Assessing environmental performance with big data: A DEA model with multiple data resources

Big data generated by environmental monitoring equipment create a good opportunity for improving perfor-mance evaluation results while also posing a challenge for DEA (Data Envelopment Analysis) model construction. This paper constructs four DEA models to deal with streaming data combined with traditional statistical data when considering undesirable output. Classic ways of transforming streaming data and LASSO (Least Absolute Shrinkage and Selection Operator) regression are both used for transforming streaming data in the new DEA approach. An empirical study shows the results of dimension reduction of big data and the difference in effi-ciency scores obtained based on them. Also, a robustness analysis illustrates how the number of variables in-fluences the efficiency result. The models presented in this paper are utilized to calculate the environmental efficiency of 252 of China's cities in 2020, considering both statistical data and daily air quality index data. The efficiency results also show a link between efficiency and city size by dividing all cities into five categories.

Automated summary

This paper constructs four DEA models to handle streaming data combined with traditional statistical data considering undesirable output, using classic ways of transforming streaming data and LASSO regression. An empirical study shows the dimension reduction results of big data and the difference in efficiency scores obtained based on them. A robustness analysis illustrates how the number of variables influences the efficiency result. The models are used to calculate the environmental efficiency of 252 Chinese cities in 2020, considering statistical data and daily air quality index data, and the efficiency results show a link between efficiency and city size.