Predicting the factors influencing construction enterprises’ adoption of green development behaviors using artificial neural network

The construction industry significantly impacts the global economy and environment. High energy consumption and resource depletion lead to severe environmental issues, including air pollution and resource degradation. In 2019, global construction CO2 emissions reached approximately 1 billion tons, representing 28% of total energy-related carbon emissions. China, facing increasingly severe environmental challenges, necessitates active green development behaviors from its construction enterprises, which constitute a substantial part of the national economy (25% of total national economic output in 2019, but also 30% of national carbon emissions). While research exists on green development in other sectors (manufacturing, chemicals), the construction industry's specific influencing factors remain understudied. This paper addresses this gap by exploring the factors influencing green development behavior adoption in Chinese construction enterprises. The study uses the TOE framework to construct an index system, analyzing the impact of technological, organizational, and environmental factors. The novelty lies in the application of an ANN model to predict these influences, offering practical suggestions for promoting green development behaviors within the construction industry and providing guidance for government policy.

The paper reviews existing literature on green development and green development behaviors within enterprises. It discusses various definitions of green development, emphasizing its importance for sustainable development. The review analyzes previous studies on green behaviors such as green purchasing, green procurement, green consumption, and green management. The literature highlights the TOE framework as a valuable theoretical lens for studying technological innovation and adoption. This framework emphasizes the interplay of technological, organizational, and environmental factors. The paper reviews several applications of the TOE framework in diverse contexts (information technology, RFID technology, online financial transparency) and shows how technological, organizational and environmental factors influence the adoption of green technology innovation. The review concludes that while there's extensive research on industrial enterprises, the construction industry's specific green development behaviors and their influencing factors require further investigation.

The study utilizes data from the National Bureau of Statistics of China (2000-2020) and other sources like the China National Intellectual Property Administration and CNKI. The data includes indicators related to technological factors (construction machinery, equipment, R&D expenditure, technological learning rates), organizational factors (enterprise size, resources), and environmental factors (market share, government policies, enterprise demand). Technological learning rates are calculated using a dynamic two-factor measurement model, considering cost per unit area, cumulative construction area, and R&D expenditure. Missing data is handled using the average value method. An artificial neural network (ANN) model, specifically a multi-layer perceptron (MLP), is employed to predict the influence of these factors on the adoption of green development behaviors (measured by the number of patents related to green building technologies). IBM SPSS Modeler 18.0 is used for model building and analysis. The data is partitioned into training and testing datasets (60:40 ratio selected for optimal accuracy), and model accuracy is assessed. Pearson correlation analysis is performed to examine the relationships among variables.

The ANN model reveals that the total output value of the construction industry (market share), R&D expenditure, and the number of construction enterprises (enterprise size) are the three most influential factors on the adoption of green development behaviors. The model's accuracy is 99.6% (60:40 training-testing split). The analysis shows a significant positive correlation (0.903) between the actual and predicted number of patents. The number of patents shows an upward trend over time, indicating increasing adoption of green development behaviors. The study also shows the relative importance of various factors, with market share having the strongest impact (0.18), followed by R&D expenditure (0.17) and number of enterprises (0.16). Other factors such as total output value of real estate, total power of machinery, employee number, building area, urbanization level, number of policies, and total profit also have some impact, but to a lesser degree.

The findings confirm the significance of the TOE framework in explaining the adoption of green development behaviors in the construction industry. The strong influence of market share highlights the crucial role of environmental factors. The importance of R&D expenditure underscores the need for technological advancements and innovation. The impact of enterprise size emphasizes the benefits of scale and resources. These findings suggest that policies promoting market growth, technology advancements, and enterprise consolidation can contribute to increased adoption of green practices within the construction sector. The study contributes to the understanding of green development behavior, specifically within the relatively under-researched construction industry context. The use of the ANN model provides a quantitative approach to predicting the impact of various factors, allowing for more targeted interventions.

This study demonstrates the effectiveness of using an ANN to predict factors influencing the adoption of green development behaviors in Chinese construction enterprises. Market share, R&D expenditure, and enterprise size emerged as the most critical factors. The upward trend in green development behavior is evident. Future research could address the limitations of the study, such as using more sophisticated missing data techniques and including qualitative aspects. Expanding the geographical scope to other countries and applying other predictive modeling techniques (Bayesian networks, decision trees) could enhance the generalizability and robustness of the findings. The results provide valuable insights for both enterprises and policymakers to promote more sustainable practices within the construction industry.

The study's limitations include the use of the average value method for handling missing data, which might affect the accuracy of the model. The analysis is limited to quantitative data, neglecting qualitative factors such as employee attitudes and perceptions. The study focuses solely on China's construction enterprises, limiting the generalizability of the findings to other national contexts. Finally, only an ANN was used. The use of alternative predictive modeling techniques could provide valuable comparative insights.