Self-organizing feature map clustering based on fuzzy equivalence relation and its application in ecological analysis

Abstract

Fuzzy equivalence clustering based on fuzzy set theory and Self organizing feature map (SOFM) clustering based on neural network are both effective methods in ecological studies. They are powerful in analyzing and solving complicated and non-linear matters and for their freedom from restrictive assumptions. The combination of Fuzzy equivalence clustering and SOFM clustering may produce better methodology. This study tried to combine them in a new method, Fuzzy equivalence SOFM clustering, and to apply it in the analysis of plant communities. The dataset was consisted of importance values of 70 species in 30 samples of 10 m × 20 m. First, we calculated fuzzy similarity matrix of samples; second, transforming fuzzy similarity matrix to fuzzy equivalence relation matrix; third, the fuzzy equivalence relation matrix was input to neural network and then SOFM was used to classified samples. The 30 samples were clustered into 5 groups, representing 5 vegetation communities. This classification result was reasonable and ecological meaningful which suggests that fuzzy equivalence SOFM clustering is effective method in ecological study. The fuzzy equivalence SOFM clustering shares both advantages of fuzzy set theory and neural network.