The rapid growth of information over the Internet makes it difficult to find the desired item among the huge amount of items. Web-based recommendation systems provide product recommendations to online consumers based on their purchase history or previous ratings of items and products, yielding the challenge of obtaining the favorite items of the active user reasonable time. This paper proposes a K-nearest neighbors (KNN)-based framework using a user-based collaborative filtering (CF) technique. The proposed framework is presented with various popular similarity measurements such as, cosine, Pearson correlation coefficient (PCC), and mean squared difference (MSD) with reduction of the execution time of the user-based CF technique. The experimental results show that the proposed approach outperforms the competing ones by an average improvement (i.e., reduction) of 2.6%, 4.17% and 3.8% in terms of the execution time the three similarity measurements mentioned above, respectively. As well, in terms of accuracy, the proposed approach outperforms the competing approaches by an average improvement of 0.35%, 7.7% and 3.15%, respectively, in terms of the root mean square error (RMSE) metric, and by an average improvement of 1.1%, 7.3% and 5.2%, respectively, in terms of the mean absolute error (MAE) metric.