Well-separated: Points belong to a cluster when they are closer to every point in the cluster than they are to any point not in the cluster. This requires each cluster of points to be separated by a distance at least equal to the diameter of the largest cluster at its longest point.
Prototype-based: Points belong to a cluster when they are closer or more similar to the cluster’s prototype, ie. a point representing the cluster as a whole, than they are to any other cluster’s prototype. This tends to result in circular clusters since the prototype (in a geospatial dataset anyway) is usually the centrepoint of the cluster.
Graph-based: A cluster is defined as a set of points that are all connected, directly or via a chain of other points, to each other, and which are not connected to any point not in the cluster.
Density-based: A cluster exists where there is a region of high density surrounded by a region of low density. These clusters can end up being any shape, and vary widely in size.
Conceptual: A cluster exists where all points in the cluster share a common attribute which is not present in any point outside the cluster.
Implementation types
Variable gridding
Interleaved digits
K-Means
QT-Clust
DBSCAN
Agglomerative Hierarchical Clustering
Grid based viral growth argorithm
Questions
what about performance?
how do filters work?
how well does it integrate with current drupal mapping technologies?
how does caching work?
how does client interaction work?
how does it work with and without spatial database capabilities?
1. Select a center point for each of k clusters, where k is a small integer.
2. Assign each data point to the cluster whose center point is closest.
3. After all the data points are assigned, move each cluster's center point to the arithmetic mean of the coordinates of all the points in that cluster, treating each dimension separately.
4. Repeat from Step 2, until the center points stop moving.
Hierachical clustering
Method
1. Assign each point to its own cluster.
2. Calculate the distance from each cluster to every other cluster, either from their respective mean centers, or from the two nearest points from each cluster.
3. Take the two closest clusters and combine them into one cluster.
4. Repeat from Step 2, until you have the right number of clusters, or the clusters are some minimum distance apart from each other, or until you have one big cluster.
Naïve grid-based clustering
grid by display pixels / marker size
assign points to clusters within grid
order clusters by points
make superclusters (heavy clusters claim their neighbors)
[1:52pm] zzolo: there is still a pretty big hole when trying to map (and handle) lots of geospatial features in Drupal
[1:53pm] phayes: zzolo - I think the answer there is a combination of ViewsGeoJSON / server-side clustering
[1:58pm] dasjo: final question before i have to leave: do we need a server-side clustering like nod_ and me discussed? http://drupal.org/node/1547610
[1:59pm] nod_: dasjo: phayes seems to agree, see backscroll
[1:59pm] Brandonian: dasjo: Definitlely think it'd be a cool feature, not sure what other work is being done with that in php/drupal land though
[2:00pm] zzolo: dasjo: my initial thought is no, if we can focus on toolking up thinks like postgis and tilestache. but since that is not really going to happen overnight, a good stop gap solon could be server side clustering
[2:00pm] phayes: dasjo: Yes we need it! That's going to be critical as we scale to support larger data sets
[2:00pm] dasjo: yep Brandonian zzolo. basically some parts of the database implementation and client-side handling should be pluggable
[2:01pm] dasjo: but maybe we can come with some basic out-of-the-box implementations and leave it open to a database to optimize the clustering
tilestash + drupal + leaflet
log
[1:59pm] tnightingale: yeah - tilestache is (using mapnik) is able to generate tiles on demand from a postgis backend
[1:59pm] mackh: of data from the oil and gas commksion - like pipeline right-of-ways, access roads, wellsites, watercrossings
[2:00pm] tnightingale: zzolo: so we can sync drupal data into postgis and render that data into tiles using mapnik+tilestache
[2:00pm] zzolo: tnightingale: and mackh just to confirm, the postgi data is drupal data? (from sync_postgis, i assume)
[2:00pm] zzolo: sweet
[2:00pm] mackh: yes and no?
[2:00pm] tnightingale: yep
[2:01pm] mackh: stored in drupal, synced to postgis
[2:01pm] tnightingale: but doesn't have to be
[2:01pm] zzolo: awesome. thats what i was hoping. this is the stack i have been envisioning, but i don't get paid to do any of this stuff so i don't have time.
[2:02pm] Brandonian: for the more python/mapnik literate in the room: How feasible is it to do mapnik renders based directly off drupal field data if the database is Postgres/postgis and stored properly?
[2:02pm] tnightingale: Brandonian: that's essentially what we're doing with sync_postgis & tilestache
[2:03pm] zzolo: Brandonian: what i image is getting your design in tile mill, export out the mapnik xml file, then it should be pretty easy
[2:03pm] tnightingale: using a mapnik xml file produced in tilemill as our style guide
[2:06pm] tnightingale: zzolo: we also have a rough cut at a tilestache config management module on g.h
[2:06pm] phayes: ooo! link tnightingale?
[2:07pm] phayes: Oh tnightingale, what's the status of D7 Spatial-tools?
[2:07pm] zzolo: tnightingale: you have a link. i am actually gonna try to get this stack running locally this week as part of my presentation at the TC Drupal camp
[2:08pm] mackh: i also want to extend search api to return WKT onto maps
[2:08pm] mackh: from faceted results, hoping to get that client funded in the next 3 months
Office hours finishing
log
[2:07pm] tom_o_t: Brandonian: perhaps schedule dedicated Q&A time for stuff general questions that aren't related to module development?
[2:08pm] Brandonian: tom_o_t: Not a bad idea.
[2:09pm] jeffschuler: tom_o_t++ … this has been super useful, but It's been more of a facilitated discussion on high-level geo in drupal topics rather than a time for folks to come get help or figure out how they can pitch in… different from core office hours http://drupal.org/node/1242856