Personalized item ranking for recommending top-N items of interest to a user is an interesting and challenging problem in e-commerce. Researchers and practitioner are continuously trying to devise new methodologies to improve the accuracy of recommendations. Recommendation problem becomes more challenging for sparse binary implicit feedback, due to the absence of explicit signals of interest and sparseness of data. In this paper, we deal with the problem of the sparseness of data and accuracy of recommendations. To address the issue, we propose an interest diffusion methodology in heterogeneous information network for items to be recommended using the meta-information related to items. In this heterogeneous information network, graph regularized interest diffusion is performed to generate personalized recommendations of top-N items. For interest diffusion, personalized weight learning is performed for different meta-information object types in the network. The experimental evaluation and comparison of the proposed methodology with the state-of-the-art techniques using the real-world datasets show the effectiveness of the proposed approach