Algorithm 4 bisecting K-means

Input:        $X_1,\dots ,X_m$– a set of points to be clustered;        K–number of clusters; Output:   $L_C$– a cluster list indicating the cluster assignments of the points;  1:Initialize a list of cluster $L_C$ to contain the cluster consisting of all the points;  2:repeat  3:   Remove the cluster C with the largest SSE from the list of clusters $L_C$;  4:   $\{C_1,C_2\}\leftarrow$K-means(C,2)  5:   for $i=2$ to number of trials do  6:       $\{C_1^{\prime },C_2^{\prime }\}\leftarrow$K-means(C,2)  7:       if $\mathit{SSE}\left(\{C_1^{\prime },C_2^{\prime }\}\right)<\mathit{SSE}\left(\{C_1,C_2\}\right)$ then  8:          $\{C_1,C_2\}\leftarrow \{C_1^{\prime },C_2^{\prime }\}$  9:     end if 10:   end for 11:   $L_C\leftarrow L_C\cup \{C_1,C_2\}$ 12:until $|L_C|==K$ (the size of $L_C$ is equal to K)