传统的模糊C均值算法直接基于原始数据进行聚类，数据的内在结构可能会被噪声、异常值或其他因素破坏，因此聚类性能会受到影响。为提升FCM算法的鲁棒性，提出了一种基于自适应近邻信息的模糊C均值聚类算法。近邻信息指的是一种基于数据点之间相似度的度量，每个数据点都可以看作其他数据点的近邻，但是不同数据点之间的相似度是不同的。将样本点的近邻信息GX和类中心点的近邻信息GV融入基础FCM模型中，为聚类过程提供更多的数据结构信息，用于指导聚类算法中的簇划分过程，以提升算法的稳定性，并提出了3个迭代算法求解本文提出的聚类模型。与其他先进聚类算法对比，在部分基准数据集上聚类性能有10%以上的提升，同时还从参数敏感性、收敛性、消融实验等方面对算法进行评价。实验结果可以充分显示本文提出的聚类算法的可行性与有效性。AI语音播报 

Traditional FCM algorithms cluster based on raw data, risking distortion from noise, outliers, or other disruptions, which can degrade clustering outcomes. To bolster FCM's resilience, this study introduces a fuzzy C-means clustering algorithm that leverages adaptive neighbor information. This concept hinges on the similarity between data points, treating each point as a potential neighbor to others, albeit with varying degrees of similarity. By integrating the neighbor information of sample points, labeled G_X, and that of cluster centers, labeled G_V, into the standard FCM framework, the algorithm gains additional insights into data structure. This aids in steering the clustering process and enhances the algorithm's robustness. Three iterative methods are presented to implement this enhanced clustering model. When compared to leading clustering techniques, our approach demonstrates over a 10% improvement in clustering efficacy on select benchmark datasets. It undergoes thorough evaluation across different dimensions, including parameter sensitivity, convergence rate, and through ablation studies, confirming its practicality and efficiency.