Geolocation data mining: flickr + Canary Islands

Abstract

• Social networks play an increasing role in the tourist market.
• This post shows an study which analyzes 150,00 photographs from Canary Islands on Flickr, a popular photo sharing site.
• There is a boom in geolocated data and services.

This project begun as a research about tags, conceived by Fernando Tricas (University of Zaragoza) and me. Fernando came with the idea to study the photos taken during Expo Zaragoza 2008, but I live in Canary Islands, so I preferred a closer region. Flickr is a photo sharing site with social network features, property of Yahoo. The study presented here analyzed 150,000 photographs and 4,000 users, obtained from the tag search "canaryislands" (and other languages as well) in Flickr, taken between 2004 and 2008. 

The first part of the study was presented in October 2009, and we analyzed photographers profiles and a basic study of tags added to images. This second part shows an interesting analysis of geolocated images to study tourist behavior and points of interests.

Geolocation

Flickr is able to store longitude and latitude data of the uploaded photographs. There is an increasing number of cameras with integrated GPS, specially mobile telephones. Also, users can manually locate in a map the place where the photo was taken. From a total of 150,000 images with the tags "canarias", "canaryislands", etc, 36,000 images had geolocation data.

The first map of the presentation (slide 18), shows the raw geolocation of the 36,000 images: a point, an image. Capitals and main touristic places stand out, along other points of photographic interest.

The second image (slide 19), shows the geolocation of the 36,000 images weighed by image popularity on Flickr (the point size is proportional to the number of views of the photographs). This gives an idea of the most photogenic places, according to image viewers. Many of our maps use the number of views' weighing.

In slide 20, geolocated photographs taken by foreign tourists are shown. During the first part of the study, we successfully geolocated ~25% of the photographers using their profile data. In this map, just the images taken by non-Spaniard users are represented. Every point is an image and its size represents the popularity (number of views). This slide gives an idea of which places are most visited by foreing tourists. For example, in Gran Canaria there aren't many photographs outside urban and touristic areas, with the exception of the center of the island. Compare this with the island of Tenerife, which shows many red coloured places besides Santa Cruz, Puerto de la Cruz y los Cristianos. But is the island of Lanzarote which stands out, both by the broadness of the points of interest and by image popularity.

Flickr assigns each place a unique identifier, identifiable using APIs de Yahoo. In slide 21, the 36,000 are grouped by places to show place popularity by number of taken photographs. Point size represents the number of photos in that place (not the number of views). Color represent type of place (town, airport, province...). Remember that maps generally show the whole set of 36,000 photographs, independently of the country of the photographers, except otherwise noted.

To display site popularity, slide 22 shows the aggregated number of views of each place, totaling the image views taken on each site. Capitals and main touristic areas are the most popular places in Flickr, but there are other zones of high visibility.

Photographs can be beautiful and/or popular, independently of the place it was taken. To know where photographers go, slide 23 shows the number of photographers who took images in each place. Point size weighs real visits, not image views. This slide displays all users, from Spain and abroad. Capitals stand out, but also touristic areas and other points of interests.

Dates

In the next four slides, from 24 to 28, photographs taken each year are shown, from 2004 to 2008. In the first part of the study, we showed a graph displaying the number of uploaded images to Flickr by date. This are the geolocated images. In this slides, 2004 shows few images compared to 2008. Each point is a photo and the circles weighs image popularity (number of views). This maps are helpful to identify photographer site preferences.

Tourists

Slide 29 displays official statistics of tourist arrivals to Canary Islands by source country between 2003 and 2007. Top sources are United Kingdom, Germany, Holland and Ireland. In the first part of this study, source country of successfully geolocated photographers were shown, and stats were not exactly the same of the official ones. Slide 30 shows source country of geolocated photographers which also have geolocated images. (Bear in mind that the origin of this set of 36,000 geolocated is a search by tags like "canarias", "canaryislands" and the like, and not a direct search by coordinates). Most geolocated images were taken by photographs from Spain, United Kingdom, Germany, Italy, Dominican Republic and Holland.

From slide 31 to 38, geolocated images are shown by source country: United Kingdom, Germany, Italy, Holland, Ireland, USA, Belgium and France. In this maps, every point is a photo and its size weighs its popularity (number of views). Clearly, images taken by foreign tourists are very popular in Flickr. Also there are behavioral differences by country. There are tourists from UK in Gomera, Tenerife, Gran Canaria, Fuerteventura and Lanzarote. In Gran Canaria and Tenerife, British stay at beaches and don't go outside touristic areas (rural tourism anyone?)... but they love to go all over Lanzarote. German photographers also take photos in rural sites of Gran Canaria and other islands.

From slide 39, the number of photographs are shown by source autonomous community. There are 17 autonomous communities in Spain. Top communities by photographers are Canary Islands followed by Valencia, Galicia, Madrid, Vasque Country and Catalonia. This numbers are a bit different to those of the whole set (that is, including un-geolocated images), which ranks first Canary Islands, Madrid, Valencia and Cataluña. In the maps, each point is a photo and its size weighs the number of views. There are maps whose points of interests are difficult to spot: this photos weren't too popular in Flickr.

And beyond

This method to study users has a great potencial, specially interpreted in a tourist context. However, bear in mind that the original set of images where found searching by tags ("canarias", "canary islands") and that analyzed users are part of a photograph sharing site (quite popular). But, this photographers could not necessary represent the average tourists -and let's also remember that a good chunk of photographs are taken by local residents. Anyway, and taking this into account, this kind of studies can be very helpful to the touristic market.

There remains other data we got but aren't showed here, like the geolocation by hour taken. We also did a preliminary analysis to relate tags and geolocation. Would be interesting to compare this results to using other tag searchs ("grancanaria", "tenerife"). Also to include geolocation searchs to the set. Now that 2009 is over, to update the data...

The increasing popularity of check-in applications, like Foursquare, Gowalla and recent movements by Twitter and Facebook in this market, proves there is a lot to be done. For many years, Google used its search technology to display ads based on content. The rise of social networks is shifting the paradigm to display ads using user profiles. Foursquare and similar mobile apps add another layer: user profile and location. Privacy issues aside, smart companies can use this data to understand better their customers.

Thanks for reading this. Please use the contact data if you want to know more.

Comments

[Skin Encyclopedia]

Brilliant study on Geolocation, really great.

[Puerto Rico Gran Canaria]

This is a pretty amazing idea, althgouh i don't know alot about geolocation. It seems the ideal way to pinpoint the most popular tourist destination o a specfic place. Gran Canaria tourist board will love this stuff!