Evaluation of regional green innovation performance in China using a support vector machine-based model optimized by the chaotic grey wolf algorithm

The green innovation performance (GIP) evaluation helps to identify strengths and weaknesses in regional innovation systems and has been crucial for policymakers in developing appropriate regional policies. Recent methodology has focused on establishing an indicator framework and calculating composite scores. It is noteworthy that the non-linear relationship between evaluation scores and indicators is rarely considered. In view of this, an evaluation model was proposed in the study which combines support vector machine (SVM) and chaotic grey wolf algorithm (CGWO). Sixteen indicators from the indicator system of European Innovation Scoreboard were retained for the GIP evaluation with an initial screening of indicators using the information entropy method. Then, four different types of optimization algorithms were used to optimize the SVM to generate non-linear predictions and GIP scores. The applicability of the model was verified for the GIP evaluation of China’s provinces. According to the training and test results, the SVM-CGWO model achieved significantly better performance than the other three algorithms, which has important benefits in improving the uniformity of the wolf distribution and the traversal of the wolf pack, together with enhancing operation speed and accuracy. It helps users to rank and benchmark regional GIP at the provincial level, taking into account performance improvement and accuracy of dimensions, as well as reliability issues.