Regional Disparities of Economic Competitiveness in China based on Artificial Neural Network and Hierarchical Cluster Analysis
中国大陆经济综合竞争力区域差异分析

Comprehensive economic competitiveness is the ability of an area to attract resources and compete within a larger region, and results from the optimization of regional resource allocation based on economic development status and development potential. It is a reflection of economic strength, economic extroversion, financial environment, innovation factors, government management and science and technology development, and results from a combination of many kinds of economic variables and the resulting dynamics. This study is based on 31 provinces, cities, and autonomous regions in mainland China. We selected 17 indices related to regional comprehensive economic competitiveness, and applied Self-Organizing Feature Maps (SOFM) based on Artificial Neural Network methods and statistical analysis. The regions were divided into four types, ranging from the highest level of economic competitiveness to the lowest. The most competitive areas are Beijing, Shanghai and Tianjin. The next level, considered to be strongly competitive, includes Guangdong, Jiangsu, Shandong and Zhejiang provinces. Medium-level competitiveness is characteristic of Liaoning, Fujian, Shanxi, Hubei, Sichuan, Hunan, Henan and Hebei. The weakest areas are Inner Mongolia, Heilongjiang, Jilin, Shaanxi, Chongqing, Anhui, Guangxi, Jiangxi, Hainan, Guizhou, Yunnan, Gansu, Qinghai, Ningxia, Xinjiang and Xizang. The research results indicate clear regional disparities of comprehensive economic competitiveness. The provinces which have strong competitiveness are mainly along the east coast of China, while the weaker provinces are mainly located in the central and western areas. The government at all levels should take effective measures to improve economic efficiency, strengthen support to less-developed areas and industries, and try to reduce regional disparities in economic competitiveness. The research results also demonstrate that application of both artificial neural network and hierarchical cluster analysis can be used for comparison of classification results, which is useful for identifying problems and improving accuracy.