Consensus clustering for urban land use analysis using cell phone network data

Abstract

Pervasive large-scale infrastructures, such as cell phone networks, have the ability to capture individual digital footprints, and as a result, generate datasets that provide a new vision on human dynamics. In this context, cell phones and cell phone networks, due to its ubiquity, can be considered one of the main sensors of human behavior. The information collected by these networks can be used to understand the dynamics of urban environments with a detail not available up to now. One of the areas that can benefit from this information is urban planning. In this paper, we present a technique for the automatic identification of land uses from the information gathered by a cell-phone network. Our approach first computes the aggregated calling patterns of the antennas and, after that, finds clusters that identify how individuals use each geographic region. Given the inherent diversity of human activities in urban landscapes, we use consensus clustering to identify land uses, characterizing only those geographical areas with well defined behaviors. We present and validate our results using cell phone records and official land use data collected for Madrid. KeywordsUrban Mobility, Urban Dynamics, Call Detail Record, Land Use, Consensus Clustering. IntroductionWith the increasing capabilities of pervasive infrastructures, individuals leave behind footprints of their interaction with urban environments. As a result, new research areas, such as urban computing and smart cities, focus on improving the quality of life by understanding city dynamics through the data provided by ubiquitous technologies [12]. Of all these new data sources, cell phone traces are becoming increasingly important, due to the fact that they contain valuable information on a variety of aspects of human dynamics (i.e. mobility, social networks, behavioral patterns, etc.) for the best part of the population.A city is inherently a self-organized human-driven organization where individuals and their behavior play an important role in defining its dynamics. This implies that in order to efficiently model a city, individual information is necessary in order to reflect that activities are, at least in part, each individual's decision [1]. The datasets captured by ubiquitous computing technologies inherently reflect individual information relating to, among others, mobility and social dynamics. This fact represents a huge improvement when we consider that up to now data for urban analysis has been typically collected using surveys. Such approaches have a variety of limitations, mainly: (1) the increasing unwillingness of individuals to answer questionnaires and (2) the cost of the process, which limits the frequency with which information is captured. We consider that the information collected from cell-phone traces provides a novel source of information for urban analysis. The models obtained can be used to complement traditional solutions, as they overcome to a large extent the previous limitations.Examples of the applications of urban computin...

Comparing and modelling land use organization in cities

“…The information is exhaustive in terms of spatial and temporal resolution, allowing for the detection of concentrations of people second by second along days, weeks and months. Information from mobile phone call records [16][17][18][19][20][21][22][23][24][25][26][27][28][29], geolocated tweets [14,[30][31][32][33][34], credit card use [35] or FourSquare [21] has been considered in the literature. Different data sources have been compared, finding a consistent agreement among the estimations on human concentrations and mobility obtained from different information and communication technology data [26], as well as between information and communication technology data and more traditional techniques [20][21][22][23]26,28,36].…”

Disadvantaged Communities Have Lower Access to Urban Infrastructure

“…Consensus clustering is an unsupervised machine learning method that is widely used in the fields of molecular biology and complex networks (71)(72)(73). While it has recently found useful applications in the field of social network analysis (74, 75), there is not much evidence for its use in urban science (76,77). The method combines and evaluates results of several clustering algorithms to find the most robust set of labels for clusters (71) in the given data set.…”

Mobile Phone Data: A Survey of Techniques, Features, and Applications

“…Using a supervised learning model to classify land use has been attempted by [ 11 ] and [ 23 ], with a sample of land use initially labeled to train the classifier. Zinman and Lerner [ 23 ] applied a random forest (RF) algorithm to classify urban areas in Tel Aviv by extracting two types of features previously used across the literature [ 8 , 9 , 10 , 12 , 13 ] along with additional feature types representing communication habits to capture human communication behaviors, such as call duration, contact type (phone calls, accessing the internet), the weekly pattern features based on differences in communication activity between weekdays and weekends, and contact features such as the average number of days on which people engaged in or performed cellular contact in a given cell over one hour. Similarly, [ 11 ] applied a support vector machine (SVM) classifier algorithm, one of the most popular supervised learning algorithms, to classify urban land use types in Beijing.…”

“…Spatiotemporal mobility patterns and mobile phone activity patterns have also been widely extracted from mobile phone data for other practical applications, such as capturing individuals’ activities associated with urban zones, transportation, and the COVID-19 pandemic. The authors of [ 8 , 9 , 10 , 11 , 12 ], among others, explored or captured human activity patterns from mobile phone data to infer land use types based on spatiotemporal call volume patterns. This feature represents the total call volume (the number of incoming calls received, and outgoing calls made from all smartphones) managed by a given base transceiver station (BTS) within a given period of time.…”