Limits...
Modelling human mobility patterns using photographic data shared online.

Barchiesi D, Preis T, Bishop S, Moat HS - R Soc Open Sci (2015)

Bottom Line: The way we move around our environment has consequences for a wide range of problems, including the design of efficient transportation systems and the planning of urban areas.Here, we gather data about the position in space and time of about 16 000 individuals who uploaded geo-tagged images from locations within the UK to the Flickr photo-sharing website.Our findings are in general agreement with official figures in the UK and on travel flows between pairs of major cities, suggesting that online data sources may be used to quantify and model large-scale human mobility patterns.

View Article: PubMed Central - PubMed

Affiliation: Department of Mathematics , University College London, Gower Street , London WC1E 6BT, UK.

ABSTRACT
Humans are inherently mobile creatures. The way we move around our environment has consequences for a wide range of problems, including the design of efficient transportation systems and the planning of urban areas. Here, we gather data about the position in space and time of about 16 000 individuals who uploaded geo-tagged images from locations within the UK to the Flickr photo-sharing website. Inspired by the theory of Lévy flights, which has previously been used to describe the statistical properties of human mobility, we design a machine learning algorithm to infer the probability of finding people in geographical locations and the probability of movement between pairs of locations. Our findings are in general agreement with official figures in the UK and on travel flows between pairs of major cities, suggesting that online data sources may be used to quantify and model large-scale human mobility patterns.

No MeSH data available.


Related in: MedlinePlus

Evaluation of the aggregate model of mobility. (a) Identification of the 20 most populous UK cities. The plot depicts a tradeoff between precision (the number of correct identifications divided by the total number of cities identified) and recall (the number of correct identifications divided by 20) when a maximum filter is applied to the function p(x). Different lines correspond to different radii of the maximum filter, and values along the lines are obtained by varying the threshold ϕ between 10−3 and 10−4. The optimal parameters for the maximum filter corresponding to an F-measure of 0.63 are r≈60 km and ϕ=2.78×10−3. (b) The number of journeys per city pair as estimated from the NTS data is consistent with a lognormal distribution, with most pairs corresponding to less than 100 observed journeys, and a few pairs with at most about 400 journeys. We generated random variables according to this lognormal distribution, and computed their distance to the NTS estimates. We find that 92 out of 100 trials, the distance between the Flickr and NTS estimates is smaller than the distance between the randomly generated estimates and the NTS estimates.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4555850&req=5

RSOS150046F3: Evaluation of the aggregate model of mobility. (a) Identification of the 20 most populous UK cities. The plot depicts a tradeoff between precision (the number of correct identifications divided by the total number of cities identified) and recall (the number of correct identifications divided by 20) when a maximum filter is applied to the function p(x). Different lines correspond to different radii of the maximum filter, and values along the lines are obtained by varying the threshold ϕ between 10−3 and 10−4. The optimal parameters for the maximum filter corresponding to an F-measure of 0.63 are r≈60 km and ϕ=2.78×10−3. (b) The number of journeys per city pair as estimated from the NTS data is consistent with a lognormal distribution, with most pairs corresponding to less than 100 observed journeys, and a few pairs with at most about 400 journeys. We generated random variables according to this lognormal distribution, and computed their distance to the NTS estimates. We find that 92 out of 100 trials, the distance between the Flickr and NTS estimates is smaller than the distance between the randomly generated estimates and the NTS estimates.

Mentions: We obtained a list of the 20 largest UK cities by number of resident population along with their geographical coordinates from Wikipedia (see the electronic supplementary material) to assess quantitatively whether the local maxima in figure 2a correspond to areas of large population. By varying the dimension of the maximum filter window and the threshold level, a different number of local maxima can be identified, hence determining a trade-off between precision (the number of correctly identified cities divided by the total number of maxima identified) and recall (the number of correctly identified cities divided by 20) of the cities' identification. For every local maximum computed on the function p(x), the point was judged to identify one of the cities in the list if it was located at a distance smaller than 15 km from the centre of the corresponding city, as computed by comparing the coordinates obtained from Wikipedia and the coordinates of the local maxima. Figure 3a depicts the tradeoff between precision and recall obtained by varying the sizes (dx,dy) between (18 km,28 km) and (90 km,140 km) expressed in terms of the radius , and the threshold ϕ between 10−3 and 10−4. The F-measure is defined as the harmonic mean between precision and recall, and is a measure of the overall accuracy of the cities' identification. The maximum F-measure obtained was 0.63, corresponding to a size of (54 km,84 km) and a threshold ϕ=2.78×10−3. Figure 2a indicates the cities identified with these parameters, highlighting the ones that appear on the list of 20 most populous UK cities.Figure 3.


Modelling human mobility patterns using photographic data shared online.

Barchiesi D, Preis T, Bishop S, Moat HS - R Soc Open Sci (2015)

Evaluation of the aggregate model of mobility. (a) Identification of the 20 most populous UK cities. The plot depicts a tradeoff between precision (the number of correct identifications divided by the total number of cities identified) and recall (the number of correct identifications divided by 20) when a maximum filter is applied to the function p(x). Different lines correspond to different radii of the maximum filter, and values along the lines are obtained by varying the threshold ϕ between 10−3 and 10−4. The optimal parameters for the maximum filter corresponding to an F-measure of 0.63 are r≈60 km and ϕ=2.78×10−3. (b) The number of journeys per city pair as estimated from the NTS data is consistent with a lognormal distribution, with most pairs corresponding to less than 100 observed journeys, and a few pairs with at most about 400 journeys. We generated random variables according to this lognormal distribution, and computed their distance to the NTS estimates. We find that 92 out of 100 trials, the distance between the Flickr and NTS estimates is smaller than the distance between the randomly generated estimates and the NTS estimates.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4555850&req=5

RSOS150046F3: Evaluation of the aggregate model of mobility. (a) Identification of the 20 most populous UK cities. The plot depicts a tradeoff between precision (the number of correct identifications divided by the total number of cities identified) and recall (the number of correct identifications divided by 20) when a maximum filter is applied to the function p(x). Different lines correspond to different radii of the maximum filter, and values along the lines are obtained by varying the threshold ϕ between 10−3 and 10−4. The optimal parameters for the maximum filter corresponding to an F-measure of 0.63 are r≈60 km and ϕ=2.78×10−3. (b) The number of journeys per city pair as estimated from the NTS data is consistent with a lognormal distribution, with most pairs corresponding to less than 100 observed journeys, and a few pairs with at most about 400 journeys. We generated random variables according to this lognormal distribution, and computed their distance to the NTS estimates. We find that 92 out of 100 trials, the distance between the Flickr and NTS estimates is smaller than the distance between the randomly generated estimates and the NTS estimates.
Mentions: We obtained a list of the 20 largest UK cities by number of resident population along with their geographical coordinates from Wikipedia (see the electronic supplementary material) to assess quantitatively whether the local maxima in figure 2a correspond to areas of large population. By varying the dimension of the maximum filter window and the threshold level, a different number of local maxima can be identified, hence determining a trade-off between precision (the number of correctly identified cities divided by the total number of maxima identified) and recall (the number of correctly identified cities divided by 20) of the cities' identification. For every local maximum computed on the function p(x), the point was judged to identify one of the cities in the list if it was located at a distance smaller than 15 km from the centre of the corresponding city, as computed by comparing the coordinates obtained from Wikipedia and the coordinates of the local maxima. Figure 3a depicts the tradeoff between precision and recall obtained by varying the sizes (dx,dy) between (18 km,28 km) and (90 km,140 km) expressed in terms of the radius , and the threshold ϕ between 10−3 and 10−4. The F-measure is defined as the harmonic mean between precision and recall, and is a measure of the overall accuracy of the cities' identification. The maximum F-measure obtained was 0.63, corresponding to a size of (54 km,84 km) and a threshold ϕ=2.78×10−3. Figure 2a indicates the cities identified with these parameters, highlighting the ones that appear on the list of 20 most populous UK cities.Figure 3.

Bottom Line: The way we move around our environment has consequences for a wide range of problems, including the design of efficient transportation systems and the planning of urban areas.Here, we gather data about the position in space and time of about 16 000 individuals who uploaded geo-tagged images from locations within the UK to the Flickr photo-sharing website.Our findings are in general agreement with official figures in the UK and on travel flows between pairs of major cities, suggesting that online data sources may be used to quantify and model large-scale human mobility patterns.

View Article: PubMed Central - PubMed

Affiliation: Department of Mathematics , University College London, Gower Street , London WC1E 6BT, UK.

ABSTRACT
Humans are inherently mobile creatures. The way we move around our environment has consequences for a wide range of problems, including the design of efficient transportation systems and the planning of urban areas. Here, we gather data about the position in space and time of about 16 000 individuals who uploaded geo-tagged images from locations within the UK to the Flickr photo-sharing website. Inspired by the theory of Lévy flights, which has previously been used to describe the statistical properties of human mobility, we design a machine learning algorithm to infer the probability of finding people in geographical locations and the probability of movement between pairs of locations. Our findings are in general agreement with official figures in the UK and on travel flows between pairs of major cities, suggesting that online data sources may be used to quantify and model large-scale human mobility patterns.

No MeSH data available.


Related in: MedlinePlus