momtr / K-Means-Clustering

🪐 K-Means clustering algorithm from scratch

K-Means-Clustering

The implementation of K-Means Clustering is based on Lloyd's algorithm:

define k (number of clusters)
define k starting points (for the clusters) from all given data points
while assignment changes:
  assign each point to a cluster where the eucledian distance between the cluster's point and the given data point is the smallest
  assignment (the cluster's mid point): point_cluster = (1/n) * sum(all points within that cluster) where n is the number of points within that cluster
end while

Details: https://en.wikipedia.org/wiki/K-means_clustering

Output for k = 3:

Run

Start a server and run index.html.

Documentation

I do not recommend using that model, it's just for undestanding the concept haha

KMeans(data, [k, accuracy])

constructor function

KMeans(data, [k, accuracy])

data - array2d, array of numbers
k - number, number of clusters (default: 3)

accuracy - number, 0 < high accuracy < 1 < small accuracy (default: 0.1)

train() : void

trains the model

kmeans.train()

predict(point) : string

predicts a new data point

kmeans.predict([value_1, value_2, ..., value_n]) 

point - array1d, length must match dimension of training data
returns the name of the cluster ('C0', 'C1', ... , 'C'n)

renameCluster(old, label) : void

renames a cluster (default names: 'C0', 'C1', ... , 'C'n (n is the dimension of the input data)

kmeans.renameCluster('C0', 'myCluster')

old - string, cluster's old name
label - string, cluster's new name

static normalize(array2d) : array2d

normalizes array2d (i.e. elements of array)

kmeans.normalize([[value_1, value_2, ..., value_n], ..., [value_1, value_2, ..., value_n]])

array2d - array2d, array to normalize
returns normalized array