Volume 15, Number 1, June 2025
| Integration of Machine Learning Algorithms with Data Envelopment Analysis for Evaluating the ESG Efficiency: A Case Study of Taiwan Semiconductor Industry |
Yi-Ting Peng 1, Chun-Yueh Lin 2*
Abstract
The semiconductor industry serves as the cornerstone of modern technological advancements, driving innovation across various sectors. However, growing concerns regarding its environmental and social impact have necessitated a shift toward sustainable development. This study focuses on upstream semiconductor companies in Taiwan, employing Data Envelopment Analysis (DEA) with Slack-Based Measure (SBM) to assess ESG efficiency and integrating machine learning through decision tree (DT) algorithms to identify key determinants of efficiency performance. The proposed integrated DEA-DT model evaluates 38 Taiwan upstream semiconductor companies from 2021-2023, using three input variables (carbon emissions per revenue, average salary, R&D expenditures) and three output variables (carbon improvement index, parental leave retention rate, ROE). Results reveal that 11 companies achieved ESG efficiently, while scale emerges as the most critical factor influencing efficiency, followed by market capitalization, company age, and employee turnover rate.
The model demonstrates strong predictive capability with 92.1% accuracy, and 75% precision. Seven decision rules are generated to guide efficiency improvement strategies. This integrated approach addresses existing gaps in ESG efficiency evaluation by combining quantitative assessment with interpretable factor analysis, providing actionable insights for semiconductor companies' sustainable development strategies and offering a replicable framework for cross-industry ESG efficiency evaluation.
Keywords: ESG efficiency evaluation, DEA, Decision tree, Machine learning, Semiconductor Industry
JEL Classification: C44, Q56
1 Department of Public Finance and Tax Administration, National Taipei University of Business
2 Department of Public Finance and Tax Administration, National Taipei University of Business