Gisruk retail identapril2010

Preview:

DESCRIPTION

Illustrates how function-form analysis can be used to automatically make explicit what is implicit in the map. Ambition is to create hierarchies of objects more 'in tune' with our perception of the world.

Citation preview

1April 2010

Automatic Classification of Retail Spaces from a Large Scale Topographc Database

William A Mackaness, Omair Z Chaudhry

School of GeoSciences, University of Edinburgh, william.mackaness@ed.ac.uk

Environmental and Geographical Sciences, Manchester Metropolitan University,

O.Chaudhry@mmu.ac.uk

2Mackaness & Chaudhry – GISRUK 2010

3Mackaness & Chaudhry – GISRUK 2010

A need to classify retail spaces

• To view distribution patterns of different types of retail space at a national scale.

• ‘essential as a means of understanding and analysing relationships in the work of retailing’ (Guy 1998, p255).

• form follows function• How might we characterise Retail Space• .. And thus reason about Retail Space• 3 methodologies

– Boolean, Fuzzy and Bayesian• Results• Conclusion

4Mackaness & Chaudhry – GISRUK 2010

Characterising Retail Space

Shopping Malls

Factory outlet

Regional centre

Retail spaces

High St Retail parks

5Mackaness & Chaudhry – GISRUK 2010

Retail Space

6Mackaness & Chaudhry – GISRUK 2010

Measures to discern different Retail Spaces

• A measure of urban centrality

• Accessibility: Bus Roads

7Mackaness & Chaudhry – GISRUK 2010

Identifying Retail Spaces

8Mackaness & Chaudhry – GISRUK 2010

The High Street

9Mackaness & Chaudhry – GISRUK 2010

Boolean Approach

• Shopping mall

• Retail park

10Mackaness & Chaudhry – GISRUK 2010

Boolean Approach

• Retail park

11Mackaness & Chaudhry – GISRUK 2010

Boolean Logic

12Mackaness & Chaudhry – GISRUK 2010

Fuzzy Logic

13Mackaness & Chaudhry – GISRUK 2010

Outline

• Broad ambition• Characterising Retail Space• Reasoning about Retail Space• Results• Conclusion

14Mackaness & Chaudhry – GISRUK 2010

Bayesian Approach

N

i

ic

h

ffK

hNcCfP

1||||

1)|(

• ‘For a given, unclassified retail space with a specific set of characteristics, what is the likelihood that it belongs to the population of ‘shopping malls’ with their specific set of characteristics?’.

15Mackaness & Chaudhry – GISRUK 2010

16Mackaness & Chaudhry – GISRUK 2010

Conclusion

• Sample size• Retail Spaces

– Need for classification national picture– Bayesian is best? (Fuzzy Logic less Black Box)

• Broader Issues– Green spaces, urban spaces– Automated derivation techniques (context of Open

Source Data)– Reasoning about space – making explicit what is

implicit

Recommended