CLARANS (Clustering Large Applications based on RANdomized Search) is efficient and effective and is the best practice for spatial data mining. CLARANS applies a strategy to search in a certain graph. A node in this graph, denoting it as , is represented by a set of objects, ,. Here, k is the predefined value to choose the k medoids; as a result, the nodes in the graph are a set of . If two nodes, , and are neighbors, then . Each node in the graph represents a set of medoids and the cluster related to it. As a result, a cost is related to each node; this cost is the total distance between any objects and the medoid represents its cluster. The cost differential of two neighbors can be calculated with the cost measure function introduced in the PAM algorithm.
18.116.24.97