An improved FCM clustering algorithm based on cosine similarity

Abstract

Based on the traditional Fuzzy C-means (FCM) clustering algorithm, this study adds cosine similarity as a correction factor and optimizes the FCM algorithm by optimizing the membership degree of the objective function. The results show that the matrix estimation error obtained by the improved algorithm is smaller and the precision is higher, which can reduce the normalized mean square error by about 20.67%, and the angular deviation is reduced by about 8° on average.