of clusters was assessed using the Silhouette index. The most representative group at the national level consisted of freight within the state of São Paulo, featuring full loads, covering distances of around 500 km, and requiring heavy-duty vehicles for transportation. In the context of São Paulo, the most significant partition comprised full freight journeys of just over five hours, also requiring heavy-duty vehicles. The higher frequency of full freight was attributed to its benefits, such as efficiency in space and resource utilization. The main strategies identified for the national context involve offering progressive discounts on additional services to carriers conducting a high volume of trips and targeted promotion of the logistics marketplace to potential customers interested in transporting heavy loads over medium distances. An interesting business opportunity was identified in Acre, where the company could expand its operations and provide support in a region of Brazil where road freight is less common. Additionally, encouraging the use of the platform in São Paulo for posting fractional loads was suggested, highlighting the advantages for owners of small vehicles. In conclusion, CLARA produced satisfactory results by reducing the computational complexity of a database with over three million entries, and the study revealed data clusters as potential opportunities for platform growth. However, there were instances of overlaps of structures clearly seen as distinct in the scatterplots.