In recent years, E-commerce had made a tremendous impact on the world. However before the emergence of E-commerce, individuals can't skim the information about the products within short time of the period, so therefore recommendation system was introduced. The principle point of the recommendation system is to prescribe the most appropriate items to the user. Many of the recommendation systems mainly use content based method, collaborative filtering method, demographic based method and hybrid method. In this paper, the major challenges such as 'data sparsity' and 'cold start problem' are addressed. To overcome these challenges, we propose a new methodology by combining the clustering algorithm with Eclat Algorithm for better rules generation. Firstly we cluster the rating matrix based on the user similarity. Then we convert the clustered data into Boolean data and applying Eclat Algorithm on Boolean data efficient rules generation takes place. At last based on rules generation recommendation takes place. Our experiments shows that approach not only decrease the sparsity level but also increase the accuracy of a system.