Sunday, July 12, 2009

The Auto Correlation Engine

Hello welcome to my blog, this is the first post, and it's going to be about what I'd like to call the Auto Correlation Engine.

An idea came to me after viewing this article.

Say you had a bunch of data, and I'm not talking a couple of spreadsheets, I'm talking tens of millions of records, each holding attribute information... so much data you literally don't know what to do with - like perhaps all the information collected by governments around the world in their yearly census. It's too big to simply browse through to find out any useful information and there's too many geographic layers to add into a G.I.S to do any manual spatial analytics on it. But you know there's gold in the data somewhere. You know there must be some correlation between separate observations.

Enter the Auto Correlation Engine.

Imagine you had a system where by, for each geographic layer (State, Suburb, Region, Census District, etc, etc) you could attach a predefined observation (e.g, count, percentage, calculation) and derive all the possible spatial correlation indices amongst the observations, and report them to you.


Let's say your a government employee in charge of deciding what to do next about the high rates of child obesity in your district. Naturally as a G.I.S user you decide to add a new layer to your system displaying the count (and the lat/long points perhaps using proportional symbols) of obese children in your district. But what next? Do you add the fast food restaurants and perhaps do some concentric ring analysis? Do compare it with a layer displaying the number of game consoles bought in the area?

What if, you had a system that had already found out, for that layer, what other geographic layers and associated attributes have a high correlation index. As soon as you added the child obesity rates to your G.I.S platform, the Auto Correlation Engine would have predetermined that there is a correlation between high child obesity rates and the number of parks in the area, and informed you of the correlation. It would then ask you if you would like to add the correlated layer to your map. Of course it wouldn't be one of those annoying Microsoft paperclips, but it might be useful.

No comments:

Post a Comment