Abstract: The distance definition in Euclidean space was weighted and the weight coefficient of samples effecting on the fuzzy clustering centers was estimated by using the similarity relations between samples and fuzzy clustering centers. As a result, the improved method was proved to have  a better result by using the method of data simulation. Meanwhile, the improved method was introduced into the red tide prediction, which visually reflects the distribution of samples of marine elements .

Key words: FCM algorithm, similarity relation, red tide, prediction and alarm