I'm running ArcMap 10.3.
I'm looking to analyse road collision data provided by a client. Each collision is recorded as a point, with a severity allocated to the accident. For example fatal accidents have a score of '3' whereas minor shunts and bumps have a score of 1. My client is interested in one kilometer stretches of roads which have a combined 'score' of more than 7 for the accidents therein.
In the past I've simply used a 'Polyline Dicer' tool to split a model of the road network into kilometer lengths. However this is a 'dumb' process, and it is possible that stretches of the road network that meet the above criteria will be missed.
I've considered cluster analysis but 1) there seems to be a slightly wooly element to how you set the parameter and 2) cluster analysis outputs won't flag up sections of the road network. For example I can identify the roads within a cluster hotspot, but I will also erroneously flag up side streets and other unrelated road lengths.
I'm aware of the 'SANET' toolset which would perform the sort of analysis I'm interested in. I can potentially commission another organization to carry out this work (I can't obtain a SANET license directly for my line of work), but I'm wondering if there is a deft combination of tools in ArcMap that would give me the same, or similar, results.
In my head there is a fuzzy workflow that involves step-wise buffer analysis of each accident to consider all other collisions within a 500 meter radius of the given collision, identifying those which have the same underlying road attribution (road name/type) and somehow working from there.