Hi all, Used to do Data Science / Analytics in a previous role but haven’t done much (any) over the past 3 years. At work, I’d offload this problem to our Data Science team, but they don’t have the bandwidth and I have a little free time, so I thought this would be a good opportunity to refresh my …

(This is to minimize the chance of any group overlapping.) What exactly do you mean by overlapping?

Good question. So additional context here is this is a marketing problem. The lat/long points are just representations of people we can target digitally but the platform we use doesn’t necessarily target people that “live” by one of the lat/long points but rather “live by or have recently visited” a…

Thanks, this did catch my eye when I was searching around, but I’m still not sure the best way to select a subset of the clusters / groups that would maximize the total distance between clusters. I could brute force it with a distance matrix but kind of thought there might be an easier or more effic…

[image] Snowy: but I’m still not sure the best way to select a subset of the clusters / groups that would maximize the total distance between clusters given what you wrote in your first post [image] Snowy: should be selected that maximizes the distance between the center of each group. (…

Maybe you could utilize k-medoids? Step 1 would be to find the optimal centers (where the distance between the centers is greatest). Step 2 would then be to use those centers as pre-defined medoids and then perform the clustering around those medoids to minimize the total within-cluster distances. Y…

Hi all, Thanks for your replies thus far. After doing some more research, I think maybe I’ve described the problem incorrectly or inefficiently. Let me try again: I’m looking for a way to select a subset of points in such a way that maximizes the diversity / entropy among the selected points. For …

[image] Snowy: This problem becomes increasingly more difficult the more points or subsets required and this is where I’m struggling. Yes - the best solution is going to depend very much on the total number of points, as well as the number that you want to select. Choosing 4 points from 100 (…

This problem objective is to find a subset of k indices p so that sum(D[p,p]) is maximum, where D is the all-to-all distances of your points. Whenever I am desperate and I have no idea what I am doing, I lean on Genetic Optimization. Just start with a pool of 50 random p “genes”. Then, “Mate” r…

Thanks for the reply… When I read your post i thought it was a good idea but either I fully don’t understand or it’s not practical in this problem. If I have 1000 points and need the best 50 so that the sum of these 50 is maximum, I would: choose 50 points at random sum the distance between all t…

How to cluster points with a maximum cluster size? Or find points with maximum diversity?

Specific Domains Data

BambOoxX November 6, 2024, 9:45pm 3

See my question from a few weeks back Balanced clustering or assignment of 3D points in space. Let me know if it can be of help (the code is not yet ready to push)

Topic		Replies	Views
Balanced clustering or assignment of 3D points in space Machine Learning question , clustering , geometry	15	492	October 27, 2024
Which JuMP.jl solver for this problem? Optimization (Mathematical) jump , optimization	19	2933	July 27, 2020
Kmeans problem General Usage	1	716	February 5, 2020
Help setting up a problem with JuMP and/or Clustering.jl Optimization (Mathematical)	5	636	July 19, 2020
K-Means with penalty on (weighted) cluster size General Usage	4	920	November 26, 2019

How to cluster points with a maximum cluster size? Or find points with maximum diversity?

Related topics