Algorithm: Two-step Hierarchical Clustering For FPP (G, ε, θ ,freq) Input: All probability subgraphs with k scale Output: frequent probability subgraph g α |
---|
1. Initilize the n graphs {g1,...,gn}as the n leaves of cluster tree ; 2.  While Change_label!=0 3.     Change_label = 0; // Change_label indicates whether the process of merging clustering operation 4.     Lc = size(ResidentGraph); //Calculation subgraph number, Lc represents the total number of clusters 5.     For i= 1 to Lc /2 6.         <Iso,inje,VMval>=IsomorphismCal(gi, gi+ L c /2, ε, θ);         // Determine gi, gi+ L c /2probability isomorphic 7.         If Iso= =TRUE 8.             gi = union(gi, gi+ L c /2); 9.             Change_label ++; 10.             ResidentGraph = {ResidentGraph i}; //if isomorphic, retaining only the subgraph label i to ResidentGraph 11.             Else             ResidentGraph = {ResidentGraph i i+ L c /2};//if not isomorphic, retaining only the subgraph label i, i+ L c /2 to ResidentGraph 12.         End if 13.     End for 14.     End while 15.         SimpleHierarchicalClusteringForFrequentSubgraphWithPro(ResidentGraph, ε, θ ,freq); //using a simple hierarchical clustering for the remaining probability subgraphs 16.         Calculate the probability isomorphic frequency p of the residual clusters gr ; |