Using Web Data and Network Science to Detect Spatial Relationships

点击翻译按钮获取中文摘要

This PhD project aims to use recently developed Graph Science methodologies and techniques to drive forward the capabilities of modelling and interpreting spatial relationships. By viewing graphs as data and vice versa, it allows different statistical questions to be answered, for example when we care about dependencies. Dynamic network embedding methods are a research area with lots of recent growth that this project will utilise. This project will use a largely unexplored source of the data - the web - to identify spatial relationships, model and predict dependencies between cities and regions. By spatial relationships we mean how two places connect, e.g. transport links, trade links, etc. The project aims to capture such relationships in new ways, in various spatial and temporal scales, while contextualising these relationships to answer interesting social science questions. There is an opportunity here to apply new graphical methods to non trivial real life data, as well as creating new questions within the graph methodology that the project can explore and endeavour to answer. It uses new data-driven ideas to answer existing questions, while utilising widely rich and vast web data, that is ever growing and available. One example of data we will use is the JISC UK Web Domain Dataset, a novel cache of geolocated, archived websites. It presents an interesting and useful application for which new mathematical theory can be applied to. The project will use big data sources in using new approaches for data science. This is a multidisciplinary project that will tackle existing problems within Geography, whilst also being an opportunity for new ideas developing within Graph theory to be applied to real data; the project thus falls within the EPSRC Statistics and applied probability research area. The findings can be interpreted to better understand how relationships and dependencies between places behave, in order to make useful inferences and predictions. It will also exemplify the usefulness of the methodologies, while highlighting and conducting further research into interesting insights the analysis returns. There is an obvious gap in utilising such web data to create meaningful geographic knowledge about urban and regional interdependencies, to help better design things such as regional policies. The vision of this PhD project is to apply novel graph theory to release the untapped potential of web data to capture spatial relationships that otherwise we would not have been able to understand. The project will positively impact our understanding of the strengths of these new methods, as well as highlighting potential opportunities to expand and develop upon. It has the capability to give a new perspective on how geographers interpret and investigate spatial relationships. In using web data, it can positively impact how we are able to view data in such a way to give it a temporal and spatial aspect, that gives it new avenues in the ways it can be used.