Clustering Function Based on K Nearest Neighbors

Finds clusters in set of observations based on their numerical attributes. Uses a variation of an graph theoretic algorithm,

This function is similar to the graph theoretic clustering function that I submitted previously (https://www.mathworks.com/matlabcentral/fileexchange/57320-clustering-algorithm-based-on-directed-graphs). The input is an observation/attribute matrix and an integer K that specifies the number of nearest neighbors for each observation. The algorithm first finds the K nearest neighbors of each observation and then a parent for each observation. The parent is the observation among the K+1 whose Kth nearest neighbor is the nearest (check the code for a more precise specification). As in the previous function, orphans become the roots of clusters and the remaining nodes are assigned recursively to the cluster of their parent.