Recommender systems have been acknowledged as efficacious tools for managing information overload. Nevertheless, conventional algorithms adopted in such systems primarily emphasize precise recommendations and, consequently, overlook other vital aspects like the coverage, diversity, and novelty of items. This approach results in less exposure for long-tail items. In this paper, to personalize the recommendations and allocate recommendation resources more purposively, a method named PIM+RA is proposed. This method utilizes a bipartite network that incorporates self-connecting edges and weights. Furthermore, an improved Pearson correlation coefficient is employed for better redistribution. The evaluation of PIM+RA demonstrates a significant enhancement not only in accuracy but also in coverage, diversity, and novelty of the recommendation. It leads to a better balance in recommendation frequency by providing effective exposure to long-tail items, while allowing customized parameters to adjust the recommendation list bias.