topic Re: OD Cost matrix gives wrong results in ArcGIS Network Analyst Questions

OD Cost matrix gives wrong results

JohnLowry — Thu, 13 Feb 2014 21:40:49 GMT

I'm using OD Cost Matrix with a very simple network dataset (no turns, etc. just defaults). I am trying to find the closest Health Facility (Destination) to many villages (origin). When I run the the OD Cost Matrix solver most of the Destination Rank = 1 lines appear correct based on a visual inspection of the road network. However, there are several where the closest Health Facility to a village is clearly wrong. I've checked for the connectivity of the road network and there don't appear to be any gaps in the next work.

I'm attaching a graphic. Notice that Health Facility 67 (Destination) is closest to village 1638 (origin) but doesn't come up with that (instead 1638 is linked (Destination Rank = 1) to a Health Facility that is a long way off (the graphic). Also notice that there does not appear to be any gap along the network (dangle errors shown). For Location Position (Load Locations dialogue) I am using a search tolerance of 1000 meters which is plenty large.

Any insights are appreciated.

[ATTACH=CONFIG]31421[/ATTACH]

Re: OD Cost matrix gives wrong results

RamB — Sun, 16 Feb 2014 04:20:42 GMT

First: you should not be using OD feature for your task. You must instead use 'closest facility' analysis. as shown here http://help.arcgis.com/en/arcgisdesktop/10.0/help../index.html#//00470000004n000000

Second: checking visually or connectivity rules is not enough. If you doubt a path, check for a simple shortest path by clicking on origin and destination and solve it for a route.

Third: results from OD are not shown with geometric alignment along the road network, it is shown as aerial distance but it contains real world travelled distance. So please do not conclude based only on graphic display of result especially with OD tool.

regards,

Re: OD Cost matrix gives wrong results

JohnLowry — Mon, 17 Feb 2014 02:06:21 GMT

Thank you very much for your helpful advice. I'm now using the right tool for my analysis. Also, I found that the reason the paths were not correct was because I thought were intersecting but instead overlapping. I fixed that with feature-to-line (new version of the old CLEAN). Thanks again for taking the time to post your advice.