Welcome to Project Global View.
We are: Sam Dixon, Umer Ijaz, Elizabeth Guest, Nick Jones and Graham Towl
The aim of this project was to explore possibilities for a form of dash-board that gives policy makers an integrated view of the state of the UK, both at the current time, and into the past (videos of a prototype we made can be seen at the foot of this page). If we are equipped with a better view of the UK, we can ensure that it is more resilient to shocks.
We might like to have policy makers who are aware of the implications of their actions on others and who understand how independent parties relate to each other. Since relationships change we might also hope our decision makers are up-to-date. Ideally they would be sensitive to subtle changes in the UK today and understand how changes could affect predictions of the future and relate to experiences in the past. But to really have a good perspective on the whole of the UK is next to impossible for any one individual. Our project investigated how novel tools from network science, statistical inference, text mining and signal processing can help those making policy decisions.
Our scoping project makes use of three types of raw materials: 1) a set of key, live, social, environmental and economic signals like the below:
2) live streams of phrases and keywords extracted from the internet, 3) a collection of interested parties (we have had meetings with the Treasury, the Office for Budget Responsiblity, the institute of public policy research (ippr) and the Department for Transport). A signal might be any time-varying quantity and relevant examples are oil prices, internet traffic or rainfall. The keyword information informs us about what people are searching for on the internet and which terms are suddenly popular.
We can ask how the signals relate to each other at the moment by using tools from statistical inference. We can, after a fashion, draw a network map where any pair of signal labels are linked by a line if the signals are related. Since we have records into the past, we can also find similar maps, or networks, for earlier periods.
These networks of how relevant signals connect could be useful for policy makers who are trying to understand how different features of the UK relate to each other (like how oil prices depend on how rainy it is). However, having a sequence of networks, or a set of unfolding signals, can be made more useful if they can be associated with known events. Alongside the quantitative signals that we record, we have also recorded keyword data which allows us to give a qualitative signature for each period of time. We are investigating how the keyword data can be coupled to the network data to allow elementary forecasting.
A video demonstrating our template version of the global view software. There's no audio.
We are: Sam Dixon, Umer Ijaz, Elizabeth Guest, Nick Jones and Graham Towl
The aim of this project was to explore possibilities for a form of dash-board that gives policy makers an integrated view of the state of the UK, both at the current time, and into the past (videos of a prototype we made can be seen at the foot of this page). If we are equipped with a better view of the UK, we can ensure that it is more resilient to shocks.
 |
| A screen-shot of our prototype dashboard. On the left are the signal data we use as input. Each signal indicates the behaviour of that variable through time (e.g. a stock price or wind speed). On the right is the inferred network of relationships between those signals. Each circle, node, represents one of the signals on the left and each link connecting a pair nodes indicates a statistical relationship between them. |
We might like to have policy makers who are aware of the implications of their actions on others and who understand how independent parties relate to each other. Since relationships change we might also hope our decision makers are up-to-date. Ideally they would be sensitive to subtle changes in the UK today and understand how changes could affect predictions of the future and relate to experiences in the past. But to really have a good perspective on the whole of the UK is next to impossible for any one individual. Our project investigated how novel tools from network science, statistical inference, text mining and signal processing can help those making policy decisions.
Our scoping project makes use of three types of raw materials: 1) a set of key, live, social, environmental and economic signals like the below:
 |
| Example time series (or signals) |
 |
| We examined Google searches for different key-words (examples on the left). We show how searches for Job centre varied through time and illustrate below this a crude method for identifying abrupt changes in search frequency. |
We can ask how the signals relate to each other at the moment by using tools from statistical inference. We can, after a fashion, draw a network map where any pair of signal labels are linked by a line if the signals are related. Since we have records into the past, we can also find similar maps, or networks, for earlier periods.
 |
| We can attempt to infer networks of relationships between each signal. We can also examine how those relationships vary through time. This is the task of inferring time-varying networks. |
These networks of how relevant signals connect could be useful for policy makers who are trying to understand how different features of the UK relate to each other (like how oil prices depend on how rainy it is). However, having a sequence of networks, or a set of unfolding signals, can be made more useful if they can be associated with known events. Alongside the quantitative signals that we record, we have also recorded keyword data which allows us to give a qualitative signature for each period of time. We are investigating how the keyword data can be coupled to the network data to allow elementary forecasting.
A video demonstrating our template version of the global view software. There's no audio.
A video demonstrating how our prototype dashboard can be used for inferring sequences of networks.
download sample data here
download prototype code here
download our scoping study report here
download some of our early motivational slides here
download a review of semantics and new media here
download overview notes on the global view pipeline here
Associated publications:
Proceedings of the Royal Society A, 467, 2088, (2011), Generalized methods and solvers
for noise removal from piecewise constant signals. I. Background theory [detecting steps in signals code here]
M.A. Little N.S. Jones
Proceedings of the Royal Society A, 467, 3115, (2011), Generalized methods and solvers
for noise removal from piecewise constant signals. II. New Methods [detecting steps in signals]
M.A. Little N.S. Jones
Phys. Rev. E 84, 026109 (2011) Temporal evolution of financial-market correlations
D.J. Fenn, M.A. Porter, S. Williams, M. McDonald, N.F. Johnson and N.S. Jones [Obtaining an overview of aspects of financial markets]
(Submitted - journal has forbidden preprint) Highly comparative time-series analysis:
the empirical structure of time series and their methods. [detecting changes in features of signals]
B.D. Fulcher, M.A. Little, N.S. Jones
Sumeet Agarwal, Gabriel Villar, and Nick S. Jones. High Throughput Network Analysis (extended abstract). In Machine Learning in Systems Biology (MLSB), Proceedings of the Fourth International Workshop, Edinburgh, Scotland, October 2010 [detecting changes in features of networks]
This project was funded by the EPSRC and the ESRC with grant number EP/I005986/1
