Developing a behavioral model to predict eco-friendly packaging use among millennials

1 Introduction

Waste management remains a significant challenge in several countries. In this context, Indonesia is ranked among the top five global waste producers, along with India, China, and Nigeria, generating approximately 17 million tons annually. However, 60% is effectively managed, and the remaining 40% are in landfills and polluted environments [1,2,3]. Inadequate waste management from marine, terrestrial, aquatic, and atmospheric pollution causes substantial health consequences for humans. A prominent example is the increased vulnerability of scavengers, whose livelihoods necessitate direct contact with waste [4,5,6]. Population growth contributes to an increased volume of waste, diminishing the natural quality of the environment [7]. The degradation manifests in various detrimental effects, including mass mortality among aquatic organisms [8], as well as the obstruction and malodor of sewage systems [9]. This facilitates the proliferation of disease vectors such as mosquitoes in developing nations [10] and reduces aeration and water percolation, leading to decreased agricultural productivity [11,12].

In response to the challenge, Indonesia is exploring a range of strategic interventions, such as the Waste Reduction Program and the Source-Based Waste Management Policy. However, obstacles to the low recycling rate prevent appropriate results. The current data show that Indonesia is experiencing an increase in the volume of plastic waste, threatening marine ecosystems and human health [13]. Therefore, more innovative solutions are urgently needed today. The solutions include environmentally friendly packaging such as biodegradable-based packaging or recycled materials.

Other strategies are followed to reduce the negative impact of conventional plastic packaging [14], such as edible films made from sweet potato starch treated with heat and water to present environmentally friendly alternatives. As an alternative, sachet-based packaging using a water-soluble film offers a promising solution for reducing daily waste production and improving sustainable practices [15,16]. In addition, automatic sorting systems and community education programs can be improved to make waste management more effective and efficient. Effective waste accumulation management is important for pollution reduction, contributing to a global effort to minimize the ecological footprint [17]. Technological innovation facilitates the creation and implementation of various solutions, such as smart bins enhanced by EfficientNet for real-time household waste classification, promoting economical recycling and environmental protection [18].

In resource-dependent countries, severe socio-economic gap is reported due to environmental crises in waste management and the intrinsic depletion of natural resources. Since the beginning of the twentieth century, there has been an increase in global temperature due to greenhouse gas emissions, the burning of fossil fuels carried out in support of human activities, and the clearing of forest land. In Indonesia, climate change poses a threat to biodiversity and species extinction [19]. The increase in environmental awareness is not substantial because there is a gap between the intentions of the millennial generation and actual behavior in implementing environmentally friendly packaging. Among the obstacles to the implementation are price, product availability, convenience, and inadequate information [20]. The solutions required to bridge the gaps and increase the implementation of green practices are education, socialization, economic incentives, and product design innovation [21,22].

The prediction model of millennial behavior using eco-friendly packaging as an indicator of environmentally friendly behavior has not been fully developed. Previous research explored green consumer behavior, but few addressed the challenges and predictors of eco-friendly packaging choices among millennials. Therefore, this research aimed to fill the gap using the Adaptive Neuro-Fuzzy Inference System (ANFIS) to predict millennial behavior regarding eco-friendly packaging. ANFIS offers valuable insights into millennial consumer behavior regarding eco-friendly packaging and the potential to mitigate plastic waste. This model contributes significantly to sustainable, environmentally friendly research by offering a more accurate and adaptive behavior prediction than traditional counterparts. The novelty lies in integrating ANFIS with Partial Least Squares Structural Equation Modeling (PLS-SEM), allowing for a deeper understanding of the relationships between different variables influencing consumer behavior.

Planned Behavior Theory (PBT) is reported as a promising framework to address the challenges. Behavior plays an important role in this theory by integrating economic, social, and environmental factors. Attitude, subjective norms, and perceived behavioral control (PBC) in individuals affect the Sustainable Development Goals (SDGs) [23,24]. Integrating the SDGs with other theoretical models, such as the Value-Belief-Norm (VBN) and the Norm Activation Model (NAM), increases the understanding of intentions to engage in eco-friendly practices [25,26]. Furthermore, factors related to travel intentions, specifically in the post-COVID-19 context, are influenced by perceived economic stability [27]. Social and environmental factors integrated with the SDGs enrich behavioral analysis. Social aspects, such as social capital, community participation, and use of social media, play an important role in forming pro-environmental intentions [28,29]. The level of green consumer behavior was effectively identified in previous research, where a neural network achieved an impressive recognition accuracy of 82% in classifying behavior as low, medium, or high. The results support the hypothesis that green environments play an important role in explaining green behavior [30]. However, this research only focuses on classification without providing an in-depth analysis of the underlying relationships between the variables driving consumer choice. The emphasis on eco-friendly packaging and risk management during the pandemic reports the need for a comprehensive method considering various influencing factors to understand behavior more deeply [27,31].

Artificial intelligence (AI) has been developed as a powerful tool for predicting consumer behavior, particularly in the context of eco-friendly packaging among millennials [32,33,34]. An AI model can identify patterns and trends in data, showing future pro-environmental behavior [35]. Machine and deep learning within AI analyze vast data from consumer surveys, social media, and online shopping behavior to predict millennials’ preferences for eco-friendly packaging based on environmental awareness and social preferences [36]. Research shows that social media significantly influences consumer purchasing behavior, with platforms such as Facebook, Twitter, and Instagram providing valuable data for behavior prediction [37].

A key innovation is integrating ANFIS with PLS-SEM to enhance consumer behavior prediction. Even though previous research used an AI-based model to classify behavior, the novelty was in the ability to model complex relationships among latent constructs and improve prediction accuracy using PLS-SEM and ANFIS, respectively. ANFIS offers the advantage of incorporating fuzzy logic to capture uncertainty and nonlinearity in consumer decision-making processes. This is particularly useful in predicting eco-friendly behavior, where human choices are influenced by multiple, conflicting, socio-economic, and psychological factors.

Prediction accuracy is enhanced beyond traditional statistical and AI methods by providing an adaptive and interpretable model. Furthermore, integrating PLS-SEM allows for a more detailed exploration of causal relationships since key behavioral determinants are systematically identified and validated. This method improves the robustness of behavioral insights and provides more actionable recommendations for policymakers and businesses to promote eco-friendly packaging adoption. The results are expected to contribute significantly to theoretical advancements in consumer behavior research and practical applications for sustainable marketing strategies. The objectives are to analyze a behavioral model of millennial generation in using Green Packaging and the factors influencing plastic waste reduction behavior.

2 Methodology

This research was conducted in Pangandaran Tourism, which had great tourism potential but produced a large amount of waste. The volume of waste on Pangandaran Beach reached 70 tons daily, increasing to 200 tons during holidays. Even though millennials are increasingly aware of the importance of being environmentally friendly, behavior in using environmentally friendly packaging is not consistent.

Before data collection, all participants were provided with detailed information regarding the purpose, procedures, potential risks, and benefits of the research. Participation was voluntary, and participants could withdraw at any time without consequences. The research protocol and consent procedures were reviewed and approved by the Director of the Department of Tourism and Culture. A pre-test was conducted to assess the validity and reliability of the instrument before distributing the questionnaire. From the initial pool of 400 respondents, 302 were determined to be valid. The research was carried out from March to June 2024 with a survey. Data were collected both online through Google Forms and offline using paper questionnaires. In addition, a total of 50 tourists visiting Pangandaran were included in the offline sample, which was obtained through systematic random sampling. The results generated by offline and online respondent sampling were validated and compared. The respondents were environmentally friendly packaging users or individuals interested in using it. The questionnaire consisted of variables Y and X. In this context, variable Y was used to determine plastic waste reduction behavior influenced by factor X, such as social, economic, tourist attraction, attitude toward behavior (ATB), subjective norm, and PBC. The breakdown of variables and the corresponding indicators comprised social factors (X11–X17), economy (X21–X27), tourist attraction (X61–X67), ATB (X31–X34), subjective norm (X41–X48), and PBC (X51–X56). The variable of plastic waste reduction behavior (Y) consisted of six questionnaires (Y1–Y6). Each variable had indicators, which used a Likert scale, as shown in Appendix 1.

Based on the TPB, attitude was crucial in shaping individuals’ engagement in plastic waste reduction [38]. Research showed that perceptions of plastic waste influenced the attitude toward plastic consumption, leading to environmentally responsible behavior (H1). Additionally, subjective norms, representing social pressure to adopt sustainable practices, significantly impact individuals’ commitment to reducing plastic waste. The perception of strong social support conformed to societal expectations [39,40] (H2).

PBC described individuals’ confidence in the ability to reduce plastic waste. Individuals with a high sense of control over behavior were more inclined to act in environmentally responsible ways [38,40] (H3). Economic factors also played a significant role in shaping plastic waste reduction behavior. Financial incentives, cost considerations, and perceived economic benefits influenced the willingness to engage in sustainable practices [39,41] (H4). Furthermore, tourist attractions could promote pro-environmental behavior. Destinations implementing environmental policies mitigated ecological degradation and inspired visitors to adopt sustainable practices, such as reducing plastic waste. Destination attractiveness was frequently correlated with its dedication to sustainability [39,41] (H5).

The TPB provided a robust framework for understanding plastic waste reduction behavior. Attitude, subjective norms, PBC, economic considerations, and the influence of tourist attractions collectively shaped the intentions and actions toward sustainable practices. Future research and policy interventions could integrate the factors to promote environmentally responsible behavior effectively.

This research adopted an ANFIS and PLS-SEM using SmartPLS 4.0. PLS-SEM evaluated a complex model, tested the validity and reliability, examined relationships between constructs, and validated a model through goodness of fit (GoF) [42,43,38,39]. Based on the formulated hypothesis, this research used inferential statistical data analysis with the PLS-SEM method through SmartPLS 3.0 software.

PLS-SEM was applied through two main stages, namely (1) evaluation of the measurement model to assess the validity and reliability of latent constructs. At this stage, convergent and discriminant validity tests were conducted with the measurement of construct reliability using indicators such as Cronbach’s Alpha and Composite Reliability. (2) Evaluation of the structural model tested the hypothesized relationships between variables and identified significant factors influencing millennials’ intention to use eco-friendly packaging. This stage comprised assessing R-square (R ²) values, conducting path analysis, and testing the significance of path coefficients through bootstrapping.

PLS-SEM was selected as the analytical method due to the ability to handle a complex model with multiple latent variables. Therefore, the method was suitable for the research objectives, which focused on predicting the environmentally friendly consumer behavior. PLS-SEM also tested causal relationships between variables that were not thoroughly explored. The criteria for a medium-to-large sample size were met with a sample size of 352 respondents, ensuring adequate statistical power and robustness of the analysis. Therefore, PLS-SEM was relevant and essential for uncovering behavioral mechanisms underlying eco-friendly packaging usage among millennials.

ANFIS was applied to predict plastic waste levels using four key input variables, namely ATB, economy, social factors, and tourist attraction. These variables represented the psychological, economic, and environmental influences on waste generation. The dataset was also subjected to preprocessing to ensure consistent scaling. An initial plastic waste variable on a 1–5 scale might be recoded into binary or multi-class categories for classification purposes.

The Fuzzy Inference System (FIS) within ANFIS defined membership functions for each input variable and established fuzzy rules to describe the relationship with plastic waste levels. ANFIS combined backpropagation and least squares estimation to optimize membership functions and reduce prediction error. The performance of model was evaluated using metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R ² score to determine the accuracy and reliability. Additionally, ANFIS offered several significant advantages, including the ability to capture nonlinear relationships, improve prediction accuracy, and generate fuzzy rules interpreted as factors affecting plastic waste. The research developed an intelligent prediction model by providing insights into waste management behavior, helping policymakers to design effective business strategies, such as promoting eco-friendly practices and reducing plastic waste. A total of five layers were used to build a model, each containing multiple nodes. Adaptive nodes were denoted by a box representing a set of customizable parameters on this node. Meanwhile, fixed nodes were denoted by a circle representing a set of parameters in the model. The simple ANFIS architecture used two variables (x and y) as input and one output. The first layer converted the input into a fuzzy set through the Member Function (MF). This layer contained adaptive nodes with the following functions.

where x and y are the inputs of node i, A and B are the linguistic labels associated with these nodes, μ(x) and μ(y) are MF. Each node in this layer is fixed, marked with a circle, and labeled Π, with the function of the node multiplied by the input signal to be the output signal.

where O _2,i is the output of Layer 2. The output of the node results in the multiplication of all incoming inputs. Each node in this layer is considered a fixed node, marked with a circle and labeled N. The node function to normalize the firing strength by calculating the ratio of the ith firing strength node to the sum of the firing strength of all rules.

where O_3,i is the output of Layer 3. The magnitude is known as normalized firing strength. This ANFIS layer contains all the adaptive nodes marked with squares. The function of the nodes is as follows:

where f1 and f2 will form the following rules:

Rule 1. If x is A1 and y is B1, THEN f1 = p1x + q1y + r1,

Rule 2. If x is A2 and y is B2, THEN f1 = p2x + q1y + r1,

where P _i , q _i , and r _i are the sets of parameters referred to as the consequent parameters. Each node on this layer is marked with a circle to calculate the overall output and the formula.

ANFIS parameters were adjusted using the Hybrid Learning Algorithm. The combination of the least squares method and the decrease of the backpropagation gradient is a hybrid algorithm for training MF to follow a predetermined training dataset. There are adaptive and consequent parameters broken down into premises in Layers 1 and 4.

The modeling process consisted of the Fuzzy Neuro Designer and Fuzzy Interface System (FIS). In Figure 1, the ANFIS model was developed for classifications that combined training and inference stages. The process started with a training phase using a primary data set derived from questionnaire responses. This dataset was subjected to preprocessing such as cleaning, normalization, and feature selection to be ready for analysis. Furthermore, the dataset was divided into training (70%) and model construction (30%) test data for performance evaluation. The ANFIS prediction model, combining the advantages of fuzzy logic and neural networks, was organized into (1) layer 1 (fuzzification) for converting the input into a fuzzy set, (2) layer 2 (rule) that applies fuzzy rules, (3) layer 3 (normalization) that standardizes the strength of the rule, (4) layer 4 (consequence) that generates output, and (5) layer 5 (aggregation) that consolidates outputs to produce the final prediction.

Figure 1

A workflow systematic approach for data preparation, model training, and classification using an ANFIS-based architecture.

The model used a backpropagation algorithm for iterative parameter adjustment, continuing until a training deadline or satisfactory performance was achieved. The training process included predicting outcomes on datasets and performance assessments with test data to ensure the model’s ability to generalize. After the training process, the model entered the inference stage, where new questionnaire data were applied. The trained model processed the inferential inputs and categorized the data into different levels. The end of the process issued a classification, providing insights based on model’s predictions (Figure 1).

Figure 2

T-value inner and outer model.

1. Informed consent: Informed consent was obtained from each participant.

2. Ethical approval: All procedures using human participants were conducted under the ethical standards of the Universitas Padjadjaran Research Committee, as well as the 1964 Helsinki Declaration and the subsequent amendments or comparable ethical guidelines. The approval was granted under letter number PK.05.02/455a/DISPARBUD/2024.

3 Results and discussion

3.4 ANFIS model

A total of 352 data points were used, with 60 and 40% allocated for training and testing, respectively. Based on the raw data obtained, normalization was carried out to obtain a variable of 0 to 1 for each value. The normalization process ensured that all features contributed equally and prevented dominant calculations due to differences in scale. The dataset consisted of four input variables, namely ATB, economy, social, and tourist attraction, represented as x1, x2, x3, and x4, respectively. In addition, this included one output variable, y, representing plastic waste. These data could be used to develop the prediction model using ANFIS. Figure 3 shows the architecture of an ANFIS model training process. The model predicted millennials’ intentions to adopt environmentally friendly packaging by considering four key variables, namely (1) ATB reflected an individual’s stance on using eco-friendly packaging, (2) Economy referred to affordability and cost factors, (3) Social included the influence of peers, family, and the surrounding environment, and (4) Tourist Attraction examined the contribution to awareness and adoption of sustainable packaging. ANFIS effectively handled nonlinear relationships between the variables, providing more accurate predictions than conventional methods.

Figure 3

(a) and (b) Setting up the ANFIS model training process.

Figure 3 shows the architecture of ANFIS model, consisting of multiple layers and the arrangement of the training process. The ANFIS model received four inputs, namely the variables x1, x2, x3, and x4 processed through the Sugeno inference system. In this context, the system used fuzzy rules to map inputs into outputs, represented by boxes labeled “Estimate_value.” Sugeno inference was simpler and produced outputs such as linear or constant values. The ANFIS prediction model comprised (1) layer 1 (fuzzification) for converting the input into a fuzzy set, (2) layer 2 (rule) for applying fuzzy rules, (3) layer 3 (normalization) for standardizing the strength of the rule, (4) layer 4 (consequence) for generating output, and (5) layer 5 (aggregation) for consolidating outputs to produce the final prediction. This structure showed the connection to a membership function through fuzzy rules, represented by connections between neurons. The training process consisted of adjusting the parameters of membership functions and rules to minimize errors between the predicted output and the actual value. ANFIS combined the capabilities of artificial neural networks in finding data patterns with a fuzzy system capable of handling uncertainty and impressions. The ANFIS model was trained over 50 epochs, with a graph of errors during the training process shown in Figure 4.

Figure 4

Error values during the model training process.

The training error graph of ANFIS model showed a consistent downward trend in error values with increased number of epochs. Therefore, the model successfully learned from the data and gradually reduced prediction errors. The error was reduced sharply at the beginning of training since the model made significant parameter adjustments. After a sharp decline, the error trend decreased, showing that the model was close to convergence. After the training process was complete, the model was tested using training and testing data. Figure 5 presents the prediction results for the datasets. The test was carried out on 212 training and 141 testing datasets. Figure 5(a) and (b) shows that the model has excellent performance on both datasets. The estimated value graph was similar to the actual value since the model had a high prediction accuracy.

Figure 5

Results of testing the model using training data: (a) testing and (b) training.

Figure 6 reports the graph of error values in tests, with training and testing data. The error value for each data point was relatively small. Therefore, the model maintained consistent performance on both datasets and was not overfitted since relevant patterns were captured without memorizing the training set. Table 3 shows the overall performance of the model on the test with both datasets. The ANFIS model was effective in learning from training data and capable of making good generalizations when tested on new data. This model had high accuracy and reliability for future predictions with relatively small errors and graphs similar to actual data. The error reduction graph, the similarity between the predicted and actual values, and the slight error in the test showed that ANFIS could effectively learn data patterns and produce the prediction model with good performance. The combination of fuzzy methods and artificial neural networks allowed the handling of complex and uncertain data. Therefore, the model is ideal for a wide range of predictive applications, requiring adaptation to uncertainty.

Figure 6

Error values during model testing: (a) training and (b) testing.

Table 3

Model performance metrics at the time of testing

	MSE	RMSE	MAE	Accuracy (%)
Training	0.011302	0.106309	0.078779	92.12
Testing	0.021345	0.1461	0.097682	90.23

Based on the performance metrics table, the ANFIS model performed excellently in tests with training and testing data. The low MSE, RMSE, and MAE values in both datasets reported that the model effectively minimized prediction errors. Low MSE and RMSE values showed that patterns were obtained to predict outputs with a small error rate. Furthermore, a low MAE showed that the average prediction error was within a small range. The accuracy level reached 92.12 and 90.23% in training and testing data, respectively. This level of accuracy showed that the built ANFIS model had quite good generalization ability to predict new data with high accuracy. The 1.89% difference in accuracy between the training and testing data suggested that the model did not experience significant overfitting problems. In contrast, this model maintained stable predictive performance.

The low MSE, RMSE, and MAE scores were interpreted since the ANFIS model successfully handled uncertainty and complexity in the data well. This is important when dealing with data containing outliers or noise. The ANFIS model is a highly efficient tool in various prediction applications, with the ability of FIS to handle uncertainty and the strength of artificial neural networks. The performance metrics evaluation results confirmed that ANFIS had high accuracy and minimal errors. Therefore, the model accurately predicted behavior of the millennial generation toward using environmentally friendly plastic. This research offered significant benefits, implications, and impacts in promoting eco-friendly packaging adoption among millennials. Accurate and reliable predictions were provided with high levels of accuracy at 92.12 and 90.23% for training and testing data, respectively, as well as minimal error rates. The inclusion of diverse input variables, such as ATB, economy, social factors, and tourist attractions, enabled a multidimensional understanding of behavioral drivers influencing eco-friendly packaging use. The results have practical implications for policymakers and businesses, offering data-driven insights to design targeted interventions, policies, and marketing strategies. Furthermore, the potential of AI-driven methods was stated in behavioral modeling, suggesting the applicability to other sustainability challenges, such as recycling, renewable energy adoption, and transportation. The impact extended to environmental conservation by promoting eco-friendly packaging to reduce plastic waste, ensuring a broader influence on other demographics, and increasing the adoption of sustainable behavior across generations. The scalable framework and method presented were adapted to various populations and geographies. This enhanced the global applicability and contribution to sustainability efforts worldwide.

The ANFIS model was implemented on a web-based platform to monitor millennials’ intent to use eco-friendly packaging. Key features included a visualization dashboard, predictive analysis, and an early warning system to detect declining interest. The platform integrated real-time data from surveys, social media, and IoT sensors at retail stores and tourist sites. This enabled stakeholders, such as governments, environmental organizations, and businesses, to take proactive measures. Beyond practical applications, the model also supported innovation in eco-friendly packaging through subsidies and tax policies on single-use plastics. An innovative, adaptive solution was provided by offering real-time monitoring and predictive insights for reducing plastic waste, shaping consumer behavior, and informing sustainable policymaking. Academically, existing knowledge was enhanced by refining the analysis method and effectively offering insights into managing independent variables. The results will assist policymakers, businesses, and other stakeholders in developing data-driven strategies. The real-time capabilities improved decision-making and promoted sustainability.

4 Conclusions

In conclusion, developing a comprehensive behavioral model to predict the use of eco-friendly packaging among millennials is essential. This research integrates ANFIS and PLS-SEM to evaluate the relationship between attitude, subjective norms, behavioral control, and social, economic, and tourist attractions, influencing consumers’ decisions to use green packaging. The results show that attitude toward the environment, PBC, social, economic, and tourist attractions play a significant role in promoting millennials’ intentions to use environmentally friendly packaging. Based on the evaluation of independent variables, several recommendations can be made to enhance sustainable millennial behavior regarding eco-friendly packaging use. Attitudes toward the environment and PBC are significant drivers of intention. Therefore, educational campaigns that build positive environmental attitude and reinforce millennials’ confidence in the ability to make sustainable choices are essential. Engaging videos on social media through influencers who resonate with millennials can improve the reach and effectiveness of the campaigns. The establishment of communities focused on peer support and the promotion of eco-friendly behavior will further enhance the effectiveness of these initiatives.

Social, economic, and tourist attraction factors have a moderate influence. In this context, stakeholders, including businesses, tourism boards, and local governments, should implement policies that incentivize the use of eco-friendly packaging. For instance, offering financial incentives such as discounts for consumers who choose eco-friendly packaging or promoting sustainable practices in tourist destinations encourages adoption. Policy measures can include the provision of subsidies or tax reductions to businesses in tourist-heavy regions. The research reports that subjective norms have a relatively lower effect on actual behavior. Therefore, increasing societal pressure through public campaigns and visible community support can help shift subjective norms and drive behavioral change. For example, initiatives that comprise local leaders or influencers advocating for eco-friendly practices create a ripple effect among consumers.

The application of ANFIS produces a model with accuracy of 92.12 and 90.23% on the training and testing data, respectively, showing effectiveness in capturing the complexity of consumer behavior. This research makes a theoretical contribution to the literature and has important practical implications. The results provide valuable insights for businesses and policymakers in designing targeted interventions, such as incentives for sustainable packaging adoption and educational campaigns to enhance environmental awareness. Businesses and governments can use the results to design policies and strategies in line with the adoption of eco-friendly packaging. However, this research has several limitations that should be acknowledged. First, the sample is limited to millennial consumers, which may restrict the generalizability to other demographic groups. Future research should expand the scope by incorporating a more diverse population to enhance external validity. Second, this research primarily relies on self-reported data, which may introduce biases such as social desirability. The use of alternative methods, such as behavioral experiments or observational research, can mitigate the limitation. Third, the ANFIS model shows high predictive accuracy, but the interpretability remains a challenge compared to conventional statistical methods. Future research should explore a hybrid model that combines ANFIS with explainable AI to improve model transparency and facilitate better decision-making.

Future research should include a broader demographic analysis, including various generational cohorts (Baby Boomers, Gen X, Y, and Z), to understand the differences in green packaging usage and develop interventions for each group. Furthermore, a more detailed exploration of specific behavior regarding eco-friendly packaging in reducing plastic bag usage can offer more profound insights. The research does not explicitly analyze external factors, such as regulatory policies or market dynamics, which can significantly impact eco-friendly packaging adoption. These aspects should be integrated to develop a more holistic framework for understanding green consumer behavior. Future research can refine the predictive model for eco-friendly packaging adoption, contributing to theoretical advancements and practical applications in sustainability efforts by addressing the limitations.