Online environments offer a major advantage that data can be accessed freely. At the same time however, they present us with an issue of trust: how much any data from online sites can be trusted. Trust and Reputation Systems (TRS), developed to address this issue of trust on network, quantify reliability in terms of semantics and derive a trust-network from a targeted online data. The performance of TRS is often hindered despite the promises because the number of links formed in the ideal scenario frequently is not reached, suffering from the problems of cold-start and sparsity. In this paper, we propose a method in which Link Prediction(LP) and Clustering are applied to TRS so that these two problems are adequately addressed. We evaluate our proposed method with a recommendation system we constructed. Our experiment results show that our method positively contributes to the performance of a recommendation system and help control the problems of cold-start and sparsity in TRS.