An algebraic multi-class classification method

An algebraic multi-class classification method AHSC, i.e., algebraic hyper surface classification, is proposed. The separating algebraic hyper surface of two-class data may be directly constructed by a single polynomial in theory, but it is too difficult to separate multi-class data by a single polynomial even though the polynomial is multivalued. AHSC can be used for classifying multi-class data by integrating a series of polynomial networks based on binary numbers, which are used for labeling the classes of samples. The problem that multi-class data cannot always be separated by a single polynomial is solved by AHSC. Moreover, using an adaptive method can choose the order of polynomial. The experimental results show that the new method can efficiently and accurately classify multi-class and high dimension data.