Millions of research papers are available to users from various digital libraries. Finding relevant research that fits a researcher's tastes is therefore a difficult process. As a result, various paper recommendation models have been put out to deal with this problem. The previous studies have concentrated on only the historical preferences of the users. The diversified user preferences as well as interests of the users were not given enough priorities. We recommend a technique for making tailored paper recommendations that includes various preferences and users’ interests. In order to understand user preferences, we look into meta-paths in the network. We use random walks over meta-paths to calculate the recommendation scores of potential papers for the intended users. Then we use Bayesian Personalised Ranking and a personalised weight learning technique to identify a user's personalised weights on various meta-paths. Combining the scores of recommendation from several meta-paths with customized weights yields a global recommendation score. The proposed model performs better than existing state-of-art models.