A soft partition discretization algorithm based on fuzzy clustering

Most traditional fuzzy discretization algorithms are sensitive to noise data and ignore the correlation between attributes. For these defects, a soft partition discretization algorithm based on improved fuzzy clustering is presented in this paper. Firstly, the algorithm selects initial clustering centers by large density area and uses density function as samples´ weights to reduce effectively noise interference. Secondly, the compatibility of decision table in rough set theory is used as criteria to adjust dynamically the parameters of the algorithm so as to achieve optimal discretization effect. Finally, experimental results validate that the algorithm has better discretization effect by using the UCI and astronomical spectrum datasets.