CACS: A Novel Classification Algorithm Based on Concept Similarity

This paper proposes a novel algorithm of classification based on the similarities among data attributes. This method assumes data attributes of dataset as basic vectors of m dimensions, and each tuple of dataset as a sum vector of all the attribute-vectors. Based on transcendental concept similarity information among attributes, this paper suggests a novel distance algorithm to compute the similarity distance of each pairs of attribute-vectors. In the method, the computing of correlation is turned to attribute-vectors and formulas of their projections on each other, and the correlation among any two tuples of dataset can be worked out by computing these vectors and formulas. Based on the correlation computing method, this paper proposes a novel classification algorithm. Extensive experiments prove the efficiency of the algorithm.