Human activities are increasingly affecting the planet and its sustainability by degrading and damaging the environment. The literature on this topic has demonstrated that Information and Communications Technology (ICT) and human development (HD) are important promoters of progress towards environmental sustainability. The impact of these factors is most often examined by using standard regression analysis which suffers from the problems of multicollinearity and non-linear dependency. In order to resolve this problem, a non-parametric method is proposed. To be specific, a Support Vector Machine (SVM) model for predicting environmental performance growth has been developed, based on various predictors- ICT and HD indexes, population growth, and an economic development indicator. The prediction is made at the macro level using a sample of 139 countries. The model was created by a prediction procedure consisting of the optimization of the SVM learner parameters using the grid-search method, as well as k-fold cross-validation. A predictive accuracy of the SVR model of 80.4% was achieved. The model predicts growth in environmental performance of 1.5% for each 1% increase in the ICT index, while an increase of the HD index of 1% produces an environmental performance increase of 4.3%. The results of the sensitivity analysis confirm that the effects of both predictors are enhanced when they operate in interaction. This the first study to apply the predictive machine learning method to the analysis of the impact of ICT and HD on environmental performance and empirically confirmed its efficiency. The obtained results contribute to the existing literature and could be beneficial to policy makers working in sustainable development.

中文翻译：

---

支持向量机方法，用于从信息和通信技术以及人类发展预测环境可持续性的进展

人类活动正在通过破坏和破坏环境来日益影响地球及其可持续性。有关该主题的文献表明，信息和通信技术（ICT）和人类发展（HD）是朝着环境可持续性发展的重要推动者。这些因素的影响通常是通过使用标准回归分析来检查的，该分析存在多重共线性和非线性相关性的问题。为了解决这个问题，提出了一种非参数方法。具体而言，已经基于各种预测因素（ICT和HD指数，人口增长和经济发展指标）开发了用于预测环境绩效增长的支持向量机（SVM）模型。该预测是使用139个国家的样本在宏观一级做出的。该模型是通过预测程序创建的，该预测程序包括使用网格搜索方法对SVM学习者参数进行优化以及k倍交叉验证。SVR模型的预测准确性达到80.4％。该模型预测，ICT指数每增加1％，环境绩效就会增长1.5％，而HD指数每增加1％，环境绩效就会增长4.3％。敏感性分析的结果证实，当两个预测变量相互作用时，它们的作用会增强。这是将预测性机器学习方法应用于ICT和HD对环境绩效影响的分析的第一项研究，并通过经验证实了其效率。