Comments on: How Safe is Your Suburb http://mashupaustralia.org/mashups/how-safe-is-your-suburb/ An initiative of the Government 2.0 Taskforce Tue, 03 Aug 2010 05:39:35 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: Hugh http://mashupaustralia.org/mashups/how-safe-is-your-suburb/comment-page-1/#comment-406 Hugh Fri, 20 Nov 2009 09:31:21 +0000 http://mashupaustralia.org/?p=685#comment-406 If it makes anyone feel better about how much work they put in... The polygon display technology in GeoDemo and CrimeFinder was somewhere between 300 and 400 hours of tooling around 5-6 nights a week over a long time... (insert unhappy wife's comments here). Online interactive polygons are far frem easy with the current generation of APIs. That said - don't forget the point of a mashup is to cobble a good idea together so people can get the jist of what you're trying to acheive. If you end up with enough time to get some build quality into it then great! But that's the secondary priority of a mashup. If it makes anyone feel better about how much work they put in…

The polygon display technology in GeoDemo and CrimeFinder was somewhere between 300 and 400 hours of tooling around 5-6 nights a week over a long time… (insert unhappy wife’s comments here). Online interactive polygons are far frem easy with the current generation of APIs.

That said – don’t forget the point of a mashup is to cobble a good idea together so people can get the jist of what you’re trying to acheive. If you end up with enough time to get some build quality into it then great! But that’s the secondary priority of a mashup.

]]> By: Mel http://mashupaustralia.org/mashups/how-safe-is-your-suburb/comment-page-1/#comment-405 Mel Tue, 17 Nov 2009 20:35:52 +0000 http://mashupaustralia.org/?p=685#comment-405 It is great that this site has a lot of visualisation tools. Only critism is that not all crimes seem to be included in the "All" figure. Looking at the original NSW Crime Excel file shows in 2008 Albury had a total of 7248 crimes (not 7169), Armidale had 3000 (not 2192), etc. It is great that this site has a lot of visualisation tools. Only critism is that not all crimes seem to be included in the “All” figure. Looking at the original NSW Crime Excel file shows in 2008 Albury had a total of 7248 crimes (not 7169), Armidale had 3000 (not 2192), etc.

]]> By: Adam Kennedy http://mashupaustralia.org/mashups/how-safe-is-your-suburb/comment-page-1/#comment-404 Adam Kennedy Sat, 14 Nov 2009 13:50:26 +0000 http://mashupaustralia.org/?p=685#comment-404 This seems to be the comment from a lot of people. They knew where they wanted to be, but got caught by timing issues, leaving sites not quite living up to their potential. This seems to be the comment from a lot of people.

They knew where they wanted to be, but got caught by timing issues, leaving sites not quite living up to their potential.

]]> By: Jo Deeker http://mashupaustralia.org/mashups/how-safe-is-your-suburb/comment-page-1/#comment-403 Jo Deeker Sat, 14 Nov 2009 05:28:40 +0000 http://mashupaustralia.org/?p=685#comment-403 Hi, thanks so much for the feedback. I think the points you raise highlight some of the challenges we face when it comes time to actually *do* a mashup. We simply left it too late to get started. We spent approximately 30 working hours putting this application together and if we were to productionise it, we would spend more than that time again getting the data right. The massaging required to join the data was more than we anticipated and we didn't get finished. What the gov 2.0 community can learn from this is that while it is good to publish your data 'as is' for mashups, it needs to be joinable to other data and that usually means a fair bit of work. Our intention was to show that for a mashup of data to be more useful, it can be very valuable to do more than simply combine different data sets in a table or a map. Our different charts and maps provide different analytical views of the data, and help tell a more compelling story with the data. We chose to use a very generalised KML layer and keep it to one layer because it results in a very small file (250kb) which then is relatively fast to load (because we didn't want an app that times out or takes ages to load a view). We do have the option to use other GIS technology like ArcGIS Server, which provides excellent caching and tiling of maps and would offer superior speed. We could also do what the GeoDemo site has done and provide different layers at different resolutions. On the data front, we did want to combine the data with ABS population data. Our original intention was to firstly, show number of offences by population count (e.g. number of offences per 1,000 people) to remove the bias of more crime occurring in more populous suburbs. Then, we intended to overlay other ABS census data (for 2006) to see if there was a relationship between demographic factors such as employment status, income ranges, age ranges, education levels etc. Now that would be a <strong>really</strong> interesting mashup. Unfortunately, we did not get that database ready in time for the competition deadline. I have a draft database but it's not complete (we still need to cross-check our classifications of the offences and fix up a few coding errors), so we decided not to go ahead with it. all the very best, Jo Hi, thanks so much for the feedback. I think the points you raise highlight some of the challenges we face when it comes time to actually *do* a mashup. We simply left it too late to get started.
We spent approximately 30 working hours putting this application together and if we were to productionise it, we would spend more than that time again getting the data right. The massaging required to join the data was more than we anticipated and we didn’t get finished. What the gov 2.0 community can learn from this is that while it is good to publish your data ‘as is’ for mashups, it needs to be joinable to other data and that usually means a fair bit of work.

Our intention was to show that for a mashup of data to be more useful, it can be very valuable to do more than simply combine different data sets in a table or a map. Our different charts and maps provide different analytical views of the data, and help tell a more compelling story with the data.

We chose to use a very generalised KML layer and keep it to one layer because it results in a very small file (250kb) which then is relatively fast to load (because we didn’t want an app that times out or takes ages to load a view). We do have the option to use other GIS technology like ArcGIS Server, which provides excellent caching and tiling of maps and would offer superior speed. We could also do what the GeoDemo site has done and provide different layers at different resolutions.

On the data front, we did want to combine the data with ABS population data. Our original intention was to firstly, show number of offences by population count (e.g. number of offences per 1,000 people) to remove the bias of more crime occurring in more populous suburbs. Then, we intended to overlay other ABS census data (for 2006) to see if there was a relationship between demographic factors such as employment status, income ranges, age ranges, education levels etc. Now that would be a really interesting mashup. Unfortunately, we did not get that database ready in time for the competition deadline. I have a draft database but it’s not complete (we still need to cross-check our classifications of the offences and fix up a few coding errors), so we decided not to go ahead with it.

all the very best,

Jo

]]> By: Hugh http://mashupaustralia.org/mashups/how-safe-is-your-suburb/comment-page-1/#comment-402 Hugh Fri, 13 Nov 2009 10:05:21 +0000 http://mashupaustralia.org/?p=685#comment-402 The polygon thinning is partly a limitation of Javascript based mapping APIs being CPU hungry (in IE primarily) but it's mainly due to the fact that you need to have a layer of polygons for each map zoom level to try and make them look Ok. I'll blow my own trumpet and point out the CrimeFinder and GeoDemo Mashup Australia entries as one way to do it. It is a fair bit of work though. The GeoDemo site, for example, has 7 sets of boundaries, but there are actually about 30 layers of polygons to make it look better whilst not degrading the performance. The polygon thinning is partly a limitation of Javascript based mapping APIs being CPU hungry (in IE primarily) but it’s mainly due to the fact that you need to have a layer of polygons for each map zoom level to try and make them look Ok.

I’ll blow my own trumpet and point out the CrimeFinder and GeoDemo Mashup Australia entries as one way to do it. It is a fair bit of work though. The GeoDemo site, for example, has 7 sets of boundaries, but there are actually about 30 layers of polygons to make it look better whilst not degrading the performance.

]]> By: Adam Kennedy http://mashupaustralia.org/mashups/how-safe-is-your-suburb/comment-page-1/#comment-401 Adam Kennedy Fri, 13 Nov 2009 06:02:37 +0000 http://mashupaustralia.org/?p=685#comment-401 This is a fairly reasonable website, with decent coverage of the statistics. Two things jump out though... Firstly, the polyogon simplifification is really really high, looking at central sydney shows a bunch of nearly useless triangles. Secondly, the data shown doesn't really seem to control for population. In a lot of cases, the top crime region is "Sydney" (i.e. the city). It's not really a big surprise to anyone that areas like this are high, as they have large populations. Controlling for population would have made the data more interesting. In general though, a competant entry. This is a fairly reasonable website, with decent coverage of the statistics. Two things jump out though…

Firstly, the polyogon simplifification is really really high, looking at central sydney shows a bunch of nearly useless triangles.

Secondly, the data shown doesn’t really seem to control for population. In a lot of cases, the top crime region is “Sydney” (i.e. the city).

It’s not really a big surprise to anyone that areas like this are high, as they have large populations. Controlling for population would have made the data more interesting.

In general though, a competant entry.

]]>