Algorithm 2 User Clustering Scheme for MIMO-NOMA with A Given Relative Weight

1:Assign the first S users of $\mathbf{U}$ to S different clusters. 2:Construct the utility function as (20) based on the given relative weight. 3:forj=S + 1:K do 4:   Sort multiple clusters based on user j’s preference. 5:   Denote the sorted sequence by ${\mathrm{\Omega }}_{\mathbf{j}}$. 6:   while ${\mathrm{\Omega }}_{\mathbf{j}}$ $\ne Ø$ do 7:     $n_j$← the first cluster in ${\mathrm{\Omega }}_{\mathbf{j}}$. 8:     Insert user j to ${\mathbf{U}}_{{\mathbf{n}}_{\mathbf{j}}}$. 9:     $NUM$← the number of the users in ${\mathbf{U}}_{{\mathbf{n}}_{\mathbf{j}}}$. 10:     if $NUM\le M$ then 11:        Break. 12:     else 13:        for i = 1:M + 1 do 14:          Remove the i-th user from ${\mathbf{U}}_{{\mathbf{n}}_{\mathbf{j}}}$. 15:          Obtain the optimal beam vector for cluster $n_j$ by Algorithm 1. 16:          ${\epsilon }_i$← the sum rate of the users in ${\mathbf{U}}_{{\mathbf{n}}_{\mathbf{j}}}$. 17:          Insert the removed user to its original position. 18:        end for 19:        $\tilde{i}\leftarrow$ the position of the maximum in $[{\epsilon }_1,{\epsilon }_2,\dots ,{\epsilon }_{M+1}]$. 20:        $x\leftarrow$ the index of the $\tilde{i}$-th user in ${\mathbf{U}}_{{\mathbf{n}}_{\mathbf{j}}}$. 21:        Remove the $\tilde{i}$-th user from ${\mathbf{U}}_{{\mathbf{n}}_{\mathbf{j}}}$. 22:     end if 23:     if $x=j$ then 24:        Remove cluster $n_j$ from ${\mathrm{\Omega }}_{\mathbf{j}}$. 25:     else 26:        Break. 27:     end if 28:   end while 29:end for