Clasificación de los municipios del país de acuerdo con el movimiento de carga por carretera utilizando modelos no supervisados de aprendizaje automático

Abstract

This document develops a no supervised clustering models applied to the colombia’s cargo flow public service, using clustering models as decision trees, k-means, k-medioids and DBSCAN. This algoritms are used with the data of the National Registry of Cargo Dispatches (RNDC) of the Colombian Ministry of Transportation. The process starts with recollecting the data, preparing the data, develop the algoritms and evaluate them to select the model that best fits in intra-group homogeneity and intergroup heterogeneity. According to the mentioned, k-medioids is the selected model, this model uses the median minimum distance between members of each group, generating three groups, each one with its own features.