A model feature analysis method based on a building indoor daylight perception evaluation prediction model

Abstract

The application provides a model feature analysis method based on a building indoor daylight perception evaluation prediction model. The method comprises the following steps: step 1, collecting building indoor light environment feature data and subjective daylight perception evaluation results, and establishing a basic database of daylight perception evaluation; step 2, carrying out exploratory factor analysis and confirmatory factor analysis, and obtaining common factors of daylight perception; step 3, screening an optimal performance machine learning algorithm, and constructing a daylight perception evaluation prediction model; and step 4, using a SHAP method to perform feature interpretation analysis on the daylight perception evaluation prediction model, and obtaining the influence weight of light environment feature parameters and their interactive combinations on the daylight perception evaluation results. The daylight perception evaluation prediction model provided by the application not only has high prediction accuracy, but also has good model interpretability, can analyze the influence weight of various environmental factors on daylight perception evaluation, and improves the accuracy and scientificity of building natural lighting design.

Description

Technical Field

[0001] This invention belongs to the field of building interior natural lighting design technology, and in particular relates to a model feature analysis method based on a building interior daylight perception evaluation and prediction model. Background Technology

[0002] The WELL Building Standard, proposed by the International WELL Building Institute, points out that natural lighting has a substantial impact on users' mood, circadian rhythms, and work efficiency. Conducting user daylight perception assessment research not only enables accurate prediction and evaluation of the indoor lighting environment from the user's perspective, but the assessment results can also significantly improve the accuracy of building lighting design and decision-making efficiency.

[0003] Visual perception is a complex response mechanism in which the brain holistically processes, analyzes, and understands external information when the human eye is stimulated by it. Since visual perception is a necessary consideration in architectural lighting design, quantitative research methods need to be introduced into this field to study the patterns of user visual perception more scientifically and effectively. Quantitative research on visual perception can provide a deeper understanding of the impact of the building's interior environment on users' visual perception, helping to optimize architectural design decisions and improve the comfort and usability of the building's interior environment. In architectural design research, subjective questionnaires and semantic differential scales are currently the most widely used quantitative analysis methods.

[0004] Existing studies typically employ different subjective evaluation scales in their experiments. Because these questionnaires vary in their questioning style and are highly subjective and empirical, it is difficult to determine whether differences in sunlight perception patterns are influenced by the questioning style. Furthermore, researchers often use traditional statistical methods such as linear regression, correlation analysis, and factor analysis to study the relationship between single environmental factors and sunlight perception evaluation. This approach lacks consideration for the potentially complex nonlinear relationships between environmental factors and evaluation results, and makes it difficult to determine the impact of combined environmental factors on sunlight perception evaluation results. Consequently, it fails to provide designers with accurate support for daylighting design decisions. Summary of the Invention

[0005] The purpose of this invention is to address the problems that existing studies on daylight perception evaluation cannot fully reflect users' subjective feelings and that the influence weights of environmental parameters are unclear. This invention proposes a model feature analysis method based on a prediction model for indoor daylight perception evaluation in buildings.

[0006] This invention is achieved through the following technical solution: This invention proposes a model feature analysis method based on a building indoor daylight perception evaluation and prediction model, the method comprising the following steps:

[0007] Step 1: Conduct a daylight perception evaluation experiment in a typical building space, collect characteristic parameters of the building's indoor light environment and subjective daylight perception evaluation results, and establish a basic database for daylight perception evaluation.

[0008] Step 2: Call the data from the basic database of sunlight perception evaluation, conduct exploratory factor analysis and confirmatory factor analysis to obtain common factors of sunlight perception;

[0009] Step 3: Based on the classification results of common factors of sunlight perception, select the best-performing machine learning algorithm and construct a sunlight perception evaluation and prediction model;

[0010] Step 4: Use the SHAP method to perform feature interpretation analysis on the sunlight perception evaluation prediction model. Calculate the average SHAP value of each light environment feature parameter, the SHAP value of the light environment feature parameters of typical samples, and the SHAP value of the interaction combination of light environment feature parameters to obtain the influence weight of the light environment feature parameters and their interaction combination on the sunlight perception evaluation results.

[0011] Furthermore, step 1 specifically includes:

[0012] Step 1.1: Use questionnaires and field surveys to obtain typical spatial environment information for different building types;

[0013] Step 1.2: Conduct a daylight perception evaluation experiment in a typical building space, collect characteristic parameters of the building's indoor light environment, and collect subjective daylight perception evaluation results;

[0014] Step 1.3: Based on the building space type and lighting form, filter and classify the light environment characteristic parameters and the daylight perception evaluation results to establish the basic database of the daylight perception evaluation model.

[0015] Furthermore, the sunlight perception evaluation consists of 15 sets of bipolar adjectives evaluating the characteristics of the indoor light environment, namely: dim-bright, low contrast-high contrast, distinct-blurred, uniform-non-uniform, stable-variable, simple-complex, continuous-discontinuous, orderly-random, interesting-uninteresting, comfortable-uncomfortable, satisfactory-unsatisfactory, pleasant-unpleasant, tense-relaxing, inhibitory-stimulating, attractive-unattractive.

[0016] Further, step 2 specifically involves: conducting reliability and validity analysis on the sunlight perception evaluation results; using principal component analysis in exploratory factor analysis to extract potential common factors among the sunlight perception evaluation items; conducting confirmatory factor analysis on the common factors; and, based on the results of convergent validity, discriminant validity, and construct validity analysis, using the standard values ​​of the model fitting indicators as a reference, pruning and eliminating items with low standard loading coefficients until all indicators reach the threshold standard, thereby obtaining the common factors of sunlight perception.

[0017] Furthermore, step 3 specifically involves:

[0018] Step 3.1: Based on the classification results of common factors of sunlight perception, perform feature selection on the dataset in the basic database and preprocess the light environment feature parameters;

[0019] Step 3.2: Construct a sunlight perception evaluation prediction model using the decision tree algorithm, random forest algorithm, and XGBoost algorithm respectively. Compare the accuracy of the prediction models using model evaluation indicators and select the optimal prediction model.

[0020] Step 3.3: Using grid search and cross-validation, the hyperparameters of the prediction model are optimized to obtain an effective prediction model for sunlight perception evaluation.

[0021] Further, step 4 specifically involves: using the SHAP method to interpret the features of the sunlight perception evaluation prediction model, analyzing the influence of light environment feature parameters on the sunlight perception evaluation results, the influence of light environment feature parameters on the sunlight perception evaluation classification results, the influence of light environment feature parameters based on typical samples on the sunlight perception evaluation results, and the influence of light environment feature parameters and their interactive combinations on the sunlight perception evaluation results, in order to explore the quantitative relationship between light environment feature elements and sunlight perception evaluation results, thereby obtaining the influence weights of light environment feature parameters and their interactive combinations on the sunlight perception evaluation results.

[0022] Furthermore, the SHAP value is used to further explore the importance of input features to the prediction result, and the specific calculation formula is as follows:

[0023]

[0024] Where g(x′) represents the prediction model, x′∈{0,1} M φ represents the feature vector, M represents the number of simplified input features; i φ represents the SHAP value of feature i; in the feature vector, "1" indicates that the corresponding feature value "exists", and "0" indicates that it "does not exist"; φ0 represents the model output when all simplified inputs are turned off, i.e., missing.

[0025] Furthermore, the SHAP method is used to analyze the features of the dataset in the learning prediction model; when the XGBoost algorithm is used to construct a sunlight perception evaluation prediction model, the SHAP method is used to analyze the final XGBoost model fc(S); the SHAP method determines the SHAP value of all features by constructing a subset of features S, and the specific calculation formula is as follows:

[0026]

[0027] Where |S| represents the number of non-zero terms in S. Let S represent all vectors S in which the non-zero elements are subsets of the non-zero elements in M.

[0028] This invention proposes an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the model feature analysis method based on a building indoor daylight perception evaluation and prediction model.

[0029] This invention proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the model feature analysis method based on a building indoor daylight perception evaluation and prediction model.

[0030] The present invention has the following beneficial effects:

[0031] This invention proposes a model feature analysis method based on a prediction model for evaluating indoor daylight perception in buildings. This method obtains common factors in daylight perception through exploratory factor analysis and confirmatory factor analysis. These factors not only accurately reflect the focus of visual attention during daylight perception but also serve as a benchmark reflecting the fundamental laws of visual cognition, aiding in the selection of subjective evaluation question scales that best reflect the basic laws of daylight perception. Using machine learning modeling techniques and the SHAP feature interpretation method, the quantitative influence and correlation laws of light environment characteristic parameters and their interaction combinations on daylight perception evaluation are obtained. By establishing a prediction model for daylight perception evaluation and interpreting its features, the influence weights of various environmental factors in daylight perception evaluation are explored. This fills the strategic gaps in the theory and technology of daylight perception research, enabling in-depth exploration of the complex influences on daylight perception evaluation and improving the accuracy and scientific rigor of building natural lighting design. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0033] Figure 1 This is a flowchart of a model feature analysis method based on a building indoor daylight perception evaluation and prediction model, as described in this invention.

[0034] Figure 2 This is a diagram illustrating the interpretative analysis results of the sunlight perception evaluation model on the overall situation of light environment characteristic parameters.

[0035] Figure 3 This is a diagram illustrating the interpretative analysis results of the classification of the sunlight perception evaluation model based on the characteristic parameters of the light environment.

[0036] Figure 4 This is a diagram showing the interpretative analysis results of the sunlight perception evaluation model based on the light environment characteristic parameters of typical samples.

[0037] Figure 5 The figure shows the interpretative analysis results of the sunlight perception evaluation model based on the interactive combination of light environment characteristic parameters. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Combination Figure 1 This embodiment illustrates a model feature analysis method based on a building indoor daylight perception evaluation and prediction model, comprising the following steps:

[0040] Step 1: Conduct a daylight perception evaluation experiment in a typical building space, collect data on the characteristics of the indoor light environment and subjective daylight perception evaluation results, and establish a basic database for daylight perception evaluation.

[0041] Step 2: Call the data from the basic database of sunlight perception evaluation, conduct exploratory factor analysis and confirmatory factor analysis to obtain common factors of sunlight perception;

[0042] Step 3: Based on the classification results of common factors of sunlight perception, select the best-performing machine learning algorithm and construct a sunlight perception evaluation and prediction model;

[0043] Step 4: Use the SHAP (SHapley Additive exPlanations) method to perform feature interpretation analysis on the sunlight perception evaluation prediction model. Calculate the average SHAP value of each light environment feature parameter, the SHAP value of the light environment feature parameters of typical samples, and the SHAP value of the interaction combination of light environment feature parameters to obtain the influence weight of the light environment feature parameters and their interaction combinations on the sunlight perception evaluation results.

[0044] Step 1 specifically involves:

[0045] Step 1.1: Use questionnaires and field surveys to obtain typical spatial environment information for different building types;

[0046] Step 1.2: Conduct a daylight perception assessment experiment in a typical building space, collect data on the characteristics of the indoor light environment of the building, and collect subjective daylight perception evaluation results.

[0047] Step 1.3: Based on the building space type and lighting form, filter and classify the light environment characteristic data and the daylight perception evaluation results to establish the basic database of the daylight perception evaluation model.

[0048] The daylight perception evaluation consists of 15 sets of bipolar adjectives evaluating the characteristics of the indoor light environment, namely: dim-bright, low contrast-high contrast, distinct-blurred, uniform-non-uniform, stable-variable, simple-complex, continuous-discontinuous, orderly-random, interesting-uninteresting, comfortable-uncomfortable, satisfactory-unsatisfactory, pleasant-unpleasant, tense-relaxing, inhibitory-stimulating, attractive-unattractive.

[0049] The sunlight perception evaluation uses a 7-level semantic difference scale as the evaluation scale. Specifically, in each evaluation item, level 1 and level 7 correspond to two adjectives in a bipolar adjective group, and level 4 is a moderate state.

[0050] The specific indoor lighting environment characteristics parameters of the building include: building space characteristics parameters (building space function, building plan dimensions, main space dimensions, window orientation, number of windows, window-to-wall ratio), outdoor environmental meteorological parameters (solar altitude angle, solar azimuth angle, sky type), and natural lighting characteristics parameters (average brightness within the subject's line of sight, vertical eye illuminance, horizontal tabletop illuminance, daylight glare probability, and daylight glare index).

[0051] Step 2 specifically involves: conducting reliability and validity analysis on the sunlight perception evaluation results; using principal component analysis in exploratory factor analysis to extract potential common factors among the sunlight perception evaluation items; conducting confirmatory factor analysis on the common factors; and, based on the results of convergent validity, discriminant validity, and construct validity analysis, using the standard values ​​of the model fitting indicators as a reference, pruning and eliminating items with low standard loading coefficients until all indicators reach the threshold standard, thereby obtaining the common factors of sunlight perception.

[0052] Exploratory factor analysis (GFA) is a statistical method used to uncover potential structural relationships within an evaluation scale, process and reduce its dimensionality to obtain common factors. Before conducting GFA, reliability and validity tests need to be performed on the items to verify the validity of the evaluation results and whether there are correlations between the items.

[0053] Reliability testing uses the Cronbach's coefficient (alpha), which examines the reliability of the assessment content. A higher Cronbach's coefficient indicates higher questionnaire reliability. In practical analysis, a Cronbach's coefficient greater than or equal to 0.7 indicates a reliable questionnaire; a Cronbach's coefficient less than 0.4 indicates low reliability, making further factor analysis unsuitable. The formula for calculating the Cronbach's coefficient is:

[0054]

[0055] In the formula: α is the reliability coefficient, and K is the questionnaire assessment item. For the variation of scores of all participants on question i, The variance of the total score obtained by all participants.

[0056] Validity testing was conducted using the Kaiser-Meyer-Olkin Test (KMO) and Bartlett's Test of Sphericity. The KMO test is used to compare the simple correlation coefficients and partial correlation coefficients between questionnaire items. The formula for calculating the KMO test is:

[0057]

[0058] In the formula: r ij For the simple correlation coefficient, r ij·1,2…kThe KMO value represents the partial correlation coefficient. The KMO test ranges from [0,1]. A KMO value closer to 1 indicates a stronger correlation between the items, resulting in better factor analysis. In practical analysis, a KMO value greater than 0.7 indicates that the questionnaire content is suitable for factor analysis; a KMO value less than 0.5 indicates that the questionnaire content is not suitable for factor analysis.

[0059] Bartlett's test of sphericity aims to test whether the null hypothesis "the correlation matrix is ​​an identity matrix" or the alternative hypothesis "the correlation matrix is ​​not an identity matrix" is true. In practical analysis, if the p-value is <0.05, the null hypothesis is rejected and the alternative hypothesis is accepted, indicating that the correlation matrix is ​​not an identity matrix and is suitable for factor analysis; if the p-value is ≥0.05, the null hypothesis is accepted, indicating that the correlation matrix is ​​an identity matrix and the data is not suitable for factor analysis.

[0060] Principal Component Analysis (PCA) is a statistical method that recombines the original variables into a new set of independent variables, and, depending on the actual needs, extracts a smaller number of summed variables to reflect as much information as possible about the original variables.

[0061] Confirmatory factor analysis (CFA) is a research method used to examine whether the correspondence between common factors and question items is consistent with the researcher's expectations. The main purpose of CFA is to verify convergent validity, discriminant validity, and construct validity.

[0062] Convergent validity refers to the similarity of measurement results when different measurement methods are used to measure the same characteristic. The validation indices for convergent validity are combined reliability, average variance extracted (AVE), and factor loadings. Combined reliability is the square of the sum of factor loadings; the stronger the correlation between items, the stronger the explanatory power of the latent variable for them. The larger the square of the sum of factor loadings, the better the internal consistency. AVE is the overall explanatory power of the latent variable for all measured variables; that is, the higher the AVE value, the stronger the ability of the latent variable to explain its corresponding assessment item. Factor loadings are the standardized regression coefficients from latent variables to measured variables; the larger the factor loadings, the stronger the explanatory power of the latent variable for the measured variable. If the AVE value of each common factor is greater than 0.5, the CR value is greater than 0.7, and the factor loadings are greater than 0.7, it indicates that the common factor has good convergent validity.

[0063] Discriminant validity measures the ability of latent variables to differentiate between them. The indicators for validating discriminant validity are the root of Ave (AVE) square root value and the correlation coefficient. When the root of Ave square root value for each common factor is greater than the absolute value of the correlation coefficient between that factor and other factors, it indicates that the common factor has good discriminant validity.

[0064] Construct validity is used to test whether the correspondence between common factors and question items meets expectations. In practical analysis, commonly used construct validity fit indices and their standards are shown in Table 1.

[0065] Table 1 Commonly used fit indices for structural validity

[0066] Fit index Chinese name standard χ2 / df Chi-square degrees of freedom Less than 3 GFI Goodness-of-fit index Greater than 0.9 RMSEA Root mean square of approximation error Less than 0.05 SRMR Standardized root mean square error Less than 0.05 CFI Comparison of fit indices Greater than 0.9 NFI Standardized fit index Greater than 0.9

[0067] Step 3 specifically involves:

[0068] Step 3.1: Based on the classification results of common factors of sunlight perception, perform feature selection on the dataset in the basic database and preprocess the light environment feature data;

[0069] Step 3.2: Construct a sunlight perception evaluation prediction model using the decision tree algorithm, random forest algorithm, and XGBoost algorithm respectively. Compare the accuracy of the prediction models using model evaluation indicators and select the optimal prediction model.

[0070] Step 3.3: Using grid search and cross-validation, the hyperparameters of the prediction model are optimized to obtain an effective prediction model for sunlight perception evaluation.

[0071] Feature selection is the process of choosing a relevant subset of features to build a model in machine learning. This method can effectively remove redundant or irrelevant variables to find the optimal subset of features in the database.

[0072] The light environment feature data preprocessing process involves normalizing continuous light environment data and performing One-Hot encoding on discrete building spatial feature data (test location, window orientation, number of windows, etc.) to reduce the impact of different data variation amplitudes on the model prediction results.

[0073] The decision tree algorithm is a classification and regression algorithm based on a tree structure, which has advantages such as ease of understanding, strong interpretability, and insensitivity to missing values. The random forest algorithm is an ensemble learning algorithm based on decision trees; by combining multiple decision trees for classification or regression prediction, it can effectively improve prediction accuracy. The XGBoost (eXtreme Gradient Boosting) algorithm is an ensemble learning algorithm that uses a greedy algorithm and gradient information to optimize the loss function, exhibiting good robustness and adaptability to imbalanced datasets.

[0074] The model evaluation metrics are shown in Table 2:

[0075] Table 2 Model Evaluation Indicators

[0076]

[0077] The grid search method refers to the process of adjusting parameters sequentially within a specified parameter range, training the learner using the adjusted parameters, and finding the parameters with the highest accuracy on the validation set from all parameters. The basic principle of cross-validation is to repeatedly use data, dividing the dataset into different training and test sets, and repeatedly training, testing, and selecting models to accurately evaluate model performance and prevent overfitting or underfitting.

[0078] Step 4 specifically involves: using the SHAP (SHapley Additive exPlanations) method to interpret the features of the sunlight perception evaluation prediction model, analyzing the influence of light environment feature parameters on the sunlight perception evaluation results, the influence of light environment feature parameters on the sunlight perception evaluation classification results, the influence of light environment feature parameters based on typical samples on the sunlight perception evaluation results, and the influence of light environment feature parameter interactions on the sunlight perception evaluation results, in order to explore the quantitative relationship between light environment feature elements and sunlight perception evaluation results, thereby obtaining the influence weights of light environment feature parameters and their interactions on the sunlight perception evaluation results.

[0079] The SHAP value is used to further explore the importance of input features to the prediction results, and the specific calculation formula is as follows:

[0080]

[0081] Where g(x′) represents the prediction model, x′∈{0,1} M φ represents the feature vector, and M represents the number of simplified input features. i This represents the SHAP value of feature i. In the feature vector, "1" indicates that the corresponding feature value "exists", and "0" indicates that it "does not exist". φ0 represents the model output when all simplified inputs are turned off (i.e., missing).

[0082] The SHAP method can also analyze the features of the dataset used to learn the prediction model. Taking the XGBoost algorithm as an example, since the SHAP method is a feature interpretation method based on the prediction model, it is only used to analyze the final XGBoost model f. x (S). The SHAP method determines the SHAP value of all features by constructing a subset of features S. The specific calculation formula is as follows:

[0083]

[0084] Where |S| represents the number of non-zero terms in S. Let S represent all vectors S in which the non-zero elements are subsets of the non-zero elements in M.

[0085] The SHAP value can be used to explain the degree of influence of light environment characteristic parameters on the evaluation results of sunlight perception. For example, the SHAP method can be used to assess the degree of influence of light environment characteristic parameters on the evaluation results of sunlight satisfaction. Figure 2 As shown. L task / L m The impact on the daylight satisfaction assessment results is greatest, followed by DGP.

[0086] The average SHAP value can be used to explain the influence of light environment characteristic parameters on the classification results of sunlight perception assessment. For example, the SHAP method can be used to assess the influence of light environment characteristic parameters on the classification results of sunlight satisfaction. Figure 3 As shown. L task / L m The impact on the "satisfactory", "unsatisfactory" and "moderate" states varies. The impact on the "unsatisfactory" state of sunlight is significant, but the impact on the "moderate" state is relatively weak.

[0087] Using SHAP values ​​and decision analysis diagrams, the influence of typical sample light environment characteristic parameters on sunlight perception evaluation results can be explained. For example, the SHAP method can be used to explain the influence of typical sample light environment characteristic parameters on sunlight satisfaction evaluation results, such as... Figure 4 As shown in the figure, each line corresponds to a sample in the dataset, and the line color represents the predicted value of the feature, corresponding to the spectral band at the top of the figure. All lines start from the bottom and move towards the top. The SHAP value of each feature is added to the base value of the evaluation result, indicating the degree of contribution of each feature parameter to the overall prediction. The slope of the straight lines at gender and experimental location changes little across all sample data, indicating that these two feature parameters have a weak impact on the evaluation results of sunlight satisfaction.

[0088] The SHAP value can be used to explain the degree of influence of the interaction combination of light environment characteristic parameters on the evaluation results of sunlight perception. For example, the SHAP method can be used to assess the degree of influence of the interaction combination of light environment characteristic parameters on the evaluation results of sunlight satisfaction. Figure 5 As shown. Solar altitude angle and L task This is the interaction combination of characteristic parameters that has the greatest impact on the assessment of daylight satisfaction, and it includes L. task / L m The combination of characteristic parameters generally has a high impact on the evaluation results.

[0089] This invention proposes a model feature analysis method based on a prediction model for evaluating indoor daylight perception in buildings. This method obtains common factors in daylight perception through exploratory factor analysis and confirmatory factor analysis. These factors not only accurately reflect the focus of visual attention during daylight perception but also serve as a benchmark reflecting the fundamental laws of visual cognition, aiding in the selection of subjective evaluation question scales that best reflect the basic laws of daylight perception. Using machine learning modeling techniques and the SHAP feature interpretation method, the quantitative influence and correlation laws of light environment characteristic parameters and their interaction combinations on daylight perception evaluation are obtained. By establishing a prediction model for daylight perception evaluation and interpreting its features, the influence weights of various environmental factors in daylight perception evaluation are explored. This fills the strategic gaps in the theory and technology of daylight perception research, enabling in-depth exploration of the complex influences on daylight perception evaluation and improving the accuracy and scientific rigor of building natural lighting design.

[0090] This invention proposes an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the model feature analysis method based on a building indoor daylight perception evaluation and prediction model.

[0091] This invention proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the model feature analysis method based on a building indoor daylight perception evaluation and prediction model.

[0092] The memory in this application embodiment can be volatile memory or non-volatile memory, or it can include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DRRAM). It should be noted that the memory used in the methods described in this invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0093] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0094] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0095] It should be noted that the processor in the embodiments of this application can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiments can be completed by the integrated logic circuitry in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above methods.

[0096] The above provides a detailed description of the model feature analysis method based on the evaluation and prediction model of indoor sunlight perception in buildings proposed in this invention. Specific examples have been used to illustrate the principle and implementation of this invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A model feature analysis method based on a building indoor daylight perception evaluation and prediction model, characterized in that: The method includes the following steps: Step 1: Conduct a daylight perception evaluation experiment in a typical building space, collect characteristic parameters of the building's indoor light environment and subjective daylight perception evaluation results, and establish a basic database for daylight perception evaluation. Step 2: Call the data from the basic database of sunlight perception evaluation, conduct exploratory factor analysis and confirmatory factor analysis to obtain common factors of sunlight perception; Step 3: Based on the classification results of common factors of sunlight perception, select the best-performing machine learning algorithm and construct a sunlight perception evaluation and prediction model; Step 4: Use the SHAP method to perform feature interpretation analysis on the sunlight perception evaluation prediction model. Calculate the average SHAP value of each light environment feature parameter, the SHAP value of the light environment feature parameters of typical samples, and the SHAP value of the interaction combination of light environment feature parameters to obtain the influence weight of the light environment feature parameters and their interaction combination on the sunlight perception evaluation results. Step 3 specifically involves: Step 3.1: Based on the classification results of common factors of sunlight perception, perform feature selection on the dataset in the basic database and preprocess the light environment feature parameters; Step 3.2: Construct a sunlight perception evaluation prediction model using the decision tree algorithm, random forest algorithm, and XGBoost algorithm respectively. Compare the accuracy of the prediction models using model evaluation indicators and select the optimal prediction model. Step 3.3: Using grid search and cross-validation, the hyperparameters of the prediction model are optimized to obtain an effective prediction model for sunlight perception evaluation. Step 4 specifically involves using the SHAP method to interpret the features of the sunlight perception evaluation prediction model, analyzing the influence of light environment feature parameters on the sunlight perception evaluation results, the influence of light environment feature parameters on the sunlight perception evaluation classification results, the influence of light environment feature parameters based on typical samples on the sunlight perception evaluation results, and the influence of light environment feature parameters and their interactive combinations on the sunlight perception evaluation results. This aims to explore the quantitative relationship between light environment feature elements and sunlight perception evaluation results, thereby obtaining the influence weights of light environment feature parameters and their interactive combinations on the sunlight perception evaluation results.

2. The method according to claim 1, characterized in that: Step 1 is as follows: Step 1.1: Use questionnaires and field surveys to obtain typical spatial environment information for different building types; Step 1.2: Conduct a daylight perception evaluation experiment in a typical building space, collect characteristic parameters of the building's indoor light environment, and collect subjective daylight perception evaluation results; Step 1.3: Based on the building space type and lighting form, filter and classify the light environment characteristic parameters and the daylight perception evaluation results to establish the basic database of the daylight perception evaluation model.

3. The method according to claim 2, characterized in that: The sunlight perception evaluation consists of 15 sets of bipolar adjectives evaluating the characteristics of the indoor light environment, namely: dim-bright, low contrast-high contrast, distinct-blurred, uniform-non-uniform, stable-variable, simple-complex, continuous-discontinuous, orderly-random, interesting-uninteresting, comfortable-uncomfortable, satisfactory-unsatisfactory, pleasant-unpleasant, tense-relaxing, inhibitory-stimulating, attractive-unattractive.

4. The method according to claim 1, characterized in that: Step 2 specifically involves: conducting reliability and validity analysis on the results of the sunlight perception evaluation, and using principal component analysis in the exploratory factor analysis to extract potential common factors among the issues in the sunlight perception evaluation. Confirmatory factor analysis was conducted on the common factors. Based on the results of convergent validity, discriminant validity and construct validity analysis, and with the standard values ​​of the model fit indicators as a reference, problem items with low standard loading coefficients were eliminated through trial and error until all indicators reached the threshold standard, thus obtaining the common factors of sunlight perception.

5. The method according to claim 1, characterized in that: The SHAP value is used to further explore the importance of input features to the prediction results, and the specific calculation formula is as follows: in, Represents the predictive model. Let M represent the feature vector, and M represent the number of simplified input features. Representation of features i The SHAP value; in the feature vector, "1" indicates that the corresponding feature value "exists", and "0" indicates that it "does not exist"; This indicates the model output when all simplified inputs are turned off, i.e., missing.

6. The method according to claim 5, characterized in that: The SHAP method is used to analyze the features of the dataset used to learn the prediction model; when the XGBoost algorithm is used to construct a sunlight perception evaluation prediction model, the SHAP method is used to analyze the final XGBoost model. The SHAP method determines the SHAP value of all features by constructing a subset of features S. The specific calculation formula is as follows: in, This represents the number of non-zero terms in S. Let S represent all vectors S in which the non-zero elements are subsets of the non-zero elements in M.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-6.

8. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-6.

Images