The overabundance of carbon emissions is widely considered as a serious world problem. This paper focuses on analyzing the influence of economic factors on carbon emissions. Based on the traditional STIRPAT model, in terms of the “pollution haven hypothesis” and “pollution halo hypothesis”, this paper employs the dynamic panel data model to explore the impact of economic elements such as economic growth, population, foreign direct investment and others on carbon emissions. Based on our research, China’s urban carbon emissions do not follow the inverted U-shaped hypothesis of the traditional EKC curve theory and presents an inverted N-type. Moreover, current foreign direct investment increases the carbon emissions of Chinese cities due to the “implicit trade carbon”. However, during the lagging period of one phase, it significantly reduced urban carbon emissions. In addition, the lag of one period of carbon emissions statistically led to carbon emissions at the current stage. According to the empirical analysis results, this paper proposes some reasonable improvements for carbon dioxide emission reduction, which have certain reference values for other developing countries facing similar carbon emission reduction challenges.

Meta learning is an advanced field of Artificial Intelligence (AI) where automatic learning algorithms are applied to acquire learning experience for a set of learning algorithms to improve learning performance. One of popular Meta learning methodologies is based on cross validation, especially for selection processes among different machine learning models. However, the challenge is that it is a very time-consuming to do cross validation among models in large data sets, especially in financial big data with high noise. This paper proposes two Asymptotic Meta learning algorithms (AML-lin and AML-xiang), which are ordinal optimization algorithms for Meta learning based on cross validation. The numerical experiments and real-world cases are conducted to illustrate its efficiency in cross validation of models in different scenarios, especially for financial data. The method proposed in this paper has significant improvement by comparing with those ones in existing algorithms OCBA and IAML (e.g., see reference [8] [9]), and it is new in dealing with financial data.