Parallel method for agglomerative clustering of non-stationary data

The disclosure is directed to clustering a stream of data points. An aspect receives the stream of data points, determines a plurality of cluster centroids, divides the plurality of cluster centroids among a plurality of threads and/or processors, assigns a portion of the stream of data points to each of the plurality of threads and/or processors, and combines a plurality of clusters generated by the plurality of threads and/or processors to generate a global universe of clusters. An aspect assigns a portion of the stream of data points to each of a plurality of threads and/or processors, wherein each of the plurality of threads and/or processors determines one or more cluster centroids and generates one or more clusters around the one or more cluster centroids, and combines the one or more clusters from each of the plurality of threads and/or processors to generate a global universe of clusters.