39th RSAI:BIS, 2-4 September 2009, Limerick, Ireland

This presentation was given at the OAC User Group visualisation event at the RSS, 15th Sept 2009. For more details see - www.areaclassification.org.uk

Transcript

1. MOVING TOWARDS REAL-TIME GEODEMOGRAPHIC SEGMENTATION Dr Alex D Singleton Department of Geography and Centre for Advance Spatial Analysis , University College London www.alex-singleton.com
4. London Terraces Blue Collar Council Flat Central Districts
5. SEGMENTATIONS ARE CREATED BY CLUSTER ANALYSIS Area V1 V2 V3 V4 V5 V6 V7 V8 V9 V10 ... Area1 Area2 Area3 Area4 Area5 Area6 Area7 Area8 ...
7. Variable 1 Variable 2 Cluster 1 Cluster 2 Cluster 3 CLUSTER ANALYSIS
8. Area Cluster Area1 1 Area2 1 Area3 2 Area4 1 Area5 3 Area6 3 Area7 3 Area8 2 ... OUTPUT
9. NEED FOR REAL TIME GEODEMOGRAPHICS • Current classifications are created using static data sources. • Rate and scale of current population change is making large surveys (census) increasingly redundant. • Significant hidden value in transactional data • Data is increasingly available in near real time • Application specific (bespoke) classifications have demonstrated utility.
10. Specification Estimation Testing REALTIME GEODEMOGRAPHICS
11. REALTIME FEEDS OF DATA • Involve integration of large and possibly disparate databases • Common protocol • XML: E.g. UK Neighbourhood Statistics API • Formal • E.g. Doctor registrations; HE Data; Census data • Informal • Is there any value in other non- traditional data sources?
12. http://www.loopt.com/phones/iphone SOCIAL GPS
13. http://smalltalkapp.com
14. http://senseable.mit.edu/
15. STORE CARDS
16. What information can be extracted?
17. ONLINE SPECIFICATION INPUTS • Usability • Expert V Non-Expert Users • Non-Expert Users • Pre selection of variables and weighting for specific application • Expert Users • Selection of any variable and any weighting
18. CLUSTERING • K-Means algorithm: Unstable - initial start conditions effect the results • Measured within sum of squares or R2 K=4
19. CLUSTERING • Alternate algorithms? • PAM (Partitioning around medoids) tries to minimize the sum of distances of the objects to their cluster centers. • CLARA draws multiple samples of the dataset, applies PAM to each sample and returns the best result. • GA (Genetic Algorithm) is inspired by models of biological evolution. Produce results through a breeding procedure.
20. CLUSTERING K-means result for 41 “OAC variables” K-means result for 26 OAC Principle Components ~99% Similar K=4 or... refine k-means
21. VISUALISATION
22. VISUALISATION
24. STATE OF THE ART • Slowly moving beyond: • Idea of expert producers • General purpose classifications • There is only one correct representation • Creating classifications which are • Responsive to changes in local populations • Fit for purpose (bespoke classifications) • Open to scrutiny and verifiable by the public
25. SO WHAT DOES A FUTURE GEODEMOGRAPHIC LOOK LIKE?