Influential data subsets labelling: A fuzzy relation approach

The selection of candidate for influential data subsets is an important step in regression analysis in order to construct a model of high quality. One of the most popular and widely used methods for labelling influential data subsets is 1-clustering method. This is an indexed hierarchical clustering based on the notion of subdominant ultrametric (SDU) of a dissimilarity matrix. In the literature, there are many different algorithms available to construct SDU. However, the computational complexity of those algorithms is high; the running time is very slow especially when the sample size is large. In this paper a method based on fuzzy relation approach, which allows us to construct a promising algorithm to obtain SDU even for a large number of data, is introduced. An example will be presented and discussed to illustrate the advantage of the proposed algorithm.