Conjoint analysis method for waste classification willingness of event activity participants

Abstract

The invention provides a conjoint analysis method for waste classification willingness of event activity participants, and the method comprises the steps: observing a plurality of factors affecting the waste classification willingness through latent variable hypothesis, collecting target sample data through a Likert scale, and analyzing and evaluating the reliability and validity of latent variables through verification factors; calculating a path coefficient based on SEM; taking the latent variable with the significant path coefficient and the observation variable corresponding to the latent variable as input features of an XGBoost model, training and optimizing hyper-parameters; sHAP values of the XGBoost model are calculated, key factors of multiple scenes are determined by combining the SEM path coefficients and the importance of the SHAP values, and differentiated waste classification management measures for different activity types are put forward. According to the method, two methods of XGBoost-SHAP and SEM are combined, so that a scientific and practical panoramic analysis framework can be provided for formulating a waste classification management strategy.

Description

Technical Field

[0001] The present application relates to the intersection of social science and environmental management, and in particular to a joint analysis method for waste classification willingness of participants in event activities. Background Art

[0002] As the global climate change problem becomes increasingly serious, the potential impact of various large-scale events (such as sports events, chamber of commerce forums, exhibitions, concerts, etc.) on the environment has gradually become the focus of social attention. These activities usually involve a large number of people moving and resource consumption, and the classification and treatment of waste plays an important role in the environmental impact of the activities. International and regional policies (such as the Paris Agreement and the United Nations Sustainable Development Goal SDG 12) have increasingly stringent requirements for the sustainability of activities, especially in terms of waste management, requiring all kinds of activities to follow the principles of "reduction, resource utilization, and harmlessness" in planning and implementation. However, due to the significant differences in the characteristics of different activity scenes, the behavioral patterns of the participating people, and the level of facility support, waste classification management needs to formulate scientific and effective strategies according to local conditions to achieve efficient classification and environmental sustainability of activities.

[0003] Structural Equation Modeling (SEM) is a commonly used statistical tool in social science research. It is based on theoretical assumptions and is good at revealing causal paths and behavioral mechanisms between potential variables. However, it has limitations when dealing with high-dimensional data and nonlinear relationships, and it is difficult to accurately rank the importance of variables. In contrast, the machine learning methods that have emerged in recent years can efficiently process complex data patterns and intuitively quantify the importance of variables, but their "black box" properties limit the in-depth understanding of behavioral logic.

[0004] Therefore, relying solely on a single method may make it difficult to fully reveal the influencing factors and mechanisms of waste sorting willingness. There is a lack of global comprehensive analysis methods for different industries and activities of different scales, as well as a lack of research strategies that can take into account both potential causal paths and feature importance. Summary of the invention

[0005] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0006] To this end, the first purpose of this application is to propose a joint analysis method for the waste classification willingness of participants in event activities. By combining the two methods of XGBoost-SHAP and structural equation model (SEM), it can not only measure the importance and influence direction of each factor from the perspective of machine learning, but also identify the potential paths and causal relationships that affect the willingness to classify waste, and provide targeted waste classification management strategies for different types of activity scenarios. The second objective of the present application is to propose a joint analysis device for the waste classification willingness of participants in event activities.

[0007] The third objective of the present application is to provide an electronic device.

[0008] A fourth objective of the present application is to provide a computer-readable storage medium.

[0009] A fifth object of the present application is to provide a computer program product.

[0010] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a joint analysis method for waste classification willingness of event activity participants, comprising: Based on the latent variable hypothesis, multiple factors affecting the willingness to sort waste were observed, and the target sample data were collected using the Likert scale; Based on the target sample data, the reliability and validity of the latent variables are evaluated through confirmatory factor analysis, and the path coefficients between the latent variables are calculated based on the structural equation model (SEM); The latent variables with significant path coefficients in the SEM analysis results and their corresponding observed variables are used as input features of the XGBoost model, and the XGBoost model is trained, and the hyperparameters are adjusted through grid search or Bayesian optimization. Use the SHAP library to build an interpreter, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to sort waste, and draw SHAP value distribution graphs, feature importance graphs, and dependency graphs to visualize the positive and negative impacts of each feature in different activity scenarios; By comparing the SEM path coefficient with the SHAP value, we can identify the key factors that play an important role in the path relationship and are also highly important in machine learning, and formulate targeted waste classification management strategies for different types of event activities.

[0011] Optionally, the multiple factors affecting the willingness to classify waste are observed based on the latent variable hypothesis, and the target sample data is collected using a Likert scale, including; The questionnaire was designed, and the factors affecting the willingness to classify waste were taken as latent variables. The corresponding observed variables were designed respectively. The basic demographic characteristics of the participants were taken as observed variables, and the Likert scale was used to score each observed variable. Target sample data is obtained through online and offline questionnaires targeting participants in sports events, chamber of commerce forums, exhibitions and entertainment activities.

[0012] Optionally, before evaluating the reliability and validity of the latent variables through confirmatory factor analysis based on the target sample data, the method further includes: Conduct a logical consistency check on the collected target sample data to check whether there are obvious contradictions or inconsistencies in the data in the same questionnaire; Process data with missing values, use multiple interpolation methods to fill missing values ​​or use mean interpolation methods to fill missing values ​​in numerical data; Use box plot screening methods to identify and remove outliers that deviate from normal distribution in the data, and use clustering-based anomaly detection methods to detect and remove potential outliers; Perform one-hot encoding transformation on all qualitative variables, converting each categorical variable into multiple binary features; All numerical variables were standardized to ensure that all numerical features were analyzed on the same scale.

[0013] Optionally, based on the target sample data, evaluating the reliability and validity of latent variables through confirmatory factor analysis, and calculating the path coefficients between latent variables based on the structural equation model (SEM), include: The reliability and validity of the scale were evaluated through confirmatory factor analysis to ensure the convergent validity and discriminant validity of the latent variables; Construct a SEM and use the maximum likelihood estimation method for parameter estimation, and evaluate the model fit using preset fit indices, where the preset fit indices are the ratio of chi-square to degrees of freedom, the fit index, the adjusted fit index, the baseline fit index, the non-standard fit index, the comparative fit index, and the root mean square error of approximation; On the premise that the model fit meets the preset requirements, the path coefficients between the latent variables are obtained according to SEM.

[0014] Optionally, the significant latent variables and observed variables in the SEM analysis results are used as input features of the XGBoost model, the XGBoost model is trained, and hyperparameters are adjusted by grid search or Bayesian optimization, including: The path-significant latent variables and their observed variables extracted from the SEM analysis results are used as the input features of the XGBoost model, and the scenario variables after one-hot encoding are merged with other latent variables to form a complete input feature set. The target variable is the waste classification willingness behavior score. The preprocessed data is divided into a training set and a test set according to a preset ratio, and the waste classification willingness or classification propensity score is determined as the prediction target; Based on the training set, using grid search or Bayesian optimization method to adjust the hyperparameters of the XGBoost model, the hyperparameters including learning rate, maximum depth of the tree, sampling rate, column sampling rate, regularization parameter, to find the optimal hyperparameter combination; K-fold cross validation is used to evaluate the fitting performance of the XGBoost model on the training set, the optimal parameter combination is screened by the mean square error minimization indicator, and the R² and root mean square error indicators are used to evaluate the overall performance of the model.

[0015] Optionally, the interpreter is constructed using the SHAP library, the SHAP value is calculated for the trained XGBoost model, the marginal contribution of each feature to the willingness to classify waste is quantified, and a SHAP value distribution graph, a feature importance graph, and a dependency graph are drawn to visualize the positive and negative effects of each feature in different activity scenarios, including: Use the SHAP library to build an interpreter and input the trained XGBoost model into the SHAP interpreter to calculate the SHAP value to quantify the marginal contribution of each input feature to the willingness to sort waste; Based on the calculated SHAP values, the impact of each feature on the prediction of waste sorting willingness is evaluated, and the importance of each feature to the target variable is revealed; According to the calculated SHAP value and the importance of each feature to the target variable, a SHAP value distribution graph, a feature importance graph and a dependency graph are drawn, wherein the SHAP value distribution graph is used to show the distribution of the SHAP value of each feature in different data samples, the feature importance graph is used to show the relative importance of different features to the willingness to classify waste, and the dependency graph is used to visualize the relationship between the feature and the willingness to classify waste.

[0016] To achieve the above-mentioned purpose, the second embodiment of the present application proposes a joint analysis device for waste classification willingness of event activity participants, comprising: The questionnaire design and data collection module is used to observe multiple factors that affect the willingness to sort waste based on the latent variable hypothesis and collect target sample data using the Likert scale; A SEM model construction and verification module is used to evaluate the reliability and validity of latent variables through confirmatory factor analysis based on the target sample data, and calculate the path coefficients between latent variables based on the structural equation model SEM; XGBoost model training and optimization module, which is used to use the latent variables with significant path coefficients in the SEM analysis results and their corresponding observed variables as input features of the XGBoost model, train the XGBoost model, and adjust the hyperparameters through grid search or Bayesian optimization; SHAP interpretability analysis module, which is used to build an interpreter using the SHAP library, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to sort waste, and draw SHAP value distribution graphs, feature importance graphs, and dependency graphs to visualize the positive and negative impacts of each feature in different activity scenarios; The fusion analysis and application module is used to compare the SEM path coefficient with the SHAP value, identify the key factors that play an important role in the path relationship and are also highly important in machine learning, and formulate targeted waste classification management strategies for different types of event activities.

[0017] To achieve the above-mentioned purpose, the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects above.

[0018] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present application proposes a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by a processor, they are used to implement the method as described in any one of the above-mentioned first aspects.

[0019] To achieve the above-mentioned purpose, the fifth aspect of the present application proposes a computer program product, including a computer program, which, when executed by a processor, implements the method as described in any one of the above-mentioned first aspects.

[0020] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects: By combining XGBoost-SHAP (XGBoost-SHapley Additive exPlanations) with the structural equation model (SEM), the advantages are complementary and suitable for two application scenarios: global causal relationship decomposition and scenario difference feature analysis. XGBoost-SHAP optimizes the variable importance ranking in a data-driven way, significantly reducing the work input of variable screening; while SEM provides a theory-driven causal path analysis, revealing the relationship between potential variables. The combination of the two effectively breaks through the limitations of a single model method, can not only deeply analyze the behavior logic, but also adapt to the differentiated needs in multiple scenarios, and provides a scientific and practical panoramic analysis framework for the formulation of waste classification management strategies. Through this method, the importance and influence direction of each factor can be accurately measured, and the potential causal paths that affect the willingness to classify waste can be identified, thereby providing tailored waste classification management solutions for different types of activity scenarios.

[0021] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A flowchart of a joint analysis method for waste classification willingness of event activity participants provided in an embodiment of the present application; Figure 2 A flowchart of a joint analysis method for waste classification willingness of event activity participants provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a joint analysis device for waste classification willingness of event activity participants provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0024] Figure 1 and Figure 2 This is a flow chart of a joint analysis method for the waste classification willingness of event participants provided in an embodiment of the present application. Figure 2 The flowchart includes two analysis strategies. The first strategy is applicable to low-dimensional data. It quantitatively analyzes the differences in complex scenarios under the assumption of known theoretical frameworks, latent variables, and explicit variables. The second strategy is applicable to high-dimensional data. It is recommended when there are many feature variables and complex nonlinear relationships, and when it is necessary to screen key variables and reduce variable noise.

[0025] Combine the following Figure 1 , a possible implementation method proposed in this application is described in detail, but this application is not limited to this. The implementation goal is to understand the influencing factors and action paths of the waste sorting willingness of participants in multi-event activity scenarios, and to rank the key influencing factors in different scenarios, namely, Figure 2 Strategy 1 in .

[0026] like Figure 1 As shown, the method comprises the following steps: Step 101, observe multiple factors that affect the willingness to classify waste based on the latent variable hypothesis, and use the Likert scale to collect target sample data.

[0027] During the implementation of this application, we first collect relevant data from participants by designing a questionnaire. The design of the questionnaire follows the principles of scientificity, feasibility and ethics to ensure the accuracy and reliability of data collection. At the same time, it follows relevant ethical regulations to ensure the informed consent and privacy protection of the participants.

[0028] In the embodiments of the present application, the selection of latent variables is based on theoretical research and literature survey results on the influencing factors of waste classification behavior. In order to quantify and observe the impact of these latent variables, the present application also designs corresponding observation variables for each latent variable.

[0029] Specifically, latent variables include multiple factors that affect the willingness to sort waste, which are quantified and measured through observed variables, and the observed variables of each latent variable are designed to target the influence of these factors. The observed variable of social norms is the participants' willingness to observe and imitate the sorting behavior of others, which can reflect the degree of influence of social pressure or the behavior of others; the observed variables of site facility convenience include the clarity of the signs of waste sorting facilities, the rationality of the location, and the convenience of using the facilities, which directly affect whether participants can conveniently sort waste; the observed variables of perceived pros and cons reflect the participants' cognition of the possible environmental benefits and personal burdens of waste sorting, and reflect their perception of the internal driving force and external motivation of the behavior; the observed variables of sense of responsibility mainly measure the participants' cognition of the responsibility for waste sorting behavior, and whether they have a strong internal driving force to perform the sorting task.

[0030] It can be understood that in this application, the selection of latent variables and observed variables is mainly based on the preliminary analysis of the factors affecting the willingness to classify waste. The selected latent variables include social norms, site facility convenience, perceived pros and cons, and sense of responsibility, while the observed variables are specifically designed for these latent variables. These latent variables and observed variables are selected based on the current research background and the results of literature research, and this application does not specifically limit the setting of latent variables and observed variables.

[0031] In fact, the specific selection of latent variables and observed variables can be adjusted according to different application scenarios, research objects and practical problems faced. In different activity scenarios or research backgrounds, there may be other potential factors that affect the willingness to classify waste, which may affect the setting of latent variables and observed variables. For example, in certain specific activity scenarios, other characteristics may be of concern, such as the socioeconomic status of participants, the degree of awareness of environmental protection, etc. These factors can be set as new latent variables or observed variables according to actual needs.

[0032] Therefore, the latent variables and observed variables mentioned in this application are only a specific implementation method, which aims to demonstrate the analysis framework and methodology of this application. Users can flexibly adjust the settings of latent variables and observed variables according to different research objectives and actual needs to adapt to the waste classification willingness analysis in different scenarios.

[0033] In addition, the questionnaire also needs to collect basic demographic characteristics data of the participants, such as age, gender, income, education level, etc. These data can help analyze how these socio-demographic factors affect the willingness to sort waste, thereby providing support for more precise formulation of management strategies for different groups.

[0034] In the embodiment of the present application, the questionnaire design uses a Likert scale (e.g., 1-5 points, representing from "completely disagree" to "completely agree") to score each observed variable to ensure that the data can be quantified and provide clear indicators for subsequent analysis. This quantification method can simplify data processing and provide an efficient basis for subsequent statistical analysis. The Likert scale is widely used in social science research and can effectively measure the attitudes and cognitive levels of participants, thereby providing strong support for further analysis of the relationship between latent variables.

[0035] It should be noted that in this application, the data collection method is not limited to the Likert scale. In practical applications, other suitable quantitative data collection methods can also be selected according to different research needs and application scenarios. For example, it is possible to directly collect data through numerical data (such as participants' actual behavior records, activity participation frequency, classification behavior scores, etc.), and perform simulation analysis based on these direct quantitative data. This method avoids the subjective bias caused by self-reporting, and can directly reflect the actual behavior of participants or the actual impact on waste classification, thereby providing more accurate input data for the model.

[0036] Therefore, the present application can not only use the Likert scale for data collection, but also perform simulations based on directly collected quantitative data. The flexible collection method enables the analysis framework of the present invention to be widely used in different scenarios and enhances its adaptability and practicality.

[0037] During the implementation of this application, the objects of data collection are participants in sports events, chamber of commerce forums, exhibitions and entertainment activities. These activity types represent different activity scenarios and can provide diverse sample data for the embodiments of this application.

[0038] As a possible implementation method, the questionnaire is distributed through a combination of online and offline methods. The online part is widely disseminated through channels such as social platforms, while the offline part is distributed directly at the event site, ensuring the diversity and representativeness of the sample.

[0039] Through the questionnaire design and data collection in the embodiments of the present application, target sample data can be obtained, which will be used for further preprocessing, causal analysis and model training in subsequent steps, providing reliable data support for a comprehensive understanding of the factors affecting the willingness to classify waste.

[0040] Step 102, based on the target sample data, the reliability and validity of the latent variables are evaluated through confirmatory factor analysis, the path coefficients and significance levels between the latent variables are calculated based on the structural equation model (SEM), and the key driving factors and causal relationship paths of waste classification willingness are identified.

[0041] In the embodiments of the present application, data preprocessing is an important step to ensure the accuracy of the model and the reliability of the results. After the collected target sample data is preprocessed, the data can be effectively cleaned, missing values ​​and outliers can be processed, and clearer and more effective input data can be provided for subsequent causal analysis and model training. The preprocessing steps specifically include the following aspects: First, a logical consistency check is performed. This check is designed to ensure that there are no obvious contradictions or inconsistencies in the data in the same questionnaire. For example, if a participant indicates that they fully support waste sorting in response to a question in the questionnaire, but indicates that they are not interested in or completely opposed to sorting in another question, this will be considered a logical error in the data. Through the logical consistency check, these data anomalies can be discovered and eliminated, thereby ensuring the internal consistency of the data.

[0042] Secondly, for the processing of missing values, some samples may have missing values ​​during the data collection process. For missing data, a multiple interpolation method is used in the embodiment of the present application to fill the missing values, avoiding the problem of sample size reduction caused by directly deleting the missing data. For missing values ​​in numerical data, the mean interpolation method can also be used to fill them. Mean interpolation replaces the missing values ​​with the mean of the variable, so that the data distribution remains consistent and reduces interference with model training.

[0043] Then, outliers are removed. Outliers refer to data points that are obviously deviated from the normal distribution, which may have an adverse effect on model training. In the embodiment of the present application, a box plot screening method is used to identify and remove outliers, and the box plot can effectively identify extreme high and low values. In addition, the present application also uses a clustering-based anomaly detection method to detect potential outliers through a clustering algorithm and remove them from the data set. This ensures that the data used for subsequent analysis is more stable and reasonable in statistical characteristics.

[0044] For qualitative variables, the embodiment of the present application uses the One-Hot Encoding method for conversion, converting each categorical variable into multiple binary features. One-Hot Encoding can avoid the influence of the model training process on the order problem between dependent variables (such as gender, activity type and other categorical variables). For example, the activity type variable may include several types such as sports events, exhibitions, forums, etc. After one-hot encoding, a new binary feature (such as TYP_Sport, TYP_Exhibition, etc.) will be created for each category. These binary features can be used as inputs to the model together with other variables.

[0045] Finally, for numerical variables, the embodiments of the present application perform standardization processing on them to ensure that all numerical features are analyzed on the same scale. Standardization is to subtract the mean of each numerical variable and divide it by the standard deviation so that the distribution of each feature has the same mean (0) and standard deviation (1). After standardization, the scale of the numerical variables is unified, which can avoid unfair effects on model training due to differences between features of different dimensions.

[0046] Through the data preprocessing in the above steps, the quality of sample data can be guaranteed, providing high-quality input data for subsequent modeling and analysis. Data cleaning and preprocessing are crucial parts of data analysis, which lay a solid foundation for subsequent causal relationship analysis, model training and result interpretation.

[0047] In the embodiment of the present application, after the preliminary data preprocessing, the reliability and validity evaluation stage of the latent variables is entered. This stage mainly uses the combination of confirmatory factor analysis (CFA) and structural equation model (SEM) to deeply analyze the relationship and influence path of the latent variables and identify the key driving factors of waste classification willingness.

[0048] First, confirmatory factor analysis (CFA) is performed to evaluate the reliability and validity of the scale corresponding to the latent variable. Reliability refers to the ability of the scale to consistently measure the latent variable, while validity refers to whether the scale can accurately measure the latent variable. In this application, CFA is used to ensure the convergent validity and discriminant validity of the latent variable. Convergent validity is used to verify whether each observed variable can effectively reflect the corresponding latent variable, while discriminant validity is used to ensure the distinction between different latent variables without confusion or overlap.

[0049] Next, based on the scale verified by CFA, the structural equation model (SEM) was constructed and run using AMOS software, and the maximum likelihood estimation (MLE) method was used for parameter estimation. The maximum likelihood estimation method is a widely used statistical estimation method that can optimize the parameters of the model so that the estimated model has the best fit with the observed data. In order to ensure the rationality of the model, it is necessary to evaluate the fit of the model based on multiple fit indicators.

[0050] However, it should be noted that this application is not limited to the maximum likelihood estimation method (MLE). In some application scenarios, the least square method (PLS-SEM) can also be used as an alternative method. The least squares structural equation model (PLS-SEM) is an effective method for processing high-dimensional data, non-normal data or small sample sizes. It uses the least squares method of the path model to estimate parameters. In some specific scenarios, PLS-SEM may be more applicable than the MLE method.

[0051] In addition, based on different software platforms, this application can also be flexibly implemented using statistical analysis tools such as Python or R. Python and R are both powerful programming languages ​​that can provide extensive statistical analysis and model building support. On these platforms, users can choose tools and libraries that suit their specific needs to implement the construction and estimation of structural equation models.

[0052] Therefore, the maximum likelihood estimation method (MLE) selected in this application is only a specific implementation method, which aims to demonstrate the usage scenarios and advantages of this method. In practical applications, according to factors such as data characteristics, sample size, and computing resources, users can flexibly choose different estimation methods and implementation platforms, such as using PLS-SEM or other suitable estimation methods to adapt to different research backgrounds and needs. This application does not make specific restrictions on this.

[0053] In the embodiment of the present application, the preset fitting indicators include: The ratio of chi-square to degrees of freedom χ² / df: used to evaluate the degree of fit between the model and the data. A smaller chi-square value means a better fit. Goodness of fit index GFI, adjusted goodness of fit index AGFI: measure the fit of the model, the closer the value is to 1, the better; Benchmark fit index NFI, non-normative fit index TLI, comparative fit index CFI: These indices are used to compare the relative goodness of model fit, with larger values ​​indicating a more appropriate model; Root mean square error (RMSEA): used to measure the model fitting error. The smaller the value, the better the fitting effect.

[0054] In the embodiment of the present application, if the above-mentioned fitting index meets the preset requirements, it means that the model fit is good and the model is acceptable, so that the subsequent path coefficient analysis is carried out.

[0055] In a possible embodiment, if the indicators satisfy the ratio of indicator chi-square to degrees of freedom χ2 / df<3, the goodness of fit index GFI>0.8, the adjusted goodness of fit index AGFI>0.8, the benchmark fit index NFI>0.9, the non-normative fit index TLI>0.9, the comparative fit index CFI>0.9, and the root mean square error RMSEA<0.08, it means that the model has a good fit and is acceptable.

[0056] After verifying that the model fit meets the preset requirements, the embodiment of the present application further analyzes the relationship between the latent variables through the structural equation model (SEM): SEM can calculate the path coefficient, that is, the strength and direction (positive or negative) of the relationship between the latent variables. In addition, the significance level of the path coefficient can also be evaluated, and the significance test can be used to determine whether the relationship between the latent variables is statistically significant. These path coefficients and significance levels provide a theoretical basis for analyzing the key driving factors of waste classification willingness.

[0057] In the embodiment of the present application, by analyzing the latent variable path relationship of waste classification willingness in different scenarios, the action path of each key factor can be identified. Specifically, this step can reveal how latent variables such as social norms, site facility convenience, perceived pros and cons, and sense of responsibility jointly affect the participants' classification willingness through different paths. Based on these causal relationship paths, it can be further clarified which factors play a decisive role in the willingness to classify waste in different scenarios, providing theoretical support for subsequent differentiated management strategies.

[0058] The analysis in this step can provide a deep theoretical understanding of the key drivers of waste sorting, help formulate more precise and targeted management strategies, and ensure that participants in different types of activities can take effective sorting behaviors according to their specific situations and characteristics.

[0059] Step 103, using the latent variables with significant path coefficients in the SEM analysis results and their corresponding observed variables as input features of the XGBoost model, training the XGBoost model, and adjusting the hyperparameters through grid search or Bayesian optimization.

[0060] In the embodiment of the present application, based on the SEM analysis results obtained in step 102, the latent variables with significant path coefficients and their corresponding observed variables are used as input features of the XGBoost model. These latent variables and observed variables have been verified for reliability and validity in previous analyses, so their impact on the willingness to classify waste has a high degree of credibility. In this way, the XGBoost model can use these key factors verified by theoretical analysis to predict the willingness to classify waste or the classification propensity score.

[0061] Specifically, this application will first extract path-significant latent variables and their observed variables from the SEM analysis as input features of the XGBoost model. After preprocessing, these features can be directly used as input to the machine learning model. In order to enhance the flexibility of the model, the scene variables (such as activity type) converted by one-hot encoding are also merged with other latent variables to form a complete input feature set. The one-hot encoding of the scene variables helps to eliminate the sequential effects of the category data, so that the model can effectively learn the feature differences in different scenarios. The target variable is the waste classification willingness behavior score, that is, the classification propensity score of the participant is predicted by the model.

[0062] Next, the preprocessed data will be divided into a training set and a test set according to a preset ratio (usually 80% training set and 20% test set, which is not specifically limited in this application) to ensure that the model can effectively learn during the training process and evaluate its generalization ability through the test set. The target variable will be set to the waste classification willingness or classification propensity score for model training and evaluation.

[0063] During the training process, in order to find the optimal model hyperparameters, this application uses grid search or Bayesian optimization methods to adjust the hyperparameters of the XGBoost model. Specific hyperparameters include learning rate (learning_rate), maximum tree depth (max_depth), sampling rate (subsample), column sampling rate (colsample_bytree), regularization parameter (gamma), etc. The adjustment of these hyperparameters can optimize the performance of the model, so that it can better fit the training data and show strong generalization ability on the test data.

[0064] In order to evaluate the performance of the model and ensure its fitting quality, the embodiment of the present application also uses k-fold cross validation (k=5) to evaluate the fitting performance of the model on the training set. Through k-fold cross validation, the data set is divided into k subsets, the model will be trained on k-1 subsets, and validated on the remaining subsets, so as to ensure that the performance of the model on different data subsets is consistent and avoid the occurrence of overfitting. Through the mean square error (MSE) minimization indicator, the optimal hyperparameter combination can be screened out and the model's prediction of waste classification willingness can be ensured to have a high accuracy.

[0065] In a possible embodiment, if the amount of data is very small (eg, the sample size of each group is <50), k=3 may be selected or adjusted as required, and this application does not make any specific limitation on this.

[0066] When evaluating the overall performance of the model, the embodiment of the present application also uses the coefficient of determination (R²) and the root mean square error (RMSE) as evaluation indicators. The closer the R² value is to 1, the stronger the model's ability to explain the target variable; RMSE is used to measure the deviation between the predicted value and the actual value. The smaller the RMSE value, the higher the prediction accuracy of the model.

[0067] It should be noted that the evaluation method in this application is not limited to the two indicators of R² and RMSE. In practical applications, other evaluation methods can be flexibly selected according to different needs and scenarios. This application does not make specific limitations on this and aims to provide an evaluation framework that is flexible and adaptable to different application scenarios.

[0068] Through these steps, the XGBoost model can make full use of the key latent variable information obtained from the SEM analysis, and optimize the model through fine hyperparameter adjustment and cross-validation, and finally achieve accurate prediction of waste classification willingness. This process provides strong data support for the subsequent formulation of waste classification management strategies, which can help better understand the impact of different factors on the classification behavior of participants and provide a strong basis for policy formulation.

[0069] Step 104, use the SHAP library to build an interpreter, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to classify waste, and draw a SHAP value distribution graph, a feature importance graph, and a dependency graph to visualize the positive and negative effects of each feature in different activity scenarios.

[0070] In the implementation of this application, the key task of step 104 is to use the SHAP library to perform interpretability analysis on the trained XGBoost model. The SHAP value can quantify the marginal contribution of each feature to the prediction target (willingness to classify waste), thereby providing a transparent explanation of the model prediction results.

[0071] In this step, the SHAP value is calculated to gain a deeper understanding of the importance of each feature in predicting the willingness to sort waste, and visualization technology is used to further reveal the positive or negative impact of the feature on the willingness to sort waste.

[0072] Specifically, this application uses the SHAP library to build an interpreter, takes the trained XGBoost model as input, and calculates the SHAP value of each input feature. The SHAP value can provide the contribution of each feature of the model to the prediction result, and quantify the marginal effect of each feature, that is, how the change of a certain feature affects the prediction result of waste classification willingness when other features are fixed. The SHAP value is an interpretation method based on cooperative game theory, which can assign a clear contribution value to each feature, helping researchers understand how the model makes predictions based on different features.

[0073] Next, based on the calculated SHAP values, this application evaluates the impact of each feature on the prediction of waste sorting willingness and reveals the importance of each feature in different activity scenarios. Through the SHAP value, it can be clarified which features have a greater impact on the prediction of waste sorting willingness and which features contribute less to the prediction. The quantification process of the SHAP value can help us determine which factors are the key drivers affecting the willingness to sort waste.

[0074] In order to intuitively display the importance of features and their impact on the target variable, the present embodiment further draws SHAP value distribution diagrams, feature importance diagrams, and dependency diagrams. These diagrams can effectively help users understand the relationship between features and waste classification willingness from different perspectives, and identify the top several key factors that have the greatest impact on waste classification willingness in different types of activity scenarios. Among them: SHAP value distribution diagram: Displays the distribution of SHAP values ​​of each feature in different data samples. Through the distribution diagram, we can observe whether there are obvious differences in the distribution of feature values ​​in different samples, and further analyze whether the impact of this feature on the willingness to classify waste is positive or negative.

[0075] Feature Importance Graph: Shows the relative importance of different features to the willingness to sort waste. Based on the average absolute value of the SHAP value, you can intuitively see which features contribute most to the model's prediction, thereby identifying the most critical factors in different activity scenarios.

[0076] Dependency graph: used to visualize the relationship between features and willingness to sort waste. The dependency graph shows how the change in the value of each feature affects the prediction of the target variable, and can reveal the specific relationship between the feature and the willingness to sort waste. Through the dependency graph, the linear or nonlinear impact of feature changes on the willingness to sort waste can be observed.

[0077] In addition, according to the research objectives, interaction terms can be constructed and introduced to further analyze the impact of complex factors and verify the significance of the interaction of multiple latent variables in different scenarios. This application does not provide specific explanations and limitations on this.

[0078] Through the interpretability analysis of the embodiments of this application, not only can the key factors affecting the willingness to classify waste be revealed, but also the performance of these factors in different activity scenarios can be visualized, so that decision makers can more clearly understand the positive and negative impacts of each feature on the willingness to classify. This provides an important basis for the subsequent formulation of targeted waste classification management strategies, ensuring that the strategies can be adjusted and optimized according to the feature differences in different scenarios.

[0079] Step 105, compare the SEM path coefficient with the SHAP value to identify key factors that play an important role in the path relationship and are also highly important in machine learning, and formulate targeted waste classification management strategies for different types of event activities.

[0080] In the implementation of this application, the core task of step 105 is to compare and analyze the latent variable path coefficient obtained through the structural equation model SEM with the SHAP value, and identify the key factors that play an important role in the causal path and are also highly important in the machine learning model. Through this process, theoretical analysis can be combined with data-driven machine learning methods to further deepen the understanding of the factors affecting the willingness to classify waste, and provide guidance for the formulation of waste classification management strategies for different types of events.

[0081] First, the SEM path coefficient and SHAP value are compared. The path coefficient obtained through SEM analysis reveals the causal relationship and influence strength between latent variables, while the SHAP value quantifies the contribution of each feature to the prediction of waste classification willingness through the machine learning model. By comparing the results of the two, we can find out the latent variables that occupy an important position in the SEM model (that is, the path coefficient is large) and also show important characteristics of high SHAP values ​​in the XGBoost model. Through this comparative analysis, we can determine which factors are key factors in both the theoretical model and the machine learning model, so as to obtain more universal conclusions.

[0082] Secondly, identify key factors. After comparative analysis, the embodiments of the present application can identify factors that have high importance in different models, which are considered to be key drivers of waste classification willingness. These key factors may have different manifestations and mechanisms of action in different types of event activity scenarios, so they are crucial for formulating targeted waste classification management strategies.

[0083] Finally, based on the results of the comparative analysis, targeted waste classification management strategies are formulated for different types of events. By clarifying which factors are of high importance in various event scenarios, decision makers can strengthen or optimize these key factors in a targeted manner during strategy formulation. For example, for some events, it may be necessary to increase the participation in waste classification by optimizing the convenience of venue facilities or enhancing the promotion of social norms; while for other events, more attention may be paid to factors such as enhancing the participants' perception of pros and cons or sense of responsibility. Through this differentiated strategy formulation, waste classification in various activities can be promoted more effectively and the classification effect can be improved.

[0084] In summary, this step fully explored the key influencing factors of waste classification willingness by combining theory-driven structural equation modeling with data-driven machine learning methods, and provided accurate management strategies for different types of events based on these factors. This comprehensive analysis method provides comprehensive, in-depth and scientific support for waste classification management, which can ensure the pertinence and effectiveness of management strategies.

[0085] It should be noted that when faced with complex high-dimensional data or scenarios with uncertainty, the joint analysis method of this application can be adjusted in sequence according to actual needs to enhance the applicability of the model and improve analysis efficiency. Especially when there are a large number of features and nonlinear relationships, the traditional direct application of the SEM model may face problems such as excessive computational burden, excessive model complexity, or insufficient path significance. Therefore, the key influencing factors can be screened by the XGBoost-SHAP method, and then the SEM model can be fitted to obtain the global key action path. The sequential adjustment of this method provides a more efficient solution for subsequent analysis, namely the attached Figure 2 The specific implementation of Strategy 2 is as follows: 1. First use XGBoost-SHAP to screen key variables: When there are many potential interference factors, directly performing SEM modeling on all observed variables may make the model too complicated and lack significance due to too many path assumptions. Therefore, the XGBoost model is first used to train the full data, and combined with SHAP value analysis, the features that contribute more to the target variable (willingness to classify waste) are screened out. The SHAP value can effectively quantify the marginal contribution of each feature to the model prediction. This method can eliminate noise features, reduce data dimensions, and retain key variables that have an important impact on the prediction of waste classification willingness.

[0086] 2. Subsequent SEM model fitting and verification: After screening out the key variables, these variables are integrated and input into the SEM model. Due to the small number of input variables, the structure of the SEM model will be simplified and the computational burden will be reduced accordingly. Through path analysis, the causal relationship between the key variables can be further verified, and their specific action paths on the willingness to classify waste can be clarified. This adjustment makes the SEM model more focused on the core factors that affect the willingness to classify waste, thereby improving the model fit and the interpretability of the results.

[0087] 3. Expansion of applicable scenarios: This order adjustment method is particularly suitable for research scenarios with large data volumes and high variable dimensions. In these scenarios, screening out key variables before SEM modeling can significantly improve the model's fitting efficiency and the interpretability of the results, and avoid statistical redundancy or model mismatch caused by too many path assumptions. In addition, this order adjustment can also reduce the calculation time during model training and improve the efficiency of the analysis process, especially when faced with large amounts of data and complex features, and can more flexibly and efficiently obtain reliable analysis results.

[0088] Through this sequential adjustment, this application can effectively cope with the challenges brought by high-dimensional data, simplify the data analysis process, improve the prediction accuracy of the model, and provide a more targeted and practical analysis framework for the study of waste classification willingness.

[0089] In order to implement the above-mentioned embodiment, the present application also proposes a joint analysis device for waste classification willingness of event activity participants. Figure 3 This is a schematic diagram of the structure of a joint analysis device for the waste classification willingness of event participants provided in an embodiment of the present application. Figure 3 As shown, the device comprises: The questionnaire design and data collection module 100 is used to observe multiple factors that affect the willingness to classify waste based on the latent variable hypothesis and collect target sample data using the Likert scale; The SEM model construction and verification module 200 is used to evaluate the reliability and validity of latent variables through confirmatory factor analysis based on the target sample data, and calculate the path coefficients between latent variables based on the structural equation model SEM; The XGBoost model training and optimization module 300 is used to use the latent variables with significant path coefficients in the SEM analysis results and their corresponding observed variables as input features of the XGBoost model, train the model using the XGBoost model, and adjust the hyperparameters through grid search or Bayesian optimization; SHAP interpretability analysis module 400, used to build an interpreter using the SHAP library, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to sort waste, and draw a SHAP value distribution map, a feature importance map, and a dependency map to visualize the positive and negative impacts of each feature in different activity scenarios; The fusion analysis and application module 500 is used to compare the SEM path coefficient with the SHAP value, identify key factors that play an important role in the path relationship and are also highly important in machine learning, and formulate targeted waste classification management strategies for different types of event activities.

[0090] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0091] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0092] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0093] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0094] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0095] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0096] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0097] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they contradict each other.

[0098] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0099] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0100] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0101] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0102] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0103] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0104] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

[0105] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0106] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A joint analysis method for waste classification willingness of event participants, characterized in that: include: Based on the latent variable hypothesis, multiple factors affecting the willingness to sort waste were observed, and the target sample data were collected using the Likert scale; Based on the target sample data, the reliability and validity of the latent variables are evaluated through confirmatory factor analysis, and the path coefficients between the latent variables are calculated based on the structural equation model (SEM); The latent variables with significant path coefficients in the SEM analysis results and their corresponding observed variables are used as input features of the XGBoost model, and the XGBoost model is trained, and the hyperparameters are adjusted through grid search or Bayesian optimization. Use the SHAP library to build an interpreter, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to sort waste, and draw SHAP value distribution graphs, feature importance graphs, and dependency graphs to visualize the positive and negative impacts of each feature in different activity scenarios; By comparing the SEM path coefficient with the SHAP value, we can identify the key factors that play an important role in the path relationship and are also highly important in machine learning, and formulate targeted waste classification management strategies for different types of event activities.

2. The method according to claim 1, characterized in that The multiple factors that affect the willingness to classify waste are observed based on the latent variable hypothesis, and the target sample data are collected using the Likert scale, including; The questionnaire was designed, and the factors affecting the willingness to classify waste were taken as latent variables. The corresponding observed variables were designed respectively. The basic demographic characteristics of the participants were taken as observed variables, and the Likert scale was used to score each observed variable. Target sample data is obtained through online and offline questionnaires targeting participants in sports events, chamber of commerce forums, exhibitions and entertainment activities.

3. The method according to claim 2, characterized in that Before evaluating the reliability and validity of the latent variables through confirmatory factor analysis based on the target sample data, it also includes: Conduct a logical consistency check on the collected target sample data to check whether there are obvious contradictions or inconsistencies in the data in the same questionnaire; Process data with missing values, use multiple interpolation methods to fill missing values ​​or use mean interpolation methods to fill missing values ​​in numerical data; Use box plot screening methods to identify and remove outliers that deviate from normal distribution in the data, and use clustering-based anomaly detection methods to detect and remove potential outliers; Perform one-hot encoding transformation on all qualitative variables, converting each categorical variable into multiple binary features; All numerical variables were standardized to ensure that all numerical features were analyzed on the same scale.

4. The method according to claim 3, characterized in that Based on the target sample data, the reliability and validity of the latent variables are evaluated through confirmatory factor analysis, and the path coefficients between the latent variables are calculated based on the structural equation model (SEM), including: The reliability and validity of the scale were evaluated through confirmatory factor analysis to ensure the convergent validity and discriminant validity of the latent variables; Construct a SEM and use the maximum likelihood estimation method for parameter estimation, and evaluate the model fit using preset fit indices, where the preset fit indices are the ratio of chi-square to degrees of freedom, the fit index, the adjusted fit index, the baseline fit index, the non-standard fit index, the comparative fit index, and the root mean square error of approximation; On the premise that the model fit meets the preset requirements, the path coefficients between the latent variables are obtained according to SEM.

5. The method according to claim 4, characterized in that The significant latent variables and observed variables in the SEM analysis results are used as input features of the XGBoost model, the XGBoost model is trained, and the hyperparameters are adjusted by grid search or Bayesian optimization, including: The path-significant latent variables and their observed variables extracted from the SEM analysis results are used as the input features of the XGBoost model, and the scenario variables after one-hot encoding are merged with other latent variables to form a complete input feature set. The target variable is the waste classification willingness behavior score. The preprocessed data is divided into a training set and a test set according to a preset ratio, and the waste classification willingness or classification propensity score is determined as the prediction target; Based on the training set, using grid search or Bayesian optimization method to adjust the hyperparameters of the XGBoost model, the hyperparameters including learning rate, maximum depth of the tree, sampling rate, column sampling rate, regularization parameter, to find the optimal hyperparameter combination; K-fold cross validation is used to evaluate the fitting performance of the XGBoost model on the training set, the optimal parameter combination is screened by the mean square error minimization indicator, and the R² and root mean square error indicators are used to evaluate the overall performance of the model.

6. The method according to claim 5, characterized in that The SHAP library is used to build an interpreter, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to classify waste, and draw a SHAP value distribution map, feature importance map, and dependency map to visualize the positive and negative impact of each feature in different activity scenarios, including: Use the SHAP library to build an interpreter and input the trained XGBoost model into the SHAP interpreter to calculate the SHAP value to quantify the marginal contribution of each input feature to the willingness to sort waste; Based on the calculated SHAP values, the impact of each feature on the prediction of waste sorting willingness is evaluated, and the importance of each feature to the target variable is revealed; According to the calculated SHAP value and the importance of each feature to the target variable, a SHAP value distribution graph, a feature importance graph and a dependency graph are drawn, wherein the SHAP value distribution graph is used to show the distribution of the SHAP value of each feature in different data samples, the feature importance graph is used to show the relative importance of different features to the willingness to classify waste, and the dependency graph is used to visualize the relationship between the feature and the willingness to classify waste.

7. A joint analysis device for waste classification willingness of event participants, characterized in that: include: The questionnaire design and data collection module is used to observe multiple factors that affect the willingness to sort waste based on the latent variable hypothesis and collect target sample data using the Likert scale; A SEM model construction and verification module is used to evaluate the reliability and validity of latent variables through confirmatory factor analysis based on the target sample data, and calculate the path coefficients between latent variables based on the structural equation model SEM; XGBoost model training and optimization module, which is used to use the latent variables with significant path coefficients in the SEM analysis results and their corresponding observed variables as input features of the XGBoost model, train the XGBoost model, and adjust the hyperparameters through grid search or Bayesian optimization; SHAP interpretability analysis module, which is used to build an interpreter using the SHAP library, calculate the SHAP value for the trained XGBoost model, quantify the marginal contribution of each feature to the willingness to sort waste, and draw SHAP value distribution graphs, feature importance graphs, and dependency graphs to visualize the positive and negative impacts of each feature in different activity scenarios; The fusion analysis and application module is used to compare the SEM path coefficient with the SHAP value, identify the key factors that play an important role in the path relationship and are also highly important in machine learning, and formulate targeted waste classification management strategies for different types of event activities.

8. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

10. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images