K-medoids is a partitioning clustering algorithm related to the K-means algorithm. Given a set of n objects and a k number that determines how many clusters you want to output, K-medoids divides the dataset into groups, trying to minimize the average quadratic error, the distance between the points of a cluster, and the point designated to be the center.
Unlike K-means, where this point is artificial (that is, the pure average of all points in the cluster also named centroid), in K-medoids, the point used is placed more centrally; so the center is one of the actual data points (medoid). A medoid can be defined as an object of a cluster whose average disparity over all objects in the cluster is minimal; thus, it will be the most central point of a given dataset. K-medoids are more robust to noise and outliers than K-means, because a mean is easily influenced by extreme values.