1: \begin{abstract}

2: We consider a general class of high order weak approximation schemes for

3: stochastic differential equations driven by L\'{e}vy processes with infinite

4: activity. These schemes combine a compound Poisson approximation for the jump

5: part of the L\'{e}vy process with a high order scheme for the Brownian driven

6: component, applied between the jump times. The overall approximation is

7: analyzed using a stochastic splitting argument. The resulting error bound

8: involves separate contributions of the compound Poisson approximation and of

9: the discretization scheme for the Brownian part, and allows, on one hand, to

10: balance the two contributions in order to minimize the computational time, and

11: on the other hand, to study the optimal design of the approximating compound

12: Poisson process. For driving processes whose L\'{e}vy measure explodes near

13: zero in a regularly varying way, this procedure allows to construct

14: discretization schemes with arbitrary order of convergence.

15: 

16: \noindent\textbf{Key words:} L\'{e}vy-driven stochastic differential

17: equations, high order discretization schemes, weak approximation, regular variation

18: 

19: \noindent\textbf{2010 Mathematics Subject Classification:} 65C30, 60G51

20: 

21: \end{abstract}

22: