This is a recommendation system based on a hybrid approach. Proposed approach uses clustering based technique to recommend items from both the content-based clusters of items and co-occurrence-based clusters of items. The items to be selected from each type of clusters depend on a parametric tuning. In order to generate content-based clustering oriented recommendation for an active user, the content based cluster that has more common item to an active user’s preferred item list is selected. Then the predicted preference of the items from the selected cluster excluding the user’s already preferred items are calculated by a weighted summation formula. After this, the top-n items based on predicted preference are selected for content-wise recommendation. A co-occurrence based recommendation list is also generated in similar way. Finally these two recommendation lists are merged together in one user’s preference list length based parametric form to generate a recommendation.