Scholarly article recommendation systems are an essential tool for effective research work. It plays a major role in retrieving relevant scientific papers in the era of big scholarly data. When researchers start working on a research problem, they are not always sure which papers to refer to learn the state-of-the-art or which papers are the most appropriate for their work. Numerous methods for generating recommendations have been proposed in the past decades. Very often, these are generalized systems, not specifically designed for scholarly articles. Moreover, they fail to capture a researcher's preferences for year, authorship, publication venue and so on. In this paper, we present an alternative approach to implementing a recommendation system based on relevance feedback to resolve these concerns. Extensive experiments have been performed on a real-world Microsoft Academic Graph (MAG) dataset to demonstrate that the proposed algorithm produces more accurate recommendations as compared to the baseline methods. Finally, the evaluation has been performed against a few search engines like Google scholar and CiteSeer to demonstrate the effectiveness and the scalability of proposed recommender system.