Distributed computation is an effective policy to increase the speed of the sparse networked systems. In a sparse network, clustering methods like k-means does not work directly as it cannot explore the connectivity of the system. To solve the problem, two modification methods are proposed in the existing graph and a new graph named Spatial Matrix is introduced in this paper. The proposed modification is a fast process and the computation time can be considered negligible compared to the rest of the process. Thus it preserves the ultimate objective of the distribution. It works as a pre-conditioning that can be used with a wide range of clustering and mathematical tools. With distributed state estimation of IEEE 14, 68, and 118-bus systems with automatic clustering, the effectiveness of the spatial matrix is demonstrated.