Improving efficiency of water resources utilization is the key to alleviate the contradiction between water supply and demand while achieving sustainable economic and social development. In order to analyze and evaluate the water consumption efficiency of the provincial capital cities in China, this paper selected five input indicators and six output indicators from agriculture, industry, life and society based on data envelopment analysis method. Then we used Akaike Information Criterion (AIC) for variable selection and constructed a scientific and rational water-use efficiency indicator system. On this basis, the Shannon Entropy Index is used to enhance the recognition ability of the traditional CCR model, and the complete ranking of water resources efficiency in the provincial capitals is given. The results show that (1) the comprehensive efficiency scores (CES) of most cities are not high. Their input-output ratio still has a lot of room for improvement. (2) The scores of CES in Lasa, Beijing, Tianjin, Yinchuan, Haikou, Shanghai etc. are relatively high, which indicate that there may be no necessary connection between the level of water use efficiency of a city and the level of economic development. At the same time, other cities should combine their own conditions to learn from cities with higher scores in CES. (3) Cities such as Chongqing, Nanning, Nanchang, and Changsha are rich in water resources, but the CES is not ideal. This shows that a large amount of water resources may be wasted and these cities should optimize their industrial structures while establishing a water-saving mechanism. In addition, due to the lack of variable selection system in most literatures and the inadequacy of traditional models in the ability of efficiency identification, this paper innovatively combined the AIC variable selection method, data envelopment analysis method, and Shannon Entropy Index method. The combination of the three methods can achieve the advantages and disadvantages of complementarity, not only fully consider a large number of input and output variables, make the analysis closer to reality, the variable indicator system obtained by the AIC criterion also perfectly conforms to the meaning of actual water resource utilization efficiency.