The smoothed aggregation algebraic multigrid (SA-AMG) method is among the fastest solver for large- scale linear equations. The SA-AMG method achieves good convergence and scalability by damping various wavelength components efficiently. To damp this, this method creates multi-level matrices which are hierarchically smaller than the dimension of the original matrix. Moreover, the convergence can be further improved by setting near-kernel vectors. Generally, the same number of near-kernel vectors is used at each level. In the present work, we propose and evaluate the method that extracts and adds them at each level. By this, the performance is improved compared with previous work.