The invention discloses a network
multimedia service semi-supervised classification method based on a t distribution
hybrid model. The method concretely includes a step of
data set preprocessing, a step of t distribution clustering, and a step of classification. In the step of
data set preprocessing, data flow samples of various
multimedia services on
the Internet are collected and then are preprocessed. In the step of t distribution clustering, fitting of a t distribution
hybrid model or a finite t distribution
hybrid model is conducted on the above
network data flow samples, and K multidimensional t distribution clusters are obtained. In the step of classification, results of the above clustering are further classified, and a total correct rate of the final classification is calculated. The t distribution
hybrid model is utilized to conduct more accurate fitting on
multimedia services, so the classification accuracy is improved. An EM
algorithm of the finite t distribution
hybrid model effectively increases the convergence speed of the t distribution
hybrid model. An experiment shows that the
algorithm is high in accuracy, and the fitting model is superior to a conventional K-means
algorithm and a conventional EM algorithm of a Gauss hybrid model.