Comprehensive Evaluation Of Global Intelligence Innovation Based On Machine Learning Methods

In the era of smart economy,the development of smart technology accelerates the process of the fourth industrial revolution with "data and computing" as the key factor of production.As the world moves towards the smart era,how to comprehensively measure the level of intelligence innovation in various countries has become an important research problem.This paper mainly carries out a comprehensive evaluation research on global intelligence innovation from three aspects:the construction of index system,the determination of index weight and the analysis of index results.The main research points are as follows:(1)In terms of the construction of the comprehensive evaluation system,this paper firstly proposes a conceptual framework of national intelligence innovation based on the traditional input-output linear industrial model,which combines the characteristics of national intelligence innovation with the innovation ecosystem perspective.Under the conceptual framework of national intelligence innovation,28 indicators are selected from three aspects:driving force of intelligence innovation,resource input of intelligence innovation and performance output of intelligence innovation to build a comprehensive evaluation system of global intelligence innovation.(2)Faced with the problem of partially missing data,this paper uses machine learning algorithm(Missing forest interpolation,K-Nearest Neighbor algorithm interpolation)methods and two types of statistical-based methods(Mean interpolation,Median interpolation)to conduct comparative interpolation experiments by randomly missing complete datasets.Taking NRMSE as the indicator of interpolation performance,the KNearest Neighbor(KNN)algorithm performs better than other methods to interpolate the missing values based on the interpolation results of the current dataset.(3)In terms of determining the indicator weights,this paper constructs a random forest model under unsupervised clustering algorithm to assign weights to the indicators.The model first uses principal component analysis(PCA)for indicator dimensionality reduction,then selects the K-means unsupervised clustering algorithm to generate clustering labels for the samples,and uses the clustering labels as the outcome variables of the sample set,and finally combines the random forest algorithm to determine the indicator weights.(4)In terms of the analysis of the index results,this paper firstly adopts a linear synthesis method to form the Intelligence Innovation Index and country rankings of 101 countries based on the results of the above index weighting results,and then develops a multi-dimensional comprehensive analysis.This includes the correlation analysis between the key indicators and the index results,as well as the grading evaluation of the global intelligence innovation index among different types of countries.The main findings show that the global intelligence innovation level presents an imbalanced situation in general,with China and the United States leading the way in intelligence innovation development.Among all indicators,there is a significant positive correlation between national GDP per capita and ICT development and the results of the Global Intelligence Innovation Index(GⅢ).In addition,class A(leading type)and class B(catching type)countries of intelligence innovation development are mainly concentrated in the group of high-middle income countries,while all low-income countries are class C(lagging type)countries of intelligence innovation development.The Global Intelligence Index provides a reference for the level of intelligence innovation in various countries around the world,enriching the research results related to comprehensive evaluation of intelligence innovation and helping policy makers in various countries around the world to better formulate relevant policies.In particular,it has theoretical and practical implications for China to improve the level of intelligence innovation and improve the intelligence innovation system,thus it promotes the highquality development of smart economy.