A novel implementation method of the FCM
Shi Hong-Bo, Yu Jian, Huang Houkuan · 2003
The paper proposes a new implementation method of the fuzzy C-means (FCM) algorithm. The traditional implementation method of the FCM algorithm mainly adopts heuristic guidelines to determine all the parameters of the FCM depending on the user's experiences, especially for the weighting exponent. The paper gives a new implementation approach to the FCM algorithm based on a new theory of choosing the weighting exponent, and uses a clustering validity function for validating the results of clustering when adopting different weighting exponents. The effectiveness of this method is verified by our experiments.