Abstract: Under the strategy of innovation-driven development, aiming at meeting the challenges of the sustainable transformation and growth and dealing the impact of external environment on efficiency in the development of low-carbon innovation network in China, this paper builds the evaluation system of regional low-carbon innovation network including undesirable output. Based on the regional panel data from 2006 to 2015, this paper employs modified J-SBM three-stage DEA model to evaluate the regional low carbon innovation network efficiency, also analyzes influence mechanism of the environment factors and the random disturbance, and discusses the production mode of low carbon innovation network. On top of that, in terms of method and content of low carbon innovation network development, the “black box” is revealed in the challenging situation of China. The research shows that: the influence of environmental factors and random disturbance on the input slack variable is heterogeneity, and the low carbon innovation network efficiency is underestimated in most areas without considering environmental factors and random disturbances. Regional low carbon innovation network efficiency has steadily risen, Beijing and Shanghai are at the forefront of efficiency, and in most parts of China, there is still a large space to improve efficiency. The production mode of regional low carbon innovation network occupies the most of quadrant of low input and low efficiency region.

Key words: innovation network efficiency, low carbon Innovation, J-SBM model, three-stage DEA model, undesirable output