Salvatore Scellato

A tale of many cities: universal patterns in human urban mobility.

The advent of geographic online social networks such as Foursquare, where users voluntarily signal their current location, opens the door to powerful studies on human movement. In particular the fine granularity of the location data, with GPS accuracy down to 10 meters, and the worldwide scale of Foursquare adoption are unprecedented. In this paper we study urban mobility patterns of people in several metropolitan cities around the globe by analyzing a large set of Foursquare users. Surprisingly, while there are variations in human movement in different cities, our analysis shows that those are predominantly due to different distributions of places across different urban environments. Moreover, a universal law for human mobility is identified, which isolates as a key component the rank-distance, factoring in the number of places between origin and destination, rather than pure physical distance, as considered in some previous works. Building on our findings, we also show how a rank-based movement model accurately captures real human movements in different cities.

NextPlace: a spatio-temporal prediction framework for pervasive systems

Accurate and fine-grained prediction of future user location and geographical profile has interesting and promising applications including targeted content service, advertisement dissemination for mobile users, and recreational social networking tools for smart-phones. Existing techniques based on linear and probabilistic models are not able to provide accurate prediction of the location patterns from a spatio-temporal perspective, especially for long-term estimation. More specifically, they are able to only forecast the next location of a user, but not his/her arrival time and residence time, i.e., the interval of time spent in that location. Moreover, these techniques are often based on prediction models that are not able to extend predictions further in the future. In this paper we present NextPlace, a novel approach to location prediction based on nonlinear time series analysis of the arrival and residence times of users in relevant places. NextPlace focuses on the predictability of single users when they visit their most important places, rather than on the transitions between different locations. We report about our evaluation using four different datasets and we compare our forecasting results to those obtained by means of the prediction techniques proposed in the literature. We show how we achieve higher performance compared to other predictors and also more stability over time, with an overall prediction precision of up to 90% and a performance increment of at least 50% with respect to the state of the art.

Structural Properties of Planar Graphs of Urban Street Patterns

Recent theoretical and empirical studies have focused on the structural properties of complex relational networks in social, biological, and technological systems. Here we study the basic properties of twenty 1-square-mile samples of street patterns of different world cities. Samples are turned into spatial valued graphs. In such graphs, the nodes are embedded in the two-dimensional plane and represent street intersections, the edges represent streets, and the edge values are equal to the street lengths. We evaluate the local properties of the graphs by measuring the meshedness coefficient and counting short cycles (of three, four, and five edges), and the global properties by measuring global efficiency and cost. We also consider, as extreme cases, minimal spanning trees (MST) and greedy triangulations (GT) induced by the same spatial distribution of nodes. The measures found in the real and the artificial networks are then compared. Surprisingly, cities of the same class, e.g., grid-iron or medieval, exhibit roughly similar properties. The correlation between a priori known classes and statistical properties is illustrated in a plot of relative efficiency vs cost.

Street Centrality and Densities of Retail and Services in Bologna, Italy:

This paper examines the relationship between street centrality and densities of commercial and service activities in the city of Bologna, northern Italy. Street centrality is calibrated in a multiple centrality assessment model composed of multiple measures such as closeness, betweenness, and straightness. Kernel density estimation is used to transform datasets of centrality and activities to one scale unit for analysis of correlation between them. Results indicate that retail and service activities in Bologna tend to concentrate in areas with better centralities. The distribution of these activities correlates highly with the global betweenness of the street network, and also, to a slightly lesser extent, with the global closeness. This confirms the hypothesis that street centrality plays a crucial role in shaping the formation of urban structure and land uses.

Small-world behavior in time-varying graphs

Connections in complex networks are inherently fluctuating over time and exhibit more dimensionality than analysis based on standard static graph measures can capture. Here, we introduce the concepts of temporal paths and distance in time-varying graphs. We define as temporal small world a time-varying graph in which the links are highly clustered in time, yet the nodes are at small average temporal distances. We explore the small-world behavior in synthetic time-varying networks of mobile agents and in real social and biological time-varying systems.

YouTube around the world: geographic popularity of videos

One of the most popular user activities on the Web is watching videos. Services like YouTube, Vimeo, and Hulu host and stream millions of videos, providing content that is on par with TV. While some of this content is popular all over the globe, some videos might be only watched in a confined, local region. In this work we study the relationship between popularity and locality of online YouTube videos. We investigate whether YouTube videos exhibit geographic locality of interest, with views arising from a confined spatial area rather than from a global one. Our analysis is done on a corpus of more than 20 millions YouTube videos, uploaded over one year from different regions. We find that about 50% of the videos have more than 70% of their views in a single region. By relating locality to viralness we show that social sharing generally widens the geographic reach of a video. If, however, a video cannot carry its social impulse over to other means of discovery, it gets stuck in a more confined geographic region. Finally, we analyze how the geographic properties of a videos views evolve on a daily basis during its lifetime, providing new insights on how the geographic reach of a video changes as its popularity peaks and then fades away. Our results demonstrate how, despite the global nature of the Web, online video consumption appears constrained by geographic locality of interest: this has a potential impact on a wide range of systems and applications, spanning from delivery networks to recommendation and discovery engines, providing new directions for future research.

SIFT Features Tracking for Video Stabilization

This paper presents a video stabilization algorithm based on the extraction and tracking of scale invariant feature transform features through video frames. Implementation of SIFT operator is analyzed and adapted to be used in a feature-based motion estimation algorithm. SIFT features are extracted from video frames and then their trajectory is evaluated to estimate interframe motion. A modified version of iterative least squares method is adopted to avoid estimation errors and features are tracked as they appear in nearby frames to improve video stability. Intentional camera motion is eventually filtered with adaptive motion vector integration. Results confirm the effectiveness of the method.

Evolution and sustainability of a wildlife monitoring sensor network

As sensor network technologies become more mature, they are increasingly being applied to a wide variety of applications, ranging from agricultural sensing to cattle, oceanic and volcanic monitoring. Significant efforts have been made in deploying and testing sensor networks resulting in unprecedented sensing capabilities. A key challenge has become how to make these emerging wireless sensor networks more sustainable and easier to maintain over increasingly prolonged deployments. In this paper, we report the findings from a one year deployment of an automated wildlife monitoring system for analyzing the social co-location patterns of European badgers (Meles meles) residing in a dense woodland environment. We describe the stages of its evolution cycle, from implementation, deployment and testing, to various iterations of software optimization, followed by hardware enhancements, which in turn triggered the need for further software optimization. We report preliminary descriptive analyses of a subset of the data collected, demonstrating the significant potential our system has to generate new insights into badger behavior. The main lessons learned were: the need to factor in the maintenance costs while designing the system; to look carefully at software and hardware interactions; the importance of a rapid initial prototype deployment (this was key to our success); and the need for continuous interaction with domain scientists which allows for unexpected optimizations.

Track globally, deliver locally: improving content delivery networks by tracking geographic social cascades

Providers such as YouTube offer easy access to multimedia content to millions, generating high bandwidth and storage demand on the Content Delivery Networks they rely upon. More and more, the diffusion of this content happens on online social networks such as Facebook and Twitter, where social cascades can be observed when users increasingly repost links they have received from others. In this paper we describe how geographic information extracted from social cascades can be exploited to improve caching of multimedia files in a Content Delivery Network. We take advantage of the fact that social cascades can propagate in a geographically limited area to discern whether an item is spreading locally or globally. This informs cache replacement policies, which utilize this information to ensure that content relevant to a cascade is kept close to the users who may be interested in it. We validate our approach by using a novel dataset which combines social interaction data with geographic information: we track social cascades of YouTube links over Twitter and build a proof-of-concept geographic model of a realistic distributed Content Delivery Network. Our performance evaluation shows that we are able to improve cache hits with respect to cache policies without geographic and social information.

Geo-spotting: mining online location-based services for optimal retail store placement

The problem of identifying the optimal location for a new retail store has been the focus of past research, especially in the field of land economy, due to its importance in the success of a business. Traditional approaches to the problem have factored in demographics, revenue and aggregated human flow statistics from nearby or remote areas. However, the acquisition of relevant data is usually expensive. With the growth of location-based social networks, fine grained data describing user mobility and popularity of places has recently become attainable. In this paper we study the predictive power of various machine learning features on the popularity of retail stores in the city through the use of a dataset collected from Foursquare in New York. The features we mine are based on two general signals: geographic, where features are formulated according to the types and density of nearby places, and user mobility, which includes transitions between venues or the incoming flow of mobile users from distant areas. Our evaluation suggests that the best performing features are common across the three different commercial chains considered in the analysis, although variations may exist too, as explained by heterogeneities in the way retail facilities attract users. We also show that performance improves significantly when combining multiple features in supervised learning algorithms, suggesting that the retail success of a business may depend on multiple factors.