This paper proposes a method to improved computation speed of forward kinematics solution of Stewart platform using neural network. Solution of the inverse kinematics is easily found by the vectors of the links which is composed of the joint coordinates in base and plate frame, but forward kinematic is not easily solved because of the nonlinearity and complexity of the Stewart platform. Actually we can get the solution of forward kinematics by solving nonlinear simultaneous equation by Newton-Raphson, but this method is too slow computation speed. Because this method has some problem about setting initial condition. Hence we suggest a method to increase the computation speed by setting the initial condition of Newton-Raphson with neural network.