MOUNTAIN METHOD-BASED FUZZY CLUSTERING: METHODOLOGICAL CONSIDERATIONS

Abstract

The mountain method is a grid-based procedure for determining the approximate locations of cluster centers in data sets with clustering tendencies. This paper supplies additional background and detail in two important areas. In the first part of the paper, crucial methodological considerations arc discussed, including the choice of grid size, the choice of parameter values, and the notion of “peak non-reusability.” The paper also introduces the possibility of using the singular value decomposition with mountain method output as a means of estimating the number of clusters in the data. In the second pan of the paper, the mountain method is discussed as part of a general grid-based approach to the location of “objects” in spatial data. An example using Anderson's iris data is included, and an application of the method to a problem in robotics is described in detail.