3.10 Spatial Indexing
4.1.2 Internet Maps
Shortly after this was released on the MapTube site, Google launched their own ‘Fusion Tables’ application, which is very similar in concept, but differs in one subtle detail. While Fusion Tables needs the data to be uploaded to a Google account, MapTube was designed from the outset to use data directly from the Internet. This feature is exploited later in section 4.4.1, where the concept of ‘mining’ an Internet data store for maps is introduced, resulting in a worked example using the UK’s 2011 Census in the data exploration of Chapter 7. ‘Data store mining’ is an appropriate name for this technique as, after collecting data on MapTube for 8 years, a valid question to ask is, “How do all the maps relate to one another?”. This question is answered using the Census 2011 data in Chapter 7. One final point to note about MapTube is the question of completeness. In order to vary the maps shown on its front page, MapTube mines the text in news stories published on a number of media websites. It uses this information, along with the text descriptions of its maps, to build a topicality ranking. The maps that show on the front page should then be related to current news items in the media, but the technique can also be used in reverse. Anything that ranks highly in the media, but does not have an associated map, suggests possible missing data. Network graphs of how maps are linked in both the data and meta-data domains give structure to the data, a concept that is explored in Chapter 5.
The final piece of infrastructure concerns real-time data. ‘Application Programming Interfaces’, or ‘APIs’, describe mechanisms by which user programs can access data or functionality provided by a third party. Through moves towards open Government, many sources of real-time data are now accessible to the general public. London Under- ground, for instance, makes information about the tube system freely available so that developers of mobile phone Apps will create utilities like ‘CityMapper’5 which helps
their customers. Another company, ‘TransportAPI’,6provides unified access to trans- port data in the UK, bringing together a number of real-time data feeds into a single interface. Before either of these were created, though, I created the ‘Adaptive Networks for complex Transport Systems’, or ANTS, project to provide real-time data to Map- Tube and subsequent systems like ‘citydashboard.org’. Along with a 2 dimensional visualisation as a map on MapTube, a 3D, browser based, visualisation was created. 5CityMapper is a mobile application which was developed to help people navigate around cities: http:// www.citymapper.com. 6TransportAPI was originally called ‘PLACR’, which was a spin-off from City University. Their website is: http:// www.
4.1. “A Place to Put Maps” 127 Figure 4.16 shows a real-time WebGL view of tube trains and buses in the Chrome web browser.
Figure 4.16: A real-time view of tube and bus positions using the Google Chrome web browser. The red cubes are buses and river services, while rectangles showing live tube train positions are shown in the regular TfL line colours.
The real-time data on tube and bus numbers, along with air quality, was used for the ‘iPad Wall’ project which was installed in the London Mayor’s office in 2012 [GMH13]. The server which collects the real-time data also archives it, which pro- vides the data for Chapter 6. Part of the motivation for writing this thesis is to use the 3GB a day of real-time data that has been collected since July 2012 to understand how the different city systems interact with one another. The incorporation of both static and real-time data is a key element. This forms the bulk of the data exploration in Chapter 8.
After creating real-time visualisations of tube positions and graphs of weather data and air quality, the realisation is that more advanced analytics are required to turn the information into knowledge. Taking the map in figure 4.17 as an example, this shows the mean wait times at stations on the London Underground network.
Figure 4.17: Mean wait times for London Underground stations plotted using QGIS.
By detecting where the real-time running data shows a wait time more than one standard deviation above the mean, unusual events can be detected. The approach is one of tightly coupling agent based modelling and data mining with real-time data and visual analytics. The solution presented in Chapter 6 is to learn from the archive data and ground the model using the derived statistics.
To conclude, the aim of this thesis is to further develop web mapping technology into a true web GIS, capable of analysing both static and real-time data. After 9 years developing the MapTube website to collect data and make maps, the goal is now to develop new ways to explore and analyse the data.