Collaborative filtering is the most widespread recommendation system technique deployed in e-commerce services nowadays. It recommends products based on the historical preference of the user. The biggest challenges in these techniques are data sparsity and growing volume of data, specifically in e-commerce sites like movie recommendation. Clustering algorithms are used for scaling up the performance of collaborative filtering in dynamically growing datasets. Most of the existing clustering based recommendation algorithms improve scalability but produce low quality recommendations. This is mainly due to data sparsity, as the user tends to rate very few items from a large number of options available. Moreover, users with a similar taste for a group of items may show different likings for another group of items over a period, i.e.., user’s interest dynamically changes over time. Finding the sub-groups that are more relevant to each other than the entire user-item matrix is more affordable. Since the user’s recent ratings can better represent their interest and preference, a Time Adaptive Collaborative Filtering Method −TACF is proposed, that adopts time to generate a recommendation. Experimental results on the MovieLens dataset show that the proposed system outperforms other state-of-art collaborative filtering algorithms in terms of accuracy and efficiency.