“…Then, a binary RVM (1, 4, 5, 6, 7, 9, 10, 11 vs. 2, 3, 8, 12) model is established on G 1 and G 2 . In the second stage, the first cluster G 1 is divided into two clusters, G 11 (1,9,10,11) and G 12 (4,5,6,7), by the KFCM algorithm. Similarly, a binary RVM (1, 9, 10, 11 vs. 4, 5, 6, 7) model is established on G 11 and G 12 .…”