The problem of link prediction in online social networks like facebook, myspace, Hi5 and in other domains like biological network of molecules, gene network to model disease have became very popular because of the structural connections and relationships among the entities. The classical methods of link prediction based on the topological structure of the graph exploit all different paths of the network which are being computationally expensive for large size of networks. In this paper, incorporating the small world phenomenon, the proposed algorithm traverses all the paths of bounded length by considering clustering information and the connection pattern of the edges as weights on the edges in the graph. As a result, the proposed algorithm will be able to predict accurately than the existing link prediction algorithms. Our analysis and experiment on real world networks shows that our algorithm outperforms other approaches in terms of time complexity and the prediction accuracy.