active-analytics.clustering.k-means — de.active-group/active-analytics 0.3.0

choose-initial-centroids clj

(choose-initial-centroids data k)

Takes k random (distinct) elements from data to use as initial centroids.

cluster clj

(cluster data centroids distance-fn)

Builds clusters by assigning each point to its nearest centroid.

default-centroid-fn clj

(default-centroid-fn points)

Calculates a centroid for points. For now, assumes the points to be neanderthal vectors.

k-means clj

(k-means data
         distance-fn
         max-number-of-iterations
         &
         {:keys [k threshold initial-centroids centroid-fn]})

Performs k-means clustering on data, given the desired number of clusters k, a function distance-fn that computes the distance between two data points, and a max-number-of-iterations after which to stop the algorithm.

Takes keyword arguments :k, :threshold, initial-centroids, and :centroid-fn.

• :k is the target number of clusters. If no :k is passed to k-means, it will assume :initial-centroids to be present and count those to use for :k.
• After each step, calculate the distance between new and previous centroids. Stop the algorithm once this is less than :threshold in every centroid. No given threshold means that max-number-of-iterations iterations will be done.
• The :initial-centroids designate the starting point for the iterations. If none are passed, the algorithm will use :k random data points.
• The :centroid-fn should take a seq of points (the cluster) and return a new point, the corresponding centroid. If no :centroid-fn is given, data is assumed to be a seq of neanderthal vectors (for now).

nearest-centroid clj

(nearest-centroid p centroids distance-fn)

Finds the centroid closest to a point p.

step clj

(step data centroids distance-fn centroid-fn)

Calculates the next generation of centroids from the current one.