Leveraging explicitly disclosed location information to understand tourist dynamics: a case study

Abstract

Abstract. In recent years, the large deployment of mobile devices has led to a massive increase in the volume of records of where people have been and when they were there. The analysis of these spatio-temporal data can supply high-level human behavior information valuable to urban planners, local authorities, and designer of location-based services. In this paper, we describe our approach to collect and analyze the history of physical presence of tourists from the digital footprints they publicly disclose on the web. Our work takes place in the Province of Florence in Italy, where the insights on the visitors' flows and on the nationalities of the tourists who do not sleep in town has been limited to information from survey-based hotel and museums frequentation. In fact, most local authorities in the world must face this dearth of data on tourist dynamics. In this case study, we used a corpus of geographically referenced photos taken in the province by 4280 photographers over a period of 2 years. Based on the disclosure of the location of the photos, we design geovisualizations to reveal the tourist concentration and spatiotemporal flows. Our initial results provide insights on the density of tourists, the points of interests they visit as well as the most common trajectories they follow.

Towards (Re)Constructing Narratives from Georeferenced Photographs through Visual Analytics

“…Nonetheless, we opted for this classification since a recent study compared the country information on Flickr with immigration information collected at national borders and was able to demonstrate that the user-provided information was representative (Wood et al, 2013). Other researchers have based their analyses on a similar approach (Girardin et al, 2007;Girardin et al, 2008b). We favoured our approach specifically over time-based approaches that build on the idea that if somebody visits a given area only during a narrow timeframe in a longer period (e.g.…”

Learning Latent Factor Models of Travel Data for Travel Prediction and Analysis

“…Furthermore, tools for analysis and prediction of movement can be useful for numerous applications. For example, travel models can help in predicting the spread of disease [6,15]; surveying tourism [9,10], traffic [2,14], and special events mobility for urban planning [3]; geolocating with computer vision [4,16]; interpreting activity from movements [19,22]; and recommending travel [7,17]. Analysis of travel data can also play a role within the larger context of analysis of social network data.…”

“…This model captures the heavy-tailed nature of travel, but ignores the actual locations being visited. Second, one can build probability tables based on empirical travel histograms [4,6,9,10,15,16]. Such models are more accurate than simple stochastic process models, but require enormous datasets while revealing little about the underlying structure of the data.…”

“…Destinations that are equidistant from one's current location are equally likely, regardless of whether they are in New York City or in the Arctic. Second, one can build probability tables based on empirical travel histograms [4,6,9,10,15,16]. For example, Kalogerakis et al [16] use counts of actual transitions in a database of 6 million travel records.…”

Inference of mobility patterns via Spectral Graph Wavelets

“…Example works include [2,3] where the authors studied semi-supervised learning problems on graphs for classification. Second, a few recent works focused on mining Flickr data with application to inference of user mobility patterns based on empirical individual traces [4,5,6,7,8,9]. In addition, our approach is also analogue to the texture classification approach in [10] in the sense that both perform classification with features computed from wavelet transforms of signals.…”