One problem with the map application is that there are tons of public art in any given area, and sometimes there are more than one public art in the exact same place. One way to solve this is to cluster the public art to not only speed up transactions but to create more readability.
Figure 1 -
This image depicts points within a certain radius being connected. From here we calculate how many other points each one touches.
Figure 2 -
The points are sorted by how many connections they have, and then for every point that one connects to, we disable so that another point may not connect to it, and so that it cannot connect to other points. Using this method we can group the most amount of points into each cluster, minimizing the number of clusters. The above image is 'unweighted' such that the circle or cluster point is not at the center of mass.
Figure 3 -
This image depicts the clusters at center of mass.
2 comments:
good ideas to explore!
You guys might be interested in this Clustering paper that came out in Science a couple years back. We used it as the basis for clustering youth caseflows through the State of Florida Mental Health system.
http://www.psi.toronto.edu/affinitypropagation/FreyDueckScience07.pdf
Thanks, we'll definitely have to take a look at that. We didn't really need it to be too precise so I just kind of threw something together, it works pretty well for our application I think. It was made in processing too which I think you had mentioned using.
Post a Comment