The shortage of resources and the increasingly severe environmental pollution have made the eco-innovation become a political goal pursued by countries all over the world. The study on the eco-innovation efficiency of the different industrial sectors in China will be of great significance for the innovative and sustainable development. Starting from the connotation of eco-innovation, an assessment index system of eco-innovation efficiency was established, including environmental benefits and economic benefits. The panel data of 34 industrial sectors from 2009 to 2015 in China was taken as a sample and the 34 industrial sectors selected were classified into resource-intensive industries, labor-intensive industries and capital-intensive industries by using the production factor intensity in this study. The projection pursuit classification model based on real number encoded accelerating genetic algorithm was used for comprehensive assessment, while this method can judge the contribution size and direction of each assessment index to the comprehensive assessment objective according to the optimal projection direction, as well as realize the unified classification and assessment for the 34 industrial sectors through using the projection index value. The results show that the eco-innovation efficiency varies significantly in the industries with different factor intensities. The eco-innovation efficiency of the capital-intensive industries is the highest, while that of the resource-intensive industries and labor-intensive industries are close to each other but both are lower than the overall average of the industry. Among them, the non-ferrous metal mining and dressing industry in labor-intensive industries shows the lowest eco-innovation efficiency. The input of manpower and financial resources in the research and development of eco-innovation and the economic benefits brought by innovation activities have a great influence on the improvement of eco-innovation efficiency, while the input of materials, such as instruments and equipment, etc., have less influence on that.