Abstract: Reservoir classification is one of the important parts of the study involving oil and gas exploration and development. Engineering data indicate that reserves of tight sandstone reservoir account for a large proportion of the total reserves in low permeability reservoirs, therefore it is meaningful to work on the tight sandstone reservoir classification. Considering the specialty and various kinds of parameters of the tight sandstone gas field of Xinchang Xu-2 formation in the west part of Sichuan Province, the four parameters, that is, porosity, permeability, flow unit, and shale content are selected as reservoir parameters, and the K-means clustering method is used to divide the formations into five categories. On basis of this, by comparing the accuracy of the discriminant analysis with that of decision tree, the decision tree classification method is used to build the quantitative evaluation of the reservoir model. It is proved by the actual data that decision tree has a better result for enhancing the effect of reservoir classification and evaluation.