Similarity assessment between hierarchical and partitioned clustering: study with fuel prices

This study aimed to compare the similarity between clusters generated by hierarchical and partitioning methods (k-means) in the analysis of fuel prices, using data on regular gasoline and ethanol in the municipality of Campina Grande – PB, Brazil, in 2019. Cluster analysis, a multivariate statistical technique, was applied to classify fuel stations into homogeneous groups based on price proximity. The analyses were performed using R software, demonstrating the feasibility of applying these techniques to real-world data. In the non-hierarchical method (k-means), the number of groups was determined using the Elbow method, with significant differences identified in the mean prices between clusters. For the hierarchical method, we employed Euclidean distance and complete linkage, the resulting clusters exhibited structures similar to those obtained with k-means. The comparison between both approaches revealed consistency in group formation and cluster structural similarity, indicating that the methods produced convergent results for fuel station price segmentation, this demonstrates the reliability of the employed techniques. In conclusion, cluster analysis proves to be an effective tool for fuel market studies, both hierarchical and partitioning methods, while distinct in their approaches, generated coherent groupings in this context.