The invention discloses a volume tense data modeling method based on a determined learning theory. The method aims at volume tense data with noise, wherein the volume tense data are ubiquitous in science researches and engineering practices and generated a dispersed nonlinear system normally. A modeling process comprises that the volume tense data are defined and preprocessed through filtering; a neural network identifier is structured by adopting a dynamic radial basis function (RBF) neural network, and a reasonable adjusting rule of RBF neural network weight is designed; and a model is built for the neural network of the volume tense data. The volume tense data modeling method has the advantages of simplifying modeling of the volume tense data and performance analysis, being applied to rapid similarity judgment of the volume tense data and the like.