Two new efficient algorithms are developed to approximate the derivatives of sufficiently smooth functions. The new techniques are based on differential quadrature method with quartic B-spline bases as test functions. To obtain the weighting coefficients of differential quadrature method (DQM), we use the midpoints of a uniform partition mixed with near boundary grid points. This enables us to obtain the weighting coefficients without adding the new extra relations. By obtaining the error bounds, it is proved that the method in its classic form is non-optimal. Then, some new weighting coefficients are constructed to obtain higher accuracy. By obtaining the error bounds, it is proved that the new algorithm is superconvergent. Afterwards, by defining some new symbols, we find a way to approximate the partial derivatives of multivariate functions. Also, some approximations are constructed to the mixed derivatives of multivariate functions. Finally, the applicability of the methods is examined by solving some well-known problems of partial differential equations. Some examples of 2D and 3D biharmonic, Poisson, and convection-diffusion equations are solved and compared to the existing methods to show the efficiency of the proposed algorithms.
