Table of Contents
Building Automated Geographical Analysis and Explanation Machines
Why?
AND...also because so few Geographers (and Social Scientists) are today trained in the essential computer and quantitative technologies necessary to perform useful applied analysis and modelling tasks
Indeed there is now a vicious circle setting in!
A New Quantitative Geography
These criteria are essentially the same as would be applied when developing Geographical Analysis Tools for GIS
What we REALLY-REALLY need are a set of GOOD general purpose Geographical Data Mining Tools
But
But
But
Currently we only have a few of the basic sub-components
In fact the main problem is ..
called ..
SPATIAL DATA
AND..
PPT Slide
PPT Slide
PPT Slide
+
More DATA
+
More DATA
and
More DATA
=
A bigger ...
NO.. just a BIGGER GIS database most of which will hardly ever be analyzed or used other than for drawing maps
Indeed
spatial data are being created and stored many times faster than they can be processed!
So the NEED is for user friendly Exploratory Spatial Data Analysis capable of being safely and easily used and understood by people who do not have higher degrees in the statistical or spatial sciences
Why Exploratory Spatial Data Analysis?
BUT
BUT
BUT
The Available Methods can be classified as follows ..
PPT Slide
Vendor Excuses Include:
but WHAT are our EXCUSES?
Its a CRIME that geographers are not taking their professional responsibilities more seriously
A basic Conjecture..
The problem is essentially that of creating a problem oriented Geographical Analysis Culture amongst End-Users
There are four additional issues:
The SECRET is to develop a GeoComputational Approach to the problem
Why AUTOMATION?
Unimaginably large Universes of Alternatives are common
In Geographical Analysis it is very important to appreciate that the aim is to do GEOGRAPHICAL ANALYSIS using statistical methods rather than doing STATISTICAL ANALYSIS on geographical data
The results should be visual and expressed as static or animated MAPS rather than as statistics that only a few understand
Also important to make the most of what computational intelligence and AI can deliver in the form of Smart Analysis Methods
Being SMART is not just a matter of methodology but also involves access, usability, relevancy, and result communication factors
The complete novice should be able to perform some sophisticated geographical analysis and get some useful and understandable results on the same day the work started
Design criteria for Smart GIS related Geographical Analysis Tools
Two Examples of Tools for doing this
Also a SMARTer version
Origins of the Technology
Mark 1 Geographical Analysis Machine (1987)
Geographical Analysis Machine (GAM) Mark 1 history
GAM/1: good aspects
BUT
GAM/1: Bads
GAM/K creates a density surface of weighted evidence of clustering which is used to suggest locations, intensities, and patterns of clustering that exists on the map
PPT Slide
and
BUT... does GAM/K work?
Results published in Alexander and Boyle (1996)
All the subsequent tools make use of the ideas first developed for the GAM/K
only the search methods differ
Putting the GAM on the Internet?
It was SURPRISINGLY easy to do!
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
PPT Slide
If you want to try out WWW-GAM http://www.ccg.leeds.ac.uk/smart/intro.html
The next problem was how to allow users to download onto their local machines the full GAM system as run at Leeds
Java GAM Run Times on Sun Ultra 30
GAM can be used to analyze virtually ANY spatial data for evidence of clustering
Applying GAM to the study of Long Term Limiting Illness data for 6,905 census areas in Northern England
Mapped Ward Level LLTI
Age-Sex Adjusted Clusters
Age Sex Clusters of Deficiency
An example of a Large Data Set: Burglary Data for Sheffield
PPT Slide
GAM can be applied to any data for which you have an EXPECTED and an OBSERVED value
Schools Performance
Bad Performing Schools
Good Performing Schools
GAM Animation
GAM/K is a descriptive tool
Well you PUT YOUR GEOGRAPHER’S head back on and start to relate the clusters to the underlying map patterns!
PPT Slide
PPT Slide
What you need is help in the form of a Geographical Explanations Machine that will trawl through lots of GIS coverages looking for localized geographical associations that might “explain” the GAM clusters or else find new ones!
?
So what is a Geographical Explanation Machine?
A version of GAM that looks at permutations of map overlays
Geographical Explanations Machine!
PPT Slide
PPT Slide
PPT Slide
GCEM too could be REVIVED but the concept needs a radical overhaul!
Why not build the search for local association into the circle search used in GAM?
GEM includes a GAM
GEM is computationally intensive
GEM can work with three types of geographical coverage
If you want to try out WWW-GEM http://www.ccg.leeds.ac.uk/smart/intro.html
Example of a GEM to “explain” the illness data
Pseudo GIS coverages created (nothing better was available) as follows..
GEM is computationally intensive
Clusters that cannot be “explained”
Clusters that can be “explained”
Explaining School Performance
Datasets used
Explaining Good Schools
Explaining Bad Schools
Good and Bad
Interpretation?
GAM and GEM Deficiencies
GAM and GEM goods
Future Plans
Further Info: Email
|