The main work of this paper is as follows. First, this paper sorts out the definition and connotation of the creational and original innovation capacity, and establishes the corresponding evaluation index system. Since the connotation of the creational and original innovation capacity is complex with many interacting factors, it is of importance to select multiple interrelated and dependent evaluation indexes accurately and comprehensively. Based on the principles of systematization, comparability,scientificalness and practicability, this paper designs an evaluation index system of the creational and original innovation capacity, with 4 first-level indexes, 12 second-level indexes and 40 third-level indexes. These indexes are selected from 4 aspects including new academic thought, new scientific discovery, new technological invention and new industrial direction. Meanwhile, three dimensions are also considered including innovation foundation, innovation input and innovation output. The evaluation system built in this paper comprehensively covers the relevant factors of the creational and original innovation capacity. 

Second, this paper selects the real data of 31 provinces and regions of China in 2017, uses AHP method to determine the index weight, and combines TOPSIS method to evaluate and rank the creational and original innovation capacity. AHP and TOPSIS are two classical decision-making methods, which are widely used in evaluation. The combination of the two methods can make the evaluation more scientific and reasonable. It not only reflects the idea of hierarchical analysis, but also pay attention to the characteristics of each index data itself. The integrated method can deal with the comprehensive evaluation information objectively and is suitable for the evaluation of the creational and original innovation capacity. According to the built evaluation index system, a three-level analysis model is established. The target level is the creational and original innovation capacity. The middle level includes 4 secondary indicators, and the indicator level includes 40 tertiary indicators. According to the experts′ opinions, the weight of each evaluation index is determined. The comprehensive evaluation values are calculated by TOPSIS. By calculating the distance between the data to the positive and negative ideal solutions, the relative closeness can be obtained. In this paper, the relative closeness shows the creational and original innovation capacity, which means the larger the relative closeness is the better the capacity will be. The combination of AHP and TOPSIS improves the operability of evaluation and the reliability of the results.

Third, this paper also uses SOM algorithm to cluster the evaluation values of the creational and original innovation capacity, which aims to divide provinces and cities in China into various groups. SOM is a kind of unsupervised machine learning method. Through repeated learning, SOM can capture the pattern characteristics of each input. Compared with other clustering algorithms, SOM can automatically find the internal regulations and essential attributes in the data and the operation result is stable. SOM method can effectively solve the problem of unclear grouping boundaries caused by small differences in data, reduce the subjectivity of artificial observation ranking, and make the grouping results more objective and scientific.

The main conclusions of this paper are as follows. First, through the interpretation of relevant policies and concepts, this paper illustrates that the concept of innovation in the new era should include the following two features:on the one hand, it should completely cover technological innovation and industrial innovation; and on the other hand, it should provide a good environment for cultivating innovation elements and innovative behaviors. Therefore, the creational and original innovation capacity refers to the innovation capacity of from scratch in the aspect of technology and industry. Moreover, a province with good creational and original innovation capacity should not only focus on innovation activities, but also guarantee the continuous progress of innovation, which means to make every potential innovation subject have the chance to innovate and grow continuously. Second, the evaluation results of AHP-TOPSIS methods show that Beijing has strong creational and original innovation capacity, followed by Jiangsu and Guangdong, while Hainan, Guangxi and Xinjiang are in a backward position. Third, the SOM clustering divides China′s 31 provinces into 4 categories. The first category is the provinces and cities with relatively excellent creational and original innovation capacity, including the first-tier cities such as Beijing and Shanghai. The second category is the provinces and cities with better creational and original innovation capacity, represented by Zhejiang and Shandong. The third and fourth categories are areas with relatively weak creational and original innovation capacity, represented by traditional agricultural provinces and some remote provinces such as Guangxi and Xinjiang. 

This paper is one of the earliest attempts to evaluate the creational and original innovation capacity of provinces and regions in China, which helps the public to better understand the differences of the creational and original innovation capacity among various provinces and regions in China.