Mining the Thin Air—for Understanding of Urban Society

Abstract

We explore the potential of crowd-sourced information on human mobility and activities in an urban population drawn from a significant fraction of smartphones in the Los Angeles basin during February–May 2015. The raw dataset was collected by WeFi, a smartphone app provider. The dataset is noisy, irregular, and lean; however, it is large scale (over a billion events), cheap to collect, and arguably unbiased. We employ the state-of-the-art Big Data techniques to turn this structurally thin dataset into semantically rich insights on commuting, overworking, recreational traveling, shopping, and fast food consumption of the Greater LA population. For example, we reveal that Greater LA residents commute substantially longer than what is reported in the US census data. Also, we show that younger individuals dine at McDonald's significantly more than the older population does. Our results have implications for public health, inequality, urban traffic, and other research areas in social sciences. The large number of phones participating in our “crowd” makes it possible to obtain those results without the risk of compromising individual privacy.