Recommendation system is a niche area of research in the field of artificial intelligence where the inputs are user-item rating matrices. Collaborative filtering is the generally acknowledged solution for constructing recommendation systems. Matrix factorization is a memory-efficient collaborative filtering scheme that creates a latent feature space for users and items. In this paper, we propose a novel hybrid combination of two matrix factorization approaches employing stochastic gradient descent and alternating least squares, respectively, as learning algorithms, and an artificial neural network as the meta-learner. K-Nearest Neighborhood computation based on the cosine similarity metric is used for shortlisting the top-rated items under consideration for the meta-learning process. The primary task of the meta-learner is to learn parallelly from multiple collaborative filtering schemes which improves the accuracy of the recommendations. The efficacy of our approach is proved by the significantly reduced Root Mean Square Error (RMSE) of 0.6459 as compared to the baseline methods for the experiments conducted on the benchmark MovieLens-100 K dataset.