Abstract
Fuzzy c-mean classification (FCM), multi-layer perceptron neural network (MLP), fuzzy ARTMAP, and linear discriminant analysis (LDA) were used to classify Forest Inventory and Analysis (FIA) plots into six ecological habitats in the U.S. Northeast. Among the four classifiers, both FCM and MLP produced “soft” classifications based on fuzzy membership values. In contrast, fuzzy ARTMAP and LDA generated only “hard” classifications in which a plot is assigned to the dominant habitat class, based on binary logic, without considering any coexisting classes. The error matrix and several accuracy indices were calculated to assess the classification accuracy of the methods and to test the differences among the four classifiers. The classification accuracies of FCM and MLP were 98% and 97%, respectively, for overall classification, while fuzzy ARTMAP and LDA had an overall classification accuracy of 92% and 85%, respectively. The χ2 test showed that there was no significant difference between the FCM and MLP methods in classification accuracy. But they were significantly better than both fuzzy ARTMAP and LDA methods at a significant level of α = 0.05. This study showed that the fuzzy classifiers such as FCM and MLP were preferable in the classification of ecological habitats using FIA plots, especially for the classification of ambiguous plots with mixed overstory and understory species compositions. FOR. SCI. 50(1):117–127.