A Topological Data Analysis of Big Spatio-Temporal Urban Data
Lead Research Organisation:
University of Liverpool
Department Name: Geography and Planning
Abstract
One challenge of working with big spatio-temporal data
relates to the extraction of information about the underlying
structure from what can be dynamic and often highly dimensional data.
Topological Data Analysis has been used within a range of applications
to extract structure from data; in particular focusing on underlying
structure from within large and complex datasets. This project will
explore the applicability of this method within the context of urban
data, and with focus on extension around temporally referenced spatial
data.
relates to the extraction of information about the underlying
structure from what can be dynamic and often highly dimensional data.
Topological Data Analysis has been used within a range of applications
to extract structure from data; in particular focusing on underlying
structure from within large and complex datasets. This project will
explore the applicability of this method within the context of urban
data, and with focus on extension around temporally referenced spatial
data.
Studentship Projects
| Project Reference | Relationship | Related To | Start | End | Student Name |
|---|---|---|---|---|---|
| ES/R501062/1 | 30/09/2017 | 29/09/2021 | |||
| 1949131 | Studentship | ES/R501062/1 | 24/09/2017 | 23/09/2021 | Krasen Samardzhiev |
| Description | The initial focus of the PhD was to use techniques and methods from the new field of topological data analysis (TDA) to address urban challenges. The specific projects were agreed upon with the supervisory team and the industry partner. First, an extensive literature review of the applications of TDA methods in different fields was undertaken. As a result two of the most widely used TDA techniques were selected and updated for use in two studies. In order to access their performance they were compared with other traditional methods. These projects showed that the TDA techniques fared worse than traditional methods and used more computational resources. These outcomes suggest that there needs to be more work on TDA techniques focusing on accuracy and better run time, before they can be advantageously applied in order to answer practical geographical questions. After discussions with the supervisory team and the industry partner, the focus of the PhD moved more towards the problem of defining and delineating urban structures. The research is ongoing and there are currently two additional outcomes from it. First, it is possible to use sound sensors in order to identify different functional areas of a city. This was experimentally shown by using data from the Newcastle Urban Observatory. By using only the sound patterns captured by the sound sensors it was possible to differentiate between different areas in the city such as transportation hubs, shopping centers or central areas. Second, the development of larger scale urban units breaks down along established administrative boundries. This was demonstrated by showing that the existence and boundaries of megaregions, large clusters of interconnected cities, in the US is affected by state boundaries. The experiment showed that of the widely established 11 megaregions in the US, only one that crossed state boundaries was reflected in employment data . |
| Exploitation Route | There are three broad ways the findings so far can be used: First, alternatives to TDA methods are suggested as well as potential areas where TDA methods need improvement. Second, sound data can be used to be provide a real-time picture of the urban landscape, that is cheaper to acquire and maintain. This means that councils or businesses that are interested this can test sound sensors as alternatives to other data sources. For example, there could be research or case studies into how sound sensors could potentially replace footfall sensors, since they are cheaper. Additionally, more effort can be put into creating better sound sensors, since they can be used in this new way. Lastly, the megaregional findings have implications for policy makers and urban planners. They shhow that the focus of interventions aiming to establish and maximize the benefits of megaregional structures should be on areas that lie on and around defined administrative boundaries. |
| Sectors | Communities and Social Services/Policy,Government, Democracy and Justice,Retail |
| Description | CARTO |
| Organisation | CARTO |
| Country | Spain |
| Sector | Private |
| PI Contribution | I am undertaking a studentship in which they are the industrial partner. |
| Collaborator Contribution | They are involved in the discussion about the direction of the PhD and contribute with advice, as well as access to their online platform and to other data partners. |
| Impact | They have been involved with all of the outcomes specific in the findings section. |
| Start Year | 2017 |
| Description | TDA summer school session |
| Form Of Engagement Activity | Participation in an activity, workshop or similar |
| Part Of Official Scheme? | No |
| Geographic Reach | International |
| Primary Audience | Postgraduate students |
| Results and Impact | Together with Vitaliy Kurlin, I created a introductory Topological data analysis(TDA) course which was presented at the DS3 data science summer school at École polytechnique. Further, I was an assistant and involved with another TDA tutorial created by Pawel Dlotko, presented at the same summer school. The purpose of both was to provide the attendees with an overview and introduction to Topological data analysis methods. The course was a good way to network with people from academia and industry interested in TDA. The university invited us back for the 2020 summer school. |
| Year(s) Of Engagement Activity | 2018 |
| URL | http://www.ds3-datascience-polytechnique.fr |