Fuzzy C-means (FCM) clustering algorithm is one of the most popular fuzzy clustering techniques because it is efficient, straightforward, and easy to implement. However, FCM is sensitive to initialization resulting in local minimization and noise points. In this paper, a novel Fuzzy C- means clustering algorithm based on breadth-first search algorithm (BFS) and coefficient of variation weighting is proposed. Breadth-first search algorithm is a global optimization tool and it is employed to determine the appropriate initial clustering centers and eliminate the noisy data. Moreover, the objective function of FCM is improved by introducing coefficient of variation weighting for reducing noise contributions. The experimental results show that our proposed method is efficient and can overcome the defects of the traditional FCM. In addition, compared to other clustering algorithms, the new one makes convergence faster, clustering accuracy better and noise immunity higher.