Classification of Users of a Health Service Provider Using Unsupervised Machine Learning Methods

In this paper, we compare three unsupervised classification methods: k-means, fuzzy clustering and Self-Organized Maps (SOM) on a database of a health service provider in Bogotá–Colombia in order to classify users who request services in different offices and to propose a reorganization of human resources of all offices according to the density of customers and their needs. To do so, the database is pre-processed to correct some data problems such as incomplete individuals, bad measurements and outliers to then apply the three selected clustering methods, compare their results and finally propose some recommendations for improving service levels and to reduce both total service and waiting times.