2011/11/20 by Yaniv Altshuler, Altshuler, Yaniv, Nadav Aharony +7 · 1 citation
Computer Science · Physics and Astronomy · Social Sciences · #Complex Network Analysis Techniques #FOS: Computer and information sciences #Human Mobility and Location-Based Analysis #Impact of Technology on Adolescents #Mobile Crowdsensing and Crowdsourcing #Social and Information Networks (cs.SI)
paper · pdf · doi:10.48550/arxiv.1111.4645
openalex publication_date 2011/11/20 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Mobile phones are quickly becoming the primary source for social, behavioral,\nand environmental sensing and data collection. Today's smartphones are equipped\nwith increasingly more sensors and accessible data types that enable the\ncollection of literally dozens of signals related to the phone, its user, and\nits environment. A great deal of research effort in academia and industry is\nput into mining this raw data for higher level sense-making, such as\nunderstanding user context, inferring social networks, learning individual\nfeatures, predicting outcomes, and so on. In this work we investigate the\nproperties of learning and inference of real world data collected via mobile\nphones over time. In particular, we look at the dynamic learning process over\ntime, and how the ability to predict individual parameters and social links is\nincrementally enhanced with the accumulation of additional data. To do this, we\nuse the Friends and Family dataset, which contains rich data signals gathered\nfrom the smartphones of 140 adult members of a young-family residential\ncommunity for over a year, and is one of the most comprehensive mobile phone\ndatasets gathered in academia to date. We develop several models that predict\nsocial and individual properties from sensed mobile phone data, including\ndetection of life-partners, ethnicity, and whether a person is a student or\nnot. Then, for this set of diverse learning tasks, we investigate how the\nprediction accuracy evolves over time, as new data is collected. Finally, based\non gained insights, we propose a method for advance prediction of the maximal\nlearning accuracy possible for the learning task at hand, based on an initial\nset of measurements. This has practical implications, like informing the design\nof mobile data collection campaigns, or evaluating analysis strategies.\n