A significant and very extended approach for Smart Cities is the use of sensors and the analysis of the data generated for the interpretation of phenomena. The proper sensor location represents a key factor for suitable data collection, especially for big data. There are different methodologies to select the places to install sensors. Such methodologies range from a simple grid of the area to the use of complex statistical models to provide their optimal number and distribution, or even the use of a random function within a set of defined positions. We propose the use of the same data generated by the sensor to locate or relocate them in real-time, through what we denominate as a ‘hot-zone’, a perimeter with significant data related to the observed phenomenon. In this paper, we present a process with four phases to calculate the best georeferenced locations for sensors and their visualization on a map. The process was applied to the Guadalajara Metropolitan Zone in Mexico where, during the last twenty years, air quality has been monitored through sensors in ten different locations. As a result, two algorithms were developed. The first one classifies data inputs in order to generate a matrix with frequencies that works along with a matrix of territorial adjacencies. The second algorithm uses training data with machine learning techniques, both running in parallel modes, in order to diagnose the installation of new sensors within the detected hot-zones.