2.1. Obstacles and Challenges in the Process of Greening of Traditional Suppliers

Financial obstacles hinder the greening of downstream supply enterprises, as suppliers feel customers demand new environmental standards but resist bearing the costs. To address this, real estate companies can share costs through rewards or commitments, and provide early guidance, allowing suppliers adequate time to meet project requirements ^{[11, 12]}. To meet the environmental mandates of green supply chains, downstream suppliers must develop new products, establish management systems, and train personnel, all of which require substantial lead time. This places higher demands on real estate enterprises’ procurement management and strategic planning. Suppliers often exhibit low cooperation due to insufficient environmental awareness or differing corporate goals when faced with new green requirements. To address this, real estate companies should consider corporate culture when selecting suppliers and guide them on corporate values after onboarding to ensure alignment and foster cooperation.

The lack of systematic training in supply enterprises, poor communication between senior management and frontline staff, and low baseline qualifications of employees can hinder meeting customer needs. Improving workforce quality can lead to win-win collaborations with clients. This issue is particularly acute for suppliers of materials and equipment, who face challenges in meeting demands for innovative products or substituting hazardous materials with safer alternatives. Suppliers are also pressured to invest in R&D for new product development. Currently, green construction standards are evolving slowly, and most companies follow internal guidelines that lack broad industry impact. Additionally, weak intellectual property protections pose risks to suppliers who invest heavily in eco-friendly products, leaving them vulnerable to competitors. Regulatory authorities should implement strict patent protection mechanisms to safeguard these innovations.

2.2. Concept Definition of Green Supply Chain

Traditional supply chain management underscores stable, efficient win-win partnerships, aiming to boost profits and secure market share. Different from the traditional supply chain, green supply chain not only in order to achieve enterprise economic benefits, but also adds two new goals of social benefits and environmental benefits ^{[13, 14]}. However, in actual operation, for a specific subject in a specific period, the above three profit indicators are very likely to contradict each other, and the goals cannot be unified. There are often management behaviors that damage the environment in order to improve economic benefits and achieve profit goals; Or one’s pursuit of environmental benefits, ignoring the market environment, and bringing profit losses to the enterprise. Green supply chain management considers the interests of these three aspects at the same time, and realizes the common development of the three aspects through advanced concepts and management methods. The update rules of the objective function and gradient descent are shown in Eqs. (3) and (4).

The emergence of traditional SCM reflects market management evolution, stemming from enterprise group self-organization. In green SCM, the stakeholder composition is more intricate, influenced by government oversight, consumer preferences, and regulatory constraints. Therefore, it can be said that the main body composition and main body behavior of the green supply chain are more complex than those of the traditional supply chain, and the behavior goals of the main body on the chain are also more diverse ^{[15, 16]}. Regulatory departments, as the guides and promoters of greening, should formulate feasible and effective regulations. Strict regulations aim to benefit consumers most directly. To promote green SCMs, consumers should adopt sustainable consumption habits.

Fig. 1 has showed the construction of intelligent learning mode. Product lifecycles span from planning to waste recycling. Traditional SCM focuses on the first five stages, while green SCM encompasses the entire cycle, from sourcing to waste recycling, applying green concepts to minimize energy use and environmental impact.

Fig. 1. Construction of intelligent learning mode.

The cornerstone of traditional supply chain management revolves around the dynamics between the various entities within the chain. However, the green supply chain extends beyond these parameters, incorporating technology as a pivotal factor. Enterprises must embrace a green strategy and foster a green corporate culture, ensuring that these principles permeate through all levels of execution, driven by employee advocacy. Given the substantial costs borne by downstream enterprises, the green supply chain strives for a more stable, thematic relationship, prioritizing seamless information exchange and pursuing mutual growth and prosperity. In terms of technical requirements, the development of new green products, the selection of environmentally friendly alternative materials, and the greening of logistics have all brought challenges to green supply chain management.

2.3. Formation and Evolution of Green Supply Chain

Facing the pressure brought by the depletion of natural resources and environmental pollution, and in order to meet the internal needs of economic development to the greatest extent, exploring a new development path has become the only way for social progress. As a self-organizing complex adaptive system, green supply chain is affected by internal power and external pressure while the system itself evolves, self-organizes, self-similarities and self-adapts. External pressure such as: Internal power such as enterprise economic benefits, marketing effects, social benefits, environmental benefits, environmental regulations, consumer demand, etc. According to the iterative theory of inductive technical organization, the formation diagram of green supply chain is drawn, as shown in Fig. 2.

Fig. 2. The formation chart of the green supply chain.

Green supply chains face external pressures from environmental regulations, policy standards, public opinion, and consumer demand. Countries worldwide, including China, have introduced regulations like ISO14000, the EU WEEE Directive, and China’s “Emerging Energy Industry Development Plan” to promote energy conservation and emission reduction. Media coverage of environmental pollution and toxic material usage has heightened public awareness, pushing both companies and consumers to prioritize environmental protection ^{[17, 18]}. Public demand for eco-friendly products has grown as consumers increasingly value sustainability alongside price and quality. Many are willing to pay a “green premium” for products from socially responsible companies with strong brand reputations, ensuring market demand for green supply chain products.

Fig. 3 compares supply chain complexity before and after optimization, highlighting how green supply chains improve raw material efficiency, reduce waste, and lower costs for hazardous material treatment. This internal drive allows companies to capture new markets, enhance their brand reputation, and achieve both economic benefits and cost savings, unlike traditional supply chains with lower recycling efficiency and higher waste disposal costs.

Fig. 3. Comparison between network complexity of supply chain and before and after optimization.

Green supply chains offer enterprises early market access, new profit opportunities, and competitive differentiation by promoting eco-friendly products and services, enhancing brand image, and building intangible assets. They improve product safety by using non-polluting materials, conserve energy, and create jobs through recycling and regeneration activities. Environmentally, they increase resource efficiency and reduce the need for hazardous material recycling. These chains function as self-powered systems where participants manage resources collaboratively, evolving through market, social, and environmental interactions. Their self-regulation ensures stability even when participants deviate from norms. Government support is vital for transitioning to green supply chains, with environmental laws, innovation incentives, and industrial plans fostering a smooth shift from traditional models. In summary, green supply chains reduce costs, improve efficiency, and strengthen brand reputation while contributing to social, environmental, and economic goals. Their adaptability and self-organization ensure resilience and continuous improvement, with government policies playing a critical role in enabling this sustainable transformation.

Fig. 4 shows the learning dynamics of the present approach and other baseline models. The real estate development process, including planning, land acquisition, approval, sales, and property management, integrates green supply chain principles. The green supply chain in real estate is a network involving developers, consulting units, suppliers, contractors, agencies, property companies, buyers, waste recyclers, and regulatory bodies. Real estate firms, leveraging green production concepts and innovative supply chain management, oversee logistics, capital flows, and information streams to drive social, environmental, and economic sustainability. Green design planning focuses on early-stage ecological design, aiming to create environmentally friendly buildings and communities. This includes minimizing environmental impact during development, promoting the use of renewable and biodegradable materials, and ensuring efficient recycling of waste. Green procurement emphasizes selecting suppliers based on green criteria, encouraging bulk purchasing, and helping suppliers develop green technologies and materials. Green logistics requires low-energy, low-consumption transportation methods, improving efficiency in warehousing and handling.

Fig. 4. Learning dynamics of the present approach and of other baseline models.

Green construction prioritizes environmentally friendly techniques, improving building material efficiency and reducing dust and noise pollution. Safe and civilized operations are emphasized to protect the surrounding environment. Green property management focuses on extending building life through maintenance, optimizing space, saving water and energy, and enhancing resource recycling for sustainable development across economic, technological, and conceptual dimensions. This also supports reverse logistics. Green marketing promotes environmentally friendly features in real estate projects, utilizing reusable and electronic marketing materials to emphasize sustainability. These practices ensure that green supply chains in real estate not only meet environmental goals but also contribute to long-term economic and social benefits.

3.1. Indicator System Construction Principles and Operation Procedures

Choosing a suitable supplier is the premise of constructing a harmonious, stable and efficient supply chain system, and also the premise of effective operation of the supply chain; The selection of general suppliers emphasizes improving quality and reducing cost; Under the supply chain thinking, the choice of partners increases the consideration of the intimacy with the supply chain, requiring suppliers to combine the requirements of the supply chain in the development strategy and high-level value orientation of the enterprise on the premise that their products meet the quality and cost; Under the green supply chain system, the supplier’s social and environmental benefits have also become the focus of the index system.

A comprehensive yet concise principle-based index system serves as the programmatic blueprint for supplier selection, necessitating a system that meticulously captures the full spectrum of supplier fundamentals. This encompasses not merely rigid, capacity-driven metrics like enterprise qualifications, output value, and product pricing, but also incorporates flexible indicators reflective of cooperation willingness and strategic alignment, thereby offering a holistic view of supplier potential. At the same time, the indicator system should also be concise, with clear layers and clear logic. And the index system should start from the strategic goals of the supply chain, pointing to clear, no duplication, and no missing items. Definparameter definition parameters and EE model formulas as Eqs. (5) and (6).

The established system should have strong operability, clear index names, easy data collection, and large-scale processing as a whole. The calculation method of each index should be easy to learn while ensuring scientific nature, so that the evaluation process is convenient to form a standardized and standardized work flow, which is beneficial to the improvement of work efficiency, reduces the management cost generated in the evaluation process, and avoids corruption and fraud. The green purchase index calculation formula and the green product design evaluation criteria are shown in Eqs. (7) and (8).

Objective principle of comparability green supply chain partner selection system. The objective principle of comparability means that the selected index system is targeted, can reflect the actual level of suppliers three-dimensionally and objectively, and at the same time take into account the actual situation of the enterprise as much as possible ^[20]. For flexible indicators that are not easy to quantify, each indicator should also be quantified as much as possible to avoid the accuracy of the results affected by subjective factors by the assessors, and to ensure that the evaluation conditions in different places, different assessors, and different periods can be compared horizontally.

The supply chain itself has different degrees of sensitivity requirements for the participants, and sensitivity also guarantees the self-adjustment and self-adaptation capabilities of the supply chain. The same evaluation indicators should also be distinguished from different market environments, industry categories, and industrial chain roles. The same enterprise will also choose different evaluation systems to choose its partners under different development stages and different strategies. Therefore, the supply chain evaluation system should be dynamically adjusted according to the characteristics of the enterprise itself and expand and evolve on the basis of many common indicators to continuously adapt to the new pattern.

3.2. Construction of Preliminary Index System

As early as the 1990s, the International Supply Chain Association SCC proposed the Scor model to provide operational reference for enterprises using supply chains. This model is suitable for supply chain operation forms of various industries ^[21]. This chapter initiates by adopting the SCOR model, specifically tailored to the realm of material and equipment suppliers, and proceeds to delineate the essential attributes that these suppliers ought to possess within the context of the real estate green supply chain. Through a methodical, layered decomposition of the model, it lays the groundwork for constructing a robust index system in subsequent sections, offering invaluable reference and support for this endeavor.

Fig. 5 shows green supply chain management efficiency improvement trend. The Scor model is mainly divided into three levels, namely the process level, the configuration level, and the execution level. Specifically, each level will refine the content on the basis of the previous level. This section needs to deduce the characteristics that qualified material and equipment suppliers in the green supply chain need to have to provide a reference for the establishment of the index system in the following text to ensure the comprehensiveness of the index system. Therefore, from the establishment of the SCOR model to the second configuration layer, the third the execution layer will be refined according to the specific characteristics of the enterprise. The formula for optimizing the green supply chain performance and the green supply chain decision model combined with the hyperparameter optimization are shown in Eqs. (9) and (10).

The second layer of configuration level, the third layer of implementation level.

Fig. 5. Green supply chain management efficiency improvement trend.

3.2.3 Green manufacturing

The key to the whole process is to do a good job in the green control of this process. The green manufacturing process is divided into: production process, warehousing process and packaging process for discussion.

Fig. 6 shows acceptance rate of green practices in supply chain management. First, do a good job in equipment maintenance, improve production technology, improve energy efficiency, and reduce energy consumption per unit of product; Second, the work of rationally recycling waste water and effectively treating the industrial “three wastes” produced in the production process; Third, build a reasonable modularization of production components, reduce leftovers generated in the production process, and establish a recycling system. The formulas involved in the statistical analysis formula of the empirical data and the interpretation and discussion of the empirical results are shown in Eqs. (15) and (16).

Fig. 6. Acceptance rate of green practices in supply chain management.

(15)

$R(\theta, x)^{l,l} = \| x_l - x_l^* \|_2^2,$

(16)

$R(\theta, x, \rho^{p,q})^{p,q} = \sum_{c \in C} \rho_c^{p,q}(x_q^*) \| (x_p - x_q^*)_c \|_2^2.$

Warehousing process, warehousing management mainly includes two aspects, one is that the backlog of warehousing will affect the cash flow of enterprises and increase supply risks; Second, many materials in building materials, if not properly kept, will greatly reduce the performance of use. Enterprises must strategically plan their production schedules, allocate warehouse space efficiently, enhance the maintenance of inventory goods, and strive to eliminate unnecessary waste.

During the packaging process, they should prioritize the use of environmentally friendly, biodegradable, and recyclable packaging materials while satisfying the logistical requirements of the products, thus preventing waste generated by excessive packaging.

In order to achieve the above goals, enterprises should have the following characteristics at the execution level:

Strictly implement the ISO14000 system;

Reasonable design of production modulus;

Reasonable recycling of corner materials;

Adopt advanced technology, maintain equipment regularly, improve production efficiency, and reduce energy consumption per unit product;

Strictly deal with “three industrial wastes”;

Establishing scientific storage management system;

Green packaging.

4.1. Applicability Analysis of BP Neural Network

4.1.4 Self-learning and self-adaptability

In the BP neural network, by bringing in the collected data samples for trial calculation training, it can be used directly after the training is completed. And we can make full use of the existing supplier evaluation database of real estate enterprises to supplement samples, and constantly improve the network model, optimize the evaluation results ^[23]. Adjustments to the number of input neural units are solely necessitated by changes in the evaluation index system, be it an expansion or reduction. In cases where a particular index’s weight is emphasized, incorporating new samples for retraining and subsequent utilization becomes a straightforward solution.

Fig. 7 shows green supply chain management cost savings situation. To sum up, the above two characteristics of BP neural network can well fit the criteria for selecting evaluation methods in this paper. BP neural network has a high simulation learning ability, the use of the real estate enterprise’s own supplier history evaluation database can better achieve the training effect of the network. Therefore, the paper will use the BP neural network method to evaluate the selection of real estate green supply chain partners.

Fig. 7. Green supply chain management cost savings situation.

4.2. Establishment of BP Neural Network for Partner Selection and Evaluation

Utilizing the fundamental BP neural network training methodology, specifically the BP algorithm, this paper exemplifies the material supplier selection evaluation system previously established. By integrating the construction principles of BP neural networks, we will construct a comprehensive system for selecting material and equipment suppliers. Bring in the known training data, train and learn the model, and know that the error value $E$ meets the requirements and is less than the allowable error $\varepsilon$, then complete the learning and training, and then calculate the final result according to the model; If the error requirement is not met, the back calculation propagation will be carried out until the desired value is reached. Ask will follow the known process step by step.

Fig. 8 shows green supply chain management efficiency improvement trend. The initial step involves determining the optimal number of neurons in each layer: the input layer, the output layer, and the hidden layer. This essentially involves deciding the appropriate number of nodes for each respective layer. Fig. 9 shows green supply chain risk assessment.

Fig. 8. Green supply chain management efficiency improvement trend.

Fig. 9. Green supply chain risk assessment.

4.2.2 Number of Hidden Layers in the Middle

When constructing a BP neural network model, the initial step involves determining the number of hidden layers. Notably, the hidden layer does not serve a specific, discrete function but primarily acts as a conduit for data transfer. The primary influence of the number of hidden layers lies in shaping the overall data processing capacity of the network. However, this decision also has implications for the network’s computational efficiency, potentially leading to prolonged training times and increased budgetary considerations ^[24]. Some scholars have proved that continuous functions in any closed interval can be simulated by a BP neural network with only one hidden layer, in which each node can use S-type functions, and can be classified by most decisions ^[25]. According to the above analysis, it is determined that the hidden layer number of the real estate green supply chain material and equipment supplier selection evaluation model established in this study is determined to be one layer. Fig. 10 shows effect of the hyperparameter optimization on the model performance. The number of neural nodes in the output layer should be determined according to the purpose of using the model and the expected effect. This chapter needs to build the real estate green supply chain material equipment supplier selection model, the purpose of the model is to supply the core enterprise of the supply chain real estate enterprises; to provide an intuitive, direct reference data, data can intuitively reflect the comprehensive level of a supplier. Therefore, an output neural node is set up in this study. To sum up, this study finally set up a 24 × 13 × 1 three-layer BP neural network model.

Fig. 10. Effect of the hyperparameter optimization on the model performance.

Fig. 11 shows model convergence rate under different optimization algorithms. DWT can decompose signals into wavelet function combinations to extract the time-frequency characteristics of supply chain data. DCT converts signals into combined cosine functions, removing redundant information and retaining important features. Both are widely used in the fields such as signal processing. The role of this step is to initialize the weights and deviations of each layer, preventing the neural network from being saturated with large weighted outputs. In general, the connection weight and value B are set in any decimal number in the (−1,1) interval, so as to ensure that the connection weight and threshold value of each layer are assigned and initialized with small random numbers, so as to achieve the purpose of using the BP neural network model. The Softmax function and Swish functions are shown in Eqs. (17) and (18).

Fig. 11. Model convergence rate under different optimization algorithms.

(17)

$y_j = \frac{e^{z_j}}{\sum_{k=1}^K e^{z_k}},$

(18)

$f(x) = x \cdot \operatorname{sigmoid}(\beta x) = \frac{x}{1 + e^{-\beta x}}.$

In general, the commonly used transfer functions include logarithmic function sigmoid, linear function purelin, tangent function simoid, etc. This model is specifically tailored for the green supply chain supplier selection process within the realm of real estate enterprises. In the construction of this model, the sigmoid function has been chosen for both the hidden layer and the output layer, underscoring its suitability for this particular application. The index data corresponding to all alternative enterprises are typed in the input layer, passed to the output layer through the above two functions, and finally calculated into data. The output vectors of the hidden layer and the output vectors of the output layer are shown in Eqs. (19) and (20).

(19)

$Y = f_1\left(\sum_i w_{ii} x_i + b_i\right),$

(20)

$Z = f_2\left(\sum_j w_{j1} + b_1\right).$

The training and application of a BP neural network involve two key steps. Initially, the neural network is trained using normalized sample data and suitable parameters until the output error falls below a preset threshold. Subsequently, test data is fed into the trained network to obtain final output values. Post-processing with normalization functions facilitates comparisons between evaluation objects, enabling the selection of suitable suppliers.