In recent years, community detection in overlapping weighted network became a research challenge. In real networks, a node can belong to two or more communities. Therefore, in this paper, we aim to address the above-mentioned problem by proposing a method to improve the modularity in overlapping weighted networks. The proposed method is based on optimizing a fitness function and fuzzy belonging degree of nodes. Experimental results on real networks, confirm a significant improvement of modularity in comparison with a similar algorithm.