Sibren Isaacman

Human mobility characterization from cellular network data

Anonymous location data from cellular phone networks sheds light on how people move around on a large scale.

Human mobility modeling at metropolitan scales

Models of human mobility have broad applicability in fields such as mobile computing, urban planning, and ecology. This paper proposes and evaluates WHERE, a novel approach to modeling how large populations move within different metropolitan areas. WHERE takes as input spatial and temporal probability distributions drawn from empirical data, such as Call Detail Records (CDRs) from a cellular telephone network, and produces synthetic CDRs for a synthetic population. We have validated WHERE against billions of anonymous location samples for hundreds of thousands of phones in the New York and Los Angeles metropolitan areas. We found that WHERE offers significantly higher fidelity than other modeling approaches. For example, daily range of travel statistics fall within one mile of their true values, an improvement of more than 14 times over a Weighted Random Waypoint model. Our modeling techniques and synthetic CDRs can be applied to a wide range of problems while avoiding many of the privacy concerns surrounding real CDRs.

DP-WHERE: Differentially private modeling of human mobility

Models of human mobility have broad applicability in urban planning, ecology, epidemiology, and other fields. Starting with Call Detail Records (CDRs) from a cellular telephone network that have gone through a straightforward anonymization procedure, the prior WHERE modeling approach produces synthetic CDRs for a synthetic population. The accuracy of WHERE has been validated against billions of location samples for hundreds of thousands of cell phones in the New York and Los Angeles metropolitan areas. In this paper, we introduce DP-WHERE, which modifies WHERE by adding controlled noise to achieve differential privacy, a strict definition of privacy that makes no assumptions about the power or background knowledge of a potential adversary. We also present experiments showing that the accuracy of DP-WHERE remains close to that of WHERE and of real CDRs. With this work, we aim to enable the creation and possible release of synthetic models that capture the mobility patterns of real metropolitan populations while preserving privacy.

Ranges of human mobility in Los Angeles and New York

The advent of ubiquitous, mobile, personal devices creates an unprecedented opportunity to improve our understanding of human movement. In this work, we study human mobility in Los Angeles and New York by analyzing anonymous records of approximate locations of cell phones belonging to residents of those cities. We examine two data sets gathered six months apart, each representing hundreds of thousands of people, containing hundreds of millions of location events, and spanning two months of activity. We present, compare, and validate the daily range of travel for people in these populations. Our findings include that human mobility changes with the seasons: both Angelenos and New Yorkers travel less in the winter, with New Yorkers showing a greater decrease in mobility during the cold months. We also show that text messaging activity does not by itself accurately characterize daily range, whereas voice calling alone suffices. Finally, we show that our methodology is accurate by comparing our results to ground truth obtained from volunteers.

Distributed rating prediction in user generated content streams

Recommender systems predict user preferences based on a range of available information. For systems in which users generate streams of content (e.g., blogs, periodically-updated newsfeeds), users may rate the produced content that they read, and be given accurate predictions about future content they are most likely to prefer. We design a distributed mechanism for predicting user ratings that avoids the disclosure of information to a centralized authority or an untrusted third party: users disclose the rating they give to certain content only to the user that produced this content. We demonstrate how rating prediction in this context can be formulated as a matrix factorization problem. Using this intuition, we propose a distributed gradient descent algorithm for its solution that abides with the above restriction on how information is exchanged between users. We formally analyse the convergence properties of this algorithm, showing that it reduces a weighted root mean square error of the accuracy of predictions. Although our algorithm may be used many different ways, we evaluate it on the Neflix data set and prediction problem as a benchmark. In addition to the improved privacy properties that stem from its distributed nature, our algorithm is competitive with current centralized solutions. Finally, we demonstrate the algorithms fast convergence in practice by conducting an online experiment with a prototype user-generated content exchange system implemented as a Facebook application.

Potential for collaborative caching and prefetching in largely-disconnected villages

In a world becoming ever more reliant on the power of information, bringing data connectivity into developing regions is becoming an important way to lift these regions out of poverty by educating and informing the population. Although many of these regions are not likely to receive the infrastructure to support fully wired (or even wireless) networks, existing cellular and delay tolerant technologies allow limited connectivity. In this paper we show that usability of highly disconnected networks can be increased through collaborative caching and data prefetching techniques. We focus on decreasing the miss rate of pages fetched in both general web access as well as more specialized education applications. We evaluate our schemes by running trace-driven simulations of internet traces from Cambodia and logs from Princeton Universitys Blackboard courseware web servers. Our caching and prefetching strategies in these environments show improvements in miss rate of up to 90% over more traditional approaches.

Low-infrastructure methods to improve internet access for mobile users in emerging regions

As information technology supports more aspects of modern life, digital access has become an important tool for developing regions to lift themselves from poverty. Though broadband internet connectivity will not be universally available in the short-term, widely-employed mobile devices coupled with novel delay-tolerant networking do allow limited forms of connectivity. This paper explores the design space for internet access systems operating with constrained connectivity. Our starting point is C-LINK, a collaborative caching system that enhances the performance of interactive web access over DTN and cellular connectivity. We discuss our experiences and results from deploying C-LINK in Nicaragua, before moving on to a broader design study of other issues that further influence operation. We consider the impact of (i) storing web content collaboratively cached across all user nodes, (ii) hybrid transport layers exploiting the best attributes of limited cellular and DTN-style connectivity. We also explore the behavior of future systems under a range of usage and mobility scenarios. Even under adverse conditions, our techniques can improve average service latency for page requests by a factor of 2X. Our results point to the considerable power of leveraging user mobility and collaboration in providing very-low-infrastructure internet access to developing regions.

Modeling human migration patterns during drought conditions in La Guajira, Colombia

Modeling human mobility is key for a variety of applications such as migratory flows, epidemic modeling or traffic estimation. Recently, cell phone traces have been successfully used to model aggregated human mobility, in particular during natural disasters such as earthquakes or flooding. Climate-related environmental change brings a decline of productive agricultural land and livestock which will push rural residents to migrate. As a result, it also has the potential of causing changes in human mobility and cause migrations that have a wider and long standing impact. In this study, using anonymized and aggregated cell phone traces, we model the migrations that happened during a severe drought that happened in La Guajira, Colombia, in 2014. Our results indicate a linear reduction of the population of 10 percent during the 6 months considered for this study. Furthermore, predicting these migrations has about a 60% success rate for both the total number of people that migrate and to where they migrate. We also introduce a modification of the Radiation model in order to capture weather as one of the factors driving mobility, showing a RSS and RMSE reduction of 4.5% when compared with the standard models.

POSTER: Exploiting Approximations for Energy/Quality Tradeoffs in Service-Based Applications

Approximations and redundancies allow mobile and distributed applications to produce answers or outcomes of lesser quality at lower costs. This paper introduces RAPID, a new programming framework and methodology for service-based applications with approximations and redundancies. Finding the best service configuration under a given resource budget becomes a constrained, dual-weight graph optimization problem.

Distributed collaborative filtering over social networks

Recommender systems predict user preferences based on a range of available information. For systems in which users generate streams of content (e.g., blogs, periodically-updated newsfeeds), users may rate the produced content that they read, and be given accurate predictions about future content they are most likely to prefer. We design a distributed mechanism for predicting user ratings that avoids the disclosure of information to a centralized authority or an untrusted third party: users disclose the rating they give to certain content only to the user that produced this content. We demonstrate how rating prediction in this context can be formulated as a matrix factorization problem. Using this intuition, we propose a distributed gradient descent algorithm for its solution that abides with the above restriction on how information is exchanged between users. We formally analyse the convergence properties of this algorithm, showing that it reduces a weighted root mean square error of the accuracy of predictions. We also demonstrate how this technique can readily be used to offer optimal recommendation.