Comparison of Dbscan and K-means clustering methods in the selection of representative clients for a vehicle routing model

This research presents the management, analysis and debugging of a database with more than 2.2 million records concerning the delivery of products by a 4PL logistics operator in Colombia, in charge of managing the supply chain of a telecommunications company. The objective of the article is to export the most representative clients of the series after comparing the clustering methods based on density (Dbscan) and centroid (Kmeans), which will be the input data for a vehicle routing model. Clustering methods are designed to reduce the size of spatial data sets of latitude and longitude, when exploring their taxonomy, parameters, and distance function in cluster generation, using Python as the programming language.