Intelligent prediction method for dynamic evolution of human settlement heritage system

Abstract

This paper proposes an intelligent prediction method for the dynamic evolution of human settlement heritage systems. The method includes the following steps: establishing a model for the evolution of heritage material spatial morphology, a dynamic interaction model of multiple community agents, and a model for the dissemination and diffusion of cultural factors; then coupling these models to form a multimodal coupling model; and finally, predicting future scenarios based on the multimodal coupling model. This method enables time-series modeling and trend prediction of core elements of human settlement systems, such as spatial structure, demographic structure, and cultural activities.

Description

Technical Field

[0001] The present invention belongs to the technical field of human settlement heritage, and in particular relates to an intelligent prediction method for the dynamic evolution of human settlement heritage systems. Background Art

[0002] Human settlement heritage refers to settlements, neighborhoods, or architectural complexes that have gradually formed over the course of historical evolution and continue to serve residential functions and maintain a living cultural heritage. This type of heritage possesses a unique dual nature: it is both a legacy that carries historical memory and cultural values ​​and a vital spatial carrier of the daily lives of contemporary residents. Its core characteristic lies in the dynamic symbiotic relationship between "people, land, and culture," demonstrating the organic integration of historical heritage and real-life experiences. From a systems theory perspective, human settlement heritage can be viewed as a complex adaptive system comprised of three core elements: physical space, social groups, and cultural traditions. This system not only possesses structural complexity, encompassing multi-level, multi-element, and multi-scale spatial and social networks, but also exhibits remarkable adaptability, capable of flexibly adjusting its functions and structure in response to changes in the external environment to maintain the community's continued survival and cultural continuity. Furthermore, this system is a constantly evolving dynamic system, constantly undergoing renewal and reconstruction driven by multiple factors, including population replacement, functional transformation, and policy intervention. In recent years, my country has continuously strengthened its policy efforts to protect human settlement heritage, promoting the coordinated advancement of heritage conservation and community development. However, in practice, there are still two prominent challenges: the need to strictly maintain the authenticity and integrity of the heritage. Currently, traditional methods of protecting and renewing residential heritage have the following limitations:

[0003] (1) The evolution process lacks quantitative modeling and prediction mechanisms. Existing research mostly stays at the level of qualitative analysis. There is a lack of systematic modeling methods that can depict the dynamic changes of human settlement systems, and there is also a lack of data-driven evolution trend prediction methods, which makes it difficult to provide effective support for the scientific formulation of protection strategies.

[0004] (2) Weak data integration and analysis capabilities. The data of human settlement heritage systems often have complex characteristics across space (spatial distribution), time (historical evolution), and dimensions (population, industry, culture, etc.). Problems such as fragmented data sources, heterogeneous structures, and inconsistent semantics have seriously hindered the development of unified modeling and temporal analysis.

[0005] (3) The evolutionary driving mechanism is complex and the coupling relationship is unclear. The evolution of human settlement systems is influenced by multiple factors, such as policy orientation, population migration, economic structure changes, and cultural identity. Traditional analytical methods are difficult to reveal the nonlinear coupling relationships and deep-seated driving mechanisms behind them.

[0006] (4) Renewal interventions lack foresight and scientific support. In urban renewal and heritage protection practices, there is a lack of predictive analysis of future evolution trends, resulting in frequent problems such as irrational functional substitution, resident loss, and spatial landscape distortion, which seriously affect the continued vitality of human settlement heritage and the true continuation of cultural value. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent prediction method for the dynamic evolution of human settlement heritage systems. This method aims to realize the time series modeling and trend prediction of core elements of human settlement systems, such as spatial structure, population structure, and cultural activities, in a multi-source heterogeneous data environment. The technical solutions adopted are:

[0008] An intelligent prediction method for the dynamic evolution of a human settlement heritage system comprises the following steps:

[0009] Establish a model for the evolution of heritage material space, a dynamic interaction model of multiple community agents, and a model for the dissemination and diffusion of cultural factors;

[0010] Then, the heritage material spatial morphological evolution model, the community multi-agent dynamic interaction model, and the cultural factor diffusion model are coupled to form a multimodal coupling model.

[0011] Finally, based on the multimodal coupling model, future scenario patterns are predicted.

[0012] Preferably, the specific steps of establishing the heritage material spatial morphological evolution model include:

[0013] Step 1A: Collect multimodal data on human settlement heritage;

[0014] Step 1B: Analyze the material spatial form of residential heritage;

[0015] Step 1C: Construct a model of the evolution of the heritage material spatial form.

[0016] Preferably, step 1B specifically includes the following steps:

[0017] Step 1B1, macroscopic external space morphology analysis;

[0018] Step 1B2: Analysis of microscopic internal space morphology.

[0019] Preferably, step 1C specifically includes the following steps:

[0020] Step 1C1, establishing a macroscopic spatial morphological evolution prediction model;

[0021] Step 1C2: Establish a microscopic spatial morphological evolution prediction model;

[0022] Step 1C3: Couple the macroscopic spatial morphology evolution prediction model and the microscopic spatial morphology evolution prediction model to form a heritage material spatial morphology evolution model.

[0023] Preferably, the steps of establishing a dynamic interaction model of multiple community agents include:

[0024] Step 2A, design of specific indicator system;

[0025] Step 2B: Multi-source heterogeneous data collection and fusion;

[0026] Step 2C: Dynamic relationship network construction and modeling.

[0027] Preferably, step 2A includes indicators: neighborhood density index NDI, intangible cultural heritage chain length TCL and local employment ratio LER.

[0028] Preferably, the specific steps of establishing a cultural factor propagation and diffusion model include:

[0029] Step 3A: node semantic encoding;

[0030] Step 3B: Build a multicultural communication network;

[0031] Step 3C: Establish a cultural factor diffusion model based on the multicultural communication network.

[0032] Preferably, the cultural communication network in step 3B includes: a master-apprentice inheritance network, a festival participation network and a language diffusion network.

[0033] Preferably, the step of forming a multimodal coupling model includes:

[0034] Step 4A: Coupling mechanism design;

[0035] Step 4B: multi-modal coupling model structure design;

[0036] Step 4C: Design cross-modal interaction mechanism.

[0037] Compared with the prior art, the advantages of the present invention are:

[0038] The human settlement heritage system is a complex adaptive system comprised of three key elements: physical space, social groups, and cultural activities. These three elements each have their own independent evolutionary laws, yet they also dynamically interact to shape the system's overall evolutionary path. Therefore, this paper first independently models each of the three subsystems to capture their respective dynamic characteristics and evolutionary trends. Then, through coupling mechanism modeling, the interactions between the three elements are identified, thereby constructing a dynamic evolution prediction model for the overall system—a multimodal coupling model.

[0039] This multimodal coupling model integrates multi-source data fusion, spatiotemporal modeling, complex network analysis and intelligent prediction algorithms such as deep learning and system dynamics. It can simulate the state changes of human-inhabited heritage systems under specific interventions or natural evolution, providing a scientific basis for policy formulation and intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Technical roadmap for intelligent prediction methods for the dynamic evolution of human settlement heritage systems:

[0041] Figure 2 A flowchart for intelligent prediction of heritage material space evolution;

[0042] Figure 3 It is a schematic diagram of a dynamic relationship network;

[0043] Figure 4 Schematic diagram for parameter optimization;

[0044] Figure 5 Design schematics for multimodal coupling models;

[0045] Figure 6 Schematic diagram of the hierarchical modeling strategy for coupled structures;

[0046] Figure 7 This is a schematic diagram of the regional grid divided based on GIS;

[0047] Figure 8 A schematic diagram of the region boundary length extracted in QGIS;

[0048] Figure 9 Schematic diagram of the integrated network module. DETAILED DESCRIPTION

[0049] The following is a more detailed description of the intelligent prediction method for the dynamic evolution of human settlement heritage systems, using schematic diagrams. These schematic diagrams illustrate preferred embodiments of the present invention. It should be understood that those skilled in the art may modify the present invention described herein while still achieving its advantageous effects. Therefore, the following description should be understood as a general guide for those skilled in the art and not as a limitation of the present invention.

[0050] This paper proposes an intelligent prediction method for the dynamic evolution of human settlement heritage systems. Its overall technical framework is as follows: Figure 1 shown.

[0051] Based on the theory of human settlement heritage, this method constructs a technical system of "four-dimensional analysis-dynamic coupling", which includes the following core modules: analysis of the evolution of heritage material space morphology, analysis of the evolution of heritage community network structure, analysis of the spread of cultural factors, and multimodal dynamic coupling model.

[0052] Each module is organically integrated through a spatiotemporal graph neural network, ultimately forming a dynamic prediction system that supports multi-scenario simulation.

[0053] The physical space evolution analysis module employs a "space-ground-calendar" data acquisition strategy to construct a systematic spatiotemporal database. This module includes two models: one that predicts how physical space structural characteristics, such as integration and connectivity, change with tourist population pressure by constructing a long- and short-term memory network; and the other, a cellular automaton model that simulates the intensity and speed of spatial expansion during the external morphological expansion process.

[0054] In the heritage community network structure evolution analysis module, questionnaire survey data and multimodal data such as mobile phone signaling are integrated to propose a quantitative indicator system that aligns with the characteristics of heritage communities and quantitatively model community network characteristics. Furthermore, an agent-based modeling approach is introduced to simulate the mechanisms by which individual and group behaviors influence the evolution of community structure.

[0055] In the cultural factor propagation analysis module, the CultureVerse multimodal cultural understanding benchmark dataset is combined with oral history speech transcription and sentiment analysis technology to encode and express cultural genes; in the propagation mechanism modeling, the Bass diffusion model that considers the spatial attenuation effect is introduced to characterize the spatial diffusion process of cultural factors.

[0056] Finally, a multimodal dynamic coupling model integrating the three subsystems of material space, community structure and cultural communication is constructed. Through the spatiotemporal graph neural network, deep integration and dynamic coordination of cross-domain features are achieved, supporting the simulation and prediction of various future scenario models.

[0057] Step 1: Analyze the morphological evolution of the heritage material space.

[0058] The morphological evolution of heritage material space is a key component of the dynamic evolution of human settlement heritage systems. This is primarily reflected in changes in building density and form, the migration and transformation of spatial functional zones, the evolution of street pattern patterns, and the preservation, reconstruction, or disappearance of landscape structures. By introducing space syntax theory and systematically analyzing street accessibility and accessibility, we can effectively reveal the feedback mechanisms of human activities on spatial structure and quantify key spatial structural indicators such as integration and connectivity, thereby reflecting the vitality of heritage spaces and their dynamic evolution.

[0059] Step 1A: Collect multimodal data on human settlement heritage.

[0060] This paper analyzes the evolution of heritage material spatial morphology through a dual analytical perspective, combining macroscopic external spatial morphology with microscopic internal spatial morphology. To address this analytical need, a three-in-one data acquisition approach, encompassing "air, ground, and history," is proposed. This approach utilizes modern drones, LiDAR scanning, and high-precision thermal infrared imaging technologies to achieve comprehensive and detailed data collection of heritage spaces.

[0061] Step 1A1: Collect macroscopic external space morphological data.

[0062] The macro level focuses on the large-scale architectural form and spatial pattern of the heritage area. The present invention uses the DJI M300 RTK drone equipped with a five-lens oblique photography system as the acquisition device. The device can achieve a ground resolution of 2.5 cm at a flight altitude of 150 meters, and uses PPK post-differential positioning technology to control the plane accuracy error within 3 cm. By generating a three-dimensional model through multi-angle high-resolution image acquisition, macro-spatial information (macro external spatial morphological data) such as the built-up area, average annual expansion area, base period area and grid side length of the heritage area can be obtained, providing high-quality data support for subsequent modeling and analysis. Regional grids divided based on GIS, such as Figure 7 shown.

[0063] Specifically, the macroscopic external space morphology data include: Annual built-up area , time interval , average annual expansion area , base period area , grid side length , the minimum number of grids covering the heritage boundary at this scale , regional plane projection area , area boundary length .

[0064] The length of the region boundary extracted based on QGIS is shown as follows Figure 8 shown.

[0065] Step 1A2: Collecting microscopic internal space morphology data.

[0066] At the microscopic level, this invention combines a handheld LiDAR with an 8K global camera to achieve detailed modeling of building interiors. The handheld LiDAR, with a scanning rate of up to 650,000 points per second, efficiently acquires high-density point cloud data, meeting the requirements for 3D modeling of complex interior structures. Furthermore, the 8K resolution global camera and a thermal infrared imager with a thermal sensitivity of 0.03°C enable not only precise image capture of building material textures but also the detection of thermal anomalies in the building structure, assisting in the identification of potential structural defects. Through the coordinated collection of multi-source data, a comprehensive reconstruction of the heritage site's interior space is achieved, resulting in a highly accurate, semantically rich 3D spatial model (microscopic internal spatial morphological data).

[0067] Among them, the micro-internal space morphological data include: roof form, bracket structure, door and window style, wall material, wall damaged area, architectural image texture data, three-dimensional point cloud data, historical building image data set, noise vector, weathering and lighting condition parameters.

[0068] Step 1A3: Data integration and spatiotemporal database construction.

[0069] To achieve efficient management and continuous updating of heritage material spatial data, this invention leverages local professional resources, connects to existing three-dimensional spatial database systems, and integrates historical data with real-time data collection to build a unified, full-time and full-spatial data management platform. This platform supports the storage, retrieval, and analysis of multi-source, heterogeneous data, providing a solid data foundation for subsequent dynamic evolution modeling and predictive analysis.

[0070] Step 1B: Analyze the material space form of residential heritage.

[0071] By systematically processing and analyzing the collected multimodal data, this paper constructs a quantitative indicator system at both macro and micro levels to reveal the dynamic evolution of the material spatial form of residential heritage. The macro level focuses on the spatial expansion characteristics at the regional scale, using a four-dimensional indicator system: expansion rate, expansion intensity, outline fractal dimension, and shape compactness. The micro level focuses on the internal characteristics of buildings, using architectural style and texture characteristics to describe them.

[0072] Step 1B1: Macroscopic external spatial morphology analysis. This is to obtain the output dataset for model training.

[0073] (1) Land expansion rate is used to measure the average annual growth rate of the built-up area of ​​the heritage area, reflecting the overall scale and development trend of spatial expansion. The calculation formula is:

[0074] (1)

[0075] in, Indicates the expansion speed, Indicates the Annual built-up area, Indicates a time interval (years);

[0076] (2) Land expansion intensity reflects the relative change in the average annual expansion per unit base period area, and describes the intensity level of the heritage area's spatial expansion. The calculation formula is:

[0077] (2)

[0078] in, Indicates the expansion strength, represents the average annual expansion area, represents the base period area;

[0079] (3) The outline fractal dimension is based on fractal geometry theory and measures the complexity and spatial fragmentation of the heritage area boundary morphology. The grid counting method is often used for estimation. The specific formula is as follows:

[0080] (3)

[0081] in, represents the contour fractal dimension, represents the grid side length, It is expressed as the minimum number of grid cells required to cover the boundaries of the property at that scale. The grid edge length is not a single fixed value, but a dynamic value, covering multiple scales from coarse to fine.

[0082] (4) The shape compactness index is used to evaluate the geometric compactness of the heritage area, usually reflecting the efficiency of space utilization and the characteristics of the development stage. Its expression is:

[0083] (4)

[0084] in represents the compactness index, represents the plane projection area of ​​the region, Indicates the length of the region boundary. In actual process, Figure 8 Read the area and length of the region boundary from .

[0085] Step 1B2: Microscopic internal spatial morphology analysis. This is to obtain the output dataset for model training.

[0086] (1) Architectural style analysis: Typical components and architectural style elements within the heritage area are extracted through roof form, bracket structure, door and window style, wall material, and wall damage area, and their style characteristics are identified. Combining deep learning with image recognition technology such as CNN networks can achieve automatic recognition of architectural styles and prediction of their evolution trends.

[0087] (2) Architectural texture analysis: Gray-level co-occurrence matrix (GLCM) is used to extract texture feature parameters such as contrast, energy, and entropy from architectural image texture data, and combined with machine learning algorithms such as random forest and support vector machines to achieve quantitative expression and temporal evolution modeling of building surface material texture.

[0088] The output dataset used for training the micro-spatial morphological evolution prediction model includes architectural style features and architectural texture features.

[0089] Step 1C: Construct a model of the evolution of the heritage material space form.

[0090] Through quantitative analysis of the material spatial form of heritage, a multi-level prediction model is constructed to accurately predict the future evolution trend of spatial form, providing scientific basis and technical support for heritage protection and renewal.

[0091] Step 1C1: Prediction of macroscopic spatial morphology evolution: Establish a macroscopic spatial morphology evolution prediction model.

[0092] Spatial evolution at the macro level is typically reflected in the dynamic changes of time-series indicators such as spatial expansion speed, expansion intensity, outline fractal dimension, and shape compactness. These indicators have obvious temporal correlation and nonlinear variation characteristics, making them suitable for time series modeling.

[0093] Long-short-term memory networks (LSTMs) have excellent capabilities for modeling long-term dependencies, effectively capturing the historical trajectory and trend characteristics of heritage spatial morphological evolution. Furthermore, the introduction of an attention mechanism automatically identifies and focuses on the historical moments most critical to current predictions, thereby enhancing the model's ability to perceive key evolutionary nodes and overcoming the "equal weight distribution" problem of traditional LSTMs in processing historical information.

[0094] Based on this, the LSTM-ATT model that integrates LSTM and Attention mechanisms can accurately predict core indicators such as the future spatial expansion speed, expansion intensity, and morphological complexity of the heritage area.

[0095] In summary, the macroscopic spatial morphology evolution prediction model, i.e., the trained LSTM-ATT model, takes as input the macroscopic external spatial morphology data collected in step 1A1, and outputs the expansion speed, expansion intensity, contour fractal dimension, and shape compactness.

[0096] Step 1C2: Prediction of microscopic spatial morphological evolution: Establish a prediction model for microscopic spatial morphological evolution.

[0097] (1) Prediction of architectural style evolution: that is, establishing an architectural style evolution prediction model.

[0098] As a visual and culturally recognizable spatial feature, architectural style evolves under the influence of multiple factors, including historical period, cultural context, material aging, and climate. Style prediction is essentially a comprehensive problem of image style generation and classification.

[0099] This model integrates three network modules: StyleGAN2, StyleNet3D, and EfficientNetB7. This model is also called an integrated network module. Figure 9 shown.

[0100] StyleGAN2 is used to generate architectural images with temporal evolution characteristics, simulating natural or human influences such as weathering and lighting changes; the input is a historical architectural image dataset, a noise vector, and weathering and lighting condition parameters, and the output is an architectural image with temporal evolution characteristics.

[0101] StyleNet3D is used to extract architectural style features from the 3D point cloud data in step 1A2 and model its evolution. The input is the 3D point cloud data and segmentation labels for components such as roofs and brackets. The output is a 3D style feature vector.

[0102] EfficientNetB7 is used to fine-tune top-level parameters based on image feature transfer learning, further improving the accuracy of style classification and evolution prediction. The input is architectural images with temporal evolution characteristics generated by StyleGAN2 and 3D style feature vectors generated by StyleNet3D. The output is the detailed style classification result.

[0103] Through the above multi-model integration strategy, the model's ability to perceive and predict the evolution trend of architectural styles is significantly enhanced.

[0104] Architectural style evolution prediction model, i.e., trained integrated network module.

[0105] (2) Prediction of architectural texture evolution: that is, establishing a prediction model for architectural texture evolution.

[0106] The evolution of building surface texture is usually driven by multiple factors such as natural weathering, artificial restoration, and functional reuse, showing significant temporal and structural change characteristics. The gray-level co-occurrence matrix (GLCM) can be used to extract a series of quantifiable texture indicators such as contrast, homogeneity, energy, and entropy to reflect the detailed changes in building surface materials.

[0107] Based on the multi-period architectural image texture data from Step 1A2, this paper constructs a GLCM-RF (Gray Level Co-occurrence Matrix-Random Forest) prediction model to further explore the evolution of building surface textures and achieve regression prediction of texture characteristics for future time periods. The model's prediction results can be further used to simulate the aging process of building materials, structural wear and tear, and visual decline trends, providing technical support for heritage reconstruction and visual restoration.

[0108] The building texture evolution prediction model, i.e. the trained GLCM-RF prediction model, takes building image texture data as input and outputs building texture features in the future time period.

[0109] In summary, the micro-space morphology evolution prediction model includes the architectural style evolution prediction model and the architectural texture evolution prediction model.

[0110] Step 1C3: Implementation of the comprehensive prediction system. This involves coupling the macroscopic spatial morphology evolution prediction model and the microscopic spatial morphology evolution prediction model to form a heritage material spatial morphology evolution model.

[0111] This system employs a "macro-micro" collaborative modeling strategy to construct an integrated system for predicting the spatial evolution of heritage materials. The macro model focuses on analyzing spatial structure and morphological evolution trends, while the micro model focuses on modeling the dynamic changes in architectural features and material details. These two models are integrated through a coupling module to achieve global collaborative prediction across spatial scales and levels of features.

[0112] The system has good interpretability, predictability and visualization capabilities, and can be widely used in scenarios such as heritage protection strategy formulation, evolution trend analysis and simulation and restoration, providing technical support for intelligent and scientific heritage management decisions. The overall framework of the intelligent prediction process of heritage material space evolution is as follows: Figure 2 As shown. First, the data is input and preprocessed. Spatiotemporal alignment is used to align the preprocessed data in time and space to ensure that data from different sources can be accurately matched and associated in time and geographical location. Feature engineering extracts, selects, transforms, and creates new features from the aligned data to improve the model's predictive ability and performance. The macro-spatial morphology prediction module uses the data processed by feature engineering to predict the spatial morphology at the macro level. Micro-spatial morphology prediction is carried out in parallel with the macro-prediction and also uses the data processed by feature engineering to predict the spatial morphology at the micro level. The results of the macro-spatial morphology prediction and the micro-spatial morphology prediction are integrated and fused to obtain a more comprehensive and accurate final prediction result. Finally, the fused prediction results are displayed in a three-dimensional visualization to facilitate user understanding and analysis.

[0113] The overall system adopts a modular + microservice architecture, encapsulating various prediction models into independent microservice modules to achieve high scalability, easy deployment and flexible upgradeability of the system.

[0114] At the data layer, the system uses distributed file systems such as HDFS and MinIO to store raw data such as images, point clouds, and gray-level co-occurrence matrices. It also uses space-time composite databases such as PostGIS + TimescaleDB to manage historical and real-time dynamic data to ensure the timeliness and spatial consistency of data access.

[0115] In the model service layer, each prediction model is encapsulated as a microservice, supporting service calls through RESTful or gRPC interfaces, including:

[0116] Microservice A is the LSTM-ATT time series prediction module. It inputs macroscopic external spatial morphological data, uses the LSTM network to model long-term evolution trends, and combines the attention mechanism to identify key time nodes. It outputs the predicted values ​​of parameters such as the profile fractal dimension and compactness index for several time periods in the future.

[0117] Microservice B is the architectural style generation and classification module. It uses StyleNet3D and StyleGAN2 to model historical architectural style features and generate images, simulating style evolution and weathering effects. It then integrates the EfficientNet network to perform style recognition and weathering stage classification on the generated images, improving the model's visualization capabilities and accuracy. The inputs are historical building images, 3D point cloud data, control parameters such as weathering degree and lighting angle, 3D point cloud data, and segmentation labels for components such as roofs and brackets. The output is the architectural style classification result, such as the Qing Dynasty official style.

[0118] Microservice C is a texture evolution regression and synthesis module. It uses GLCM to extract building surface texture features (such as contrast, entropy, and energy). It then uses a random forest model to regress and predict the future evolution of these texture features. It also synthesizes aging effect images to achieve a time-series simulation of the visual degradation of building materials. The input is high-resolution surface images and a time series of historical texture parameters collected by the historical heritage community. The output is a synthesized aging effect image and texture evolution prediction values.

[0119] Finally, the material space characteristics are formed by integrating the macroscopic morphological prediction results with the evolution trend of microscopic style and texture, and the material space characteristics construct a unified spatial situation simulation map.

[0120] The system uses the output of all microservices, such as the macro prediction of microservice A, the style probability of microservice B, and the texture parameters of microservice C, as input and models them using a graph neural network (GNN). Nodes are morphological indicators, style labels, and texture features; edges are dynamic association strengths, such as the probability of a decrease in compactness or an increase in texture entropy. The output is a simulation map of the heritage space situation, which explores the dynamic association path between "morphology-style-texture" and reveals its implicit evolutionary logic chain.

[0121] Step 2: Analysis of the evolution of the heritage community network structure.

[0122] Based on Complex Adaptive Systems (CAS) theory, this paper proposes a three-dimensional analytical framework: "environment-individual-interaction." This framework systematically models the dynamic evolution of multiple actors within heritage communities across social relationships, cultural heritage, and economic activities. By integrating heterogeneous data from multiple sources, this model reveals the evolutionary mechanisms of community network structures, providing a scientific basis for understanding and intervening in group behavior in cultural heritage conservation.

[0123] Step 2A: Design of specific indicator system.

[0124] In order to quantitatively characterize the evolutionary characteristics of the network structure of heritage communities, three core indicator systems are constructed, which are developed from the three dimensions of community structure, cultural heritage, and economic resilience.

[0125] (1) The Neighborhood Density Index (NDI) is used to quantify the spatial social density of residents in heritage communities and reflects the ability of the physical spatial layout to support social interaction. It is defined as follows:

[0126] (5)

[0127] in, Indicates residents and The strength of social connections between Indicates the distance between residents’ living spaces, represents the distance decay function, Represents the effective social interaction area.

[0128] (2) The length of the intangible cultural heritage chain (TCL) is used to measure the length of intergenerational transmission of an intangible cultural heritage from its original inheritors to contemporary practitioners. The longer the intergenerational span, the deeper the cultural heritage. Its definition is as follows:

[0129] (6)

[0130] in, represents the maximum traceability algebra, Indicates the The effectiveness of generational inheritance, represents the generational weight.

[0131] (3) Local Employment Rate (LER) reflects the degree of continuity of traditional production and lifestyle in heritage areas and is defined as follows:

[0132] (7)

[0133] in, Indicates the number of local residents engaged in traditional related jobs, It represents the total employed population within the heritage protection area, obtained through data collected through the following social questionnaire survey and other methods.

[0134] The neighborhood density index is used to calculate the real-time spatial interaction distance in step 2C and dynamically update the adjacency matrix weights.

[0135] The length of the intangible cultural heritage inheritance chain is used to construct the inheritance relationship map in step 2C.

[0136] The local employment ratio is used in the social relationship statement in step 2C to quantify the strength of social connections.

[0137] Step 2B: Multi-source heterogeneous data collection and fusion.

[0138] The key players in heritage communities include residents and tourists. These two groups have differences and tensions in economic interests, values, and development aspirations, forming a complex and dynamically evolving social network system. To support the modeling and evolutionary analysis of the aforementioned characteristic indicators, a multi-scale, multi-level crowd behavior data system must be constructed, encompassing the following two core components:

[0139] Resident data collection: The social network structure of residents is obtained through structured methods (such as social relationship questionnaires) to model the social relationships and interaction patterns among residents.

[0140] Visitor Dataset: Utilizing ticket scanning records at tourist attractions and camera video analysis, such as crowd flow statistics and heat maps, this dataset captures the spatial movement and aggregation dynamics of visitors at heritage sites, reflecting their visit patterns and potential impact on heritage sites.

[0141] Semantic Data Integration: This integrates semantic attributes from resident data, including cultural identity tags, intangible cultural heritage participation, and spatial cognitive maps, to construct a social relationship database encompassing both temporal and spatial dimensions. This database serves as the core foundation for modeling and predicting the evolution of heritage community networks, enabling analysis of the evolutionary process from the individual-to-relationship-to-group level.

[0142] The data in step 2B, such as the resident dataset and tourist dataset, are used to capture the spatiotemporal trajectories and face-to-face contact records in the dynamic relationship network in step 2C.

[0143] Step 2C: Dynamic relationship network construction and modeling implementation. That is, building a dynamic interaction model of multiple community agents.

[0144] 1. Model framework composition

[0145] The present invention is based on agent-based modeling (ABM) technology to simulate the dynamic interaction process of multiple types of agents in heritage communities.

[0146] The model regards key participants such as residents, tourists, intangible cultural heritage inheritors, and policymakers as "intelligent agents" with autonomous decision-making capabilities. They make decisions and interact in the heritage space based on their respective behavioral rules such as economic incentives, cultural preferences, and social intentions, forming an evolvable dynamic relationship network.

[0147] The modeling framework mainly consists of three parts: environment layer, individual layer and interaction layer.

[0148] The environmental layer describes the spatial structural basis of the heritage site, covering spatial topological relationships, building function distribution, transportation accessibility network, layout of intangible cultural heritage inheritance points, and spatiotemporal semantic layers of historical evolution, etc., to construct the spatial context of agent activities.

[0149] The individual layer mainly includes three types of subjects: community residents, tourists, and intangible cultural heritage inheritors. Each type of subject has autonomous attributes and behavioral strategies, and their behavior is driven by multiple factors such as cultural identity, social ties, and economic returns.

[0150] The interaction layer simulates the social relationship network between agents, the transmission path of cultural elements and the response behavior of individuals to the spatial environment, capturing the evolution of social networks and the process of cultural diffusion. The structure of this dynamic network is as follows: Figure 3As shown in the figure, signaling positioning primarily relies on various signaling data generated by mobile communication networks when users use their mobile phones. This data is originally used for network management and communication, but through specific technologies and algorithms, users' geographic location information can be extracted from it to generate or update spatiotemporal trajectories. These trajectories represent the movement paths of entities in time and space. This spatiotemporal trajectory information is then transmitted and integrated into the relationship graph. Questionnaire data is collected primarily through a series of designed questions from respondents. After processing, the questionnaire data is used to generate social relationship statements. Sensor data refers to the real-time or near-real-time perception of the environment or object status through physical sensor devices. For example, smart bracelets and GPS track recorders capture residents' activity patterns and spatial behavior in real time. This data is used to identify and record face-to-face contact events (face-to-face contact records) between entities. These records reveal close-range physical interactions. After the spatiotemporal trajectories, social relationship statements, and face-to-face contact records are generated, they work together to dynamically update the adjacency matrix, which is continuously adjusted and optimized based on the real-time information received.

[0151] (1) Spatial interaction distance Based on the geographical distance between residents and the neighborhood density index, the specific formula is defined as follows:

[0152]

[0153] in Indicates residents and The geographical distance between Neighborhood density index

[0154] (2) Inheritance relationship map Based on the length of the intangible cultural heritage inheritance chain, the specific formula is defined as follows:

[0155]

[0156] in, Indicates the length of the intangible cultural heritage chain

[0157] (3) Social Relationship Statement Based on the proportion of local employment, the specific formula is defined as follows:

[0158]

[0159] in, represents the local employment share of entity i.

[0160] (4) Social connection strength Based on the social relationship declaration, the specific formula is defined as follows:

[0161]

[0162] 2. Parameter optimization and model calibration

[0163] To improve the authenticity and prediction accuracy of the simulation, this paper introduces a parameter optimization and learning mechanism. By using historical survey data and behavioral trajectory data, the prior distribution of behavioral rule parameters is constructed, and it is continuously optimized iteratively during the simulation process. The specific implementation includes:

[0164] Introduce intelligent algorithms such as particle swarm optimization (PSO) or genetic algorithm (GA) to adjust key parameters in behavioral rules; implement multiple rounds of "simulation-verification-correction" mechanism, compare simulation results with the actual community evolution history, and continuously adjust the model to improve its fit and interpretability; support parameter sensitivity analysis and uncertainty quantification to ensure that the model has stable response capabilities when facing different scenarios and policy interventions. Figure 4 As shown in the figure. First, a historical dataset is input to drive the subsequent parameter search and model optimization. Then, a set of parameters is selected from a predefined parameter space for evaluation, providing an initial exploration range for Bayesian optimization. Bayesian optimization is an efficient global optimization algorithm that constructs a Gaussian process to fit the objective function and uses an acquisition function to determine the next evaluation point, aiming to find the optimal solution in the fewest possible iterations. After Bayesian optimization generates a set of parameters, these parameters are input into a simulator or model to generate the corresponding simulation results. The simulation results are compared and evaluated with the observed "real community structure." The root mean square error (RMSE) is used to measure the deviation between the model prediction and the true value. When the RMSE is less than 0.1, the simulation results match the real community structure closely enough and meet the requirements. When the "Comparison to Real Community Structure" result meets the RMSE requirement of <0.1, the current parameter set is considered the optimal parameter set.

[0165] Step 3: Analysis of the spread of cultural factors.

[0166] This paper considers the cultural transmission process within a human-inhabited heritage system as a complex adaptive system. The diverse individuals within this system, such as residents, families, mentors, and tourists, interact with each other based on factors such as geographic location, social networks, cognitive differences, and willingness to absorb culture, exhibiting systemic characteristics such as nonlinearity, emergence, and self-organizing evolution.

[0167] Combining classical cultural communication theory with diffusion dynamics models, a three-dimensional modeling method of "structure-semantics-behavior" is constructed. The communication path is simulated at the network structure layer, cultural factors are encoded at the semantic layer, and the communication probability function is defined at the behavioral layer, thereby realizing the spatiotemporal diffusion modeling of cultural elements and the prediction of future trends.

[0168] This model consists of three core building blocks:

[0169] The node semantic coding layer is used to structure and encode the cultural factors and their transmission carriers into semantic vectors;

[0170] The multi-layer communication network construction layer is used to construct multiple communication paths such as "mentor-apprentice relationship network", "festival participation network", and "language diffusion network".

[0171] The diffusion evolution simulation layer introduces an improved Bass diffusion model to simulate the evolution of cultural factors in the network, taking into account dynamic factors such as imitation propagation and spatial attenuation.

[0172] Step 3A: Node semantic encoding.

[0173] This model divides the cultural factors of heritage communities into two categories: explicit factors and implicit factors. Explicit factors include traditional crafts and festival rituals, while implicit factors include values ​​and oral traditions.

[0174] For the traditional processes in the explicit factors, a "process knowledge graph" can be constructed, where nodes represent process operation steps, edges represent operation dependency paths, and node attributes include materials used, tools, process difficulty, etc.

[0175] Construct a "space-time behavior pattern matrix" for festival rituals, encoding participants, time nodes, spatial locations, and behavior sequences.

[0176] For the implicit factors of values ​​and oral traditions, a dual encoding method of "semantic embedding + pragmatic labeling" is adopted. BERT is first used to semantically embed the text, and then cultural pragmatic attribute labels are introduced to enhance its contextual expression ability.

[0177] Step 3B: Construct a multicultural communication network.

[0178] To comprehensively depict the cultural communication paths and mechanisms in the human settlement heritage system, this paper constructs the following three types of cultural communication network layers based on the heterogeneity of cultural subjects and communication media:

[0179] 1. Master-apprentice inheritance network.

[0180] The master-apprentice inheritance network mainly simulates the "point-to-point" intergenerational inheritance path of intangible culture such as traditional crafts, opera, and folk skills in the master-apprentice relationship.

[0181] Each node represents an individual with a heritage relationship, and its attributes include the level of the inheritor (national, provincial, etc.), skill category, active years, and geographic location. Edges represent the existence of skill transmission. Edge weight It is defined as the imparting intensity index, which is defined as follows:

[0182] (8)

[0183] in, Indicates the number of years of teaching. Indicates the frequency of teaching, 、 Indicates the adjustment parameter.

[0184] 2. Festival participation network.

[0185] The festival participation network mainly depicts the collective participation behavior of community residents in festival activities, highlighting their emotional connection and sharing.

[0186] Nodes are families, community organizations, or festival organizers. Edges represent two nodes that have a record of participating in a festival together. The edge weight is defined as follows:

[0187] (9)

[0188] in, K Indicates the number of festivals. Represents family 、 The number of joint participations in k festivals. Indicates the cultural weight of Qing.

[0189] 3. Language diffusion network.

[0190] Used to simulate the spread and evolution of language and culture such as dialects, slang and oral traditions in geographical space.

[0191] Nodes represent language geographical units such as towns, streets, and dialect areas. Attributes include population size, language type, and frequency of use. Edges represent communication relationships. Edge weights The definition is as follows:

[0192] (10)

[0193] in, Represents the vocabulary similarity rate, which can be calculated by Jaccard coefficient or edit distance. represents the population mobility ratio, , Indicates the adjustment parameter.

[0194] Step 3C: Cultural diffusion simulation and model enhancement strategy.

[0195] Based on the above communication network, the improved Bass diffusion model is used to simulate the evolution and diffusion process of cultural factors:

[0196] (11)

[0197] in, Indicates the deadline The number of cultural factor adopters;

[0198] represents the maximum size of the potential adoption group;

[0199] represents the innovation adoption coefficient;

[0200] represents the imitation propagation coefficient, represents the spatial propagation attenuation factor;

[0201] Represents geographical distance, Represents the propagation resistance in the simulated space or social network.

[0202] Formula (11) is the cultural factor diffusion model.

[0203] The mentor-apprentice network (Formula 8) provides the formula in Formula 11 and The longer the teaching period, the more systematic the skills, the higher the imitation transmission coefficient q, the faster the teaching frequency, and the maximum size of the adoption group. the bigger it is;

[0204] Here (Formula 8) provides the formula 11 and Relationship:

[0205]

[0206]

[0207] represents the maximum value of all edge weights in the network, Indicates the basic scale (through historical data statistics, such as the peak number of people covered by the skill in the past)

[0208] The Festival Participation Network (Formula 9) provides the The more times they celebrate together, the easier it is for them to imitate each other, and the imitation propagation coefficient The higher it is.

[0209] Here (Formula 9) provides the formula 11 The relationship:

[0210]

[0211] Represents the festival influence adjustment parameter, and the default setting is 0.3.

[0212] The language diffusion network (Formula 10) provides the and , the higher the vocabulary similarity, the smaller the cultural gap, the weaker the resistance to social network communication, and when population mobility is frequent, the innovation adoption coefficient is high.

[0213] Here (Formula 10) provides the formula 11 The relationship:

[0214]

[0215] To improve the performance of the model, the present invention implements the following three model enhancement strategies:

[0216] (1) Introducing the “Cultural Absorption Willingness Function” : Reflects the individual In time Modeling the adoption tendency of a certain cultural factor by combining behavioral psychological characteristics such as cognitive load and cultural preference

[0217] (2) Constructing heterogeneous propagation kernel functions : Dynamically adjust the communication path and intensity according to the type of cultural factors;

[0218] (3) Modeling of cultural loss mechanisms: Considering the actual loss mechanisms such as the interruption of inheritance and migration, the scale of dissemination Dynamically update and adjust with the propagation path map.

[0219] Step 4: Establish a multi-modal coupling model.

[0220] In the three steps above, we constructed a model for the evolution of heritage material spatial form, a model for the evolution of community networks (a dynamic interaction model of multiple community entities), and a model for the diffusion of cultural factors. However, in a human settlement heritage system, these three subsystems are interconnected and mutually influential, exhibiting significant coupling. Therefore, this paper proposes a multimodal coupling modeling method to predict the overall dynamic evolution of the system, which can be used to conduct multi-scenario simulations. The schematic diagram of the multimodal coupling model design is shown in the figure. Figure 5As shown in the figure. First, the input is a heritage material spatial morphological evolution model, a community multi-agent dynamic interaction model, and a cultural communication and diffusion model. These are coupled through the ST-GNN multimodal spatiotemporal network, effectively linking and integrating these different modalities and spatiotemporal information to understand the interactions and influences between them. One output of the coupled model is used to dynamically predict the "human settlement heritage system." This model can comprehensively consider material, social, and cultural factors to predict the future state and trends of heritage in the human settlement environment. Another output is to support "intervention simulation platform simulation." This model can serve as the basis for a simulation platform to test the impact of different intervention measures, such as policy adjustments and protection strategies, on the dynamic evolution of the heritage system, thereby providing support for decision-making.

[0221] Step 4A: Coupling mechanism design.

[0222] The present invention constructs a "three-dimensional coupling mechanism" and clarifies its dynamic coupling path by analyzing the interactive relationship between each subsystem:

[0223] (1) Physical space-social network coupling: Architectural layout and spatial form directly affect the spatial accessibility and interaction frequency between individuals, thereby driving the evolution of the social network structure within the community. Therefore, the architectural form-spatial accessibility-social interaction-network structure evolution can be established.

[0224] (2) Social network-cultural communication coupling: Architectural layout and spatial form directly affect the spatial accessibility and interaction frequency between individuals, thereby driving the evolution of the social network structure within the community. Therefore, it is possible to establish network centrality-information flow efficiency-intangible cultural heritage memory transmission-cultural factor diffusion.

[0225] (3) Cultural Communication-Material Space Coupling: The practical activities of certain cultural crafts place specific demands on spatial form, which in turn promotes the adjustment of architectural functions and layout, leading to the reshaping of physical space. The path of action is: Therefore, a process of craft activity – spatial use – architectural transformation – spatial evolution can be constructed.

[0226] Step 4B: Multi-modal coupling model structure design.

[0227] That is, the interrelationships between the three models should be coupled. For example, physical space may be influenced by social networks, so we need to establish a relationship between physical space and social networks. Social networks may be influenced by cultural communication networks, so we need to establish a relationship between social networks and cultural communication networks. The same is true for the relationship between cultural communication and physical space.

[0228] Step 4B1. Modeling framework: ST-GNN multimodal spatiotemporal network

[0229] This paper selects the spatiotemporal graph neural network ST-GNN as the coupling backbone network for the following reasons:

[0230] Use graph structures to model irregular spatial units such as blocks and nodes, and support topological modeling to handle spatial heterogeneity.

[0231] The time series learning LSTM gating mechanism captures the historical evolution trend of multimodal indicators.

[0232] Finally, the graph attention mechanism (GAT) realizes the dynamic weighting and coupling modeling of features of different subsystems to enhance the feature fusion capability.

[0233] Compared with models such as system dynamics (SD) and cellular automata (CA), ST-GNN has advantages in spatial resolution, data fusion capability, and prediction accuracy.

[0234] Step 4B2: Setting up the coupled equation system.

[0235] Taking the material space M, social group structure S and cultural communication state C as dynamic state variables, the following multimodal coupled evolution equations are constructed, namely formula (12):

[0236]

[0237]

[0238]

[0239] in, It represents the reaction effect of the community network model and the cultural transmission factor diffusion model on the material space evolution network;

[0240] Indicates the diffusion coefficient of the material space form, which is the set value;

[0241] It represents the reconstruction effect of the material space evolution network model and the cultural factor diffusion model on the community network.

[0242] represents the social network adjacency matrix, which is calculated from formula (13);

[0243] It represents the supporting effect of the material space evolution network model and the community network model on the cultural factor diffusion model.

[0244] The Laplace operator representing the cultural factor is a set value.

[0245] Formula 12 outputs the future physical space layout, social network topology, and cultural communication scope.

[0246] The dynamic state variables M, S, and C here refer to the heritage material space evolution model obtained in step 1C, the dynamic interaction model of community multi-agents obtained in step 2C, and the cultural factor propagation and diffusion model obtained in step 3C, respectively.

[0247] The ST-GNN multimodal spatiotemporal network in step 4B1 is to fit this coupling equation, that is, to fit formula 12, and find the optimal solution, that is, 、 、 .

[0248] Step 4C: Design cross-modal interaction mechanism.

[0249] 1. Material Space—Social Network:

[0250] Use the spatial accessibility matrix to set the initial social network edge weights and initialize the social network adjacency matrix :

[0251] (13)

[0252] in, represents the geographical distance between nodes, is the attenuation coefficient.

[0253] 2. Social Network - Cultural Communication:

[0254] Adjusting the rate of cultural dissemination based on the network centrality of nodes , Used to adjust the cultural transmission equation function:

[0255] (14)

[0256] in, Central node centrality.

[0257] 3. Cultural Communication - Material Space:

[0258] Increase the probability of building updates based on the activity of cultural crafts , Used to adjust the material space evolution equation function

[0259] (15)

[0260] in, Indicates the level of craft inheritance. The adaptability of building functions.

[0261] Step 4D: Hierarchical modeling strategy for coupled structures

[0262] In order to clarify the functional positioning and coupling direction of each subsystem, this paper introduces a three-layer coupling modeling approach to clarify the functional positioning of each subsystem. The bottom-level material space provides a physical carrier for cultural activities and a spatial platform for residents to interact. The middle-level social network becomes a transmission channel for the flow of culture and behavior, and has dynamic evolution capabilities. The upper-level cultural genes dominate heritage recognition, space use intentions, and system evolution direction. Specifically, Figure 6 As shown in Figure 2, physical space, including geographic location and building types, constitutes the physical foundation and environmental constraints for system operation. Spatial indicators at the physical level act on the social level through "spatial constraints." This means that material attributes such as physical space layout, accessibility, and geographical barriers can restrict or influence the formation and development of social relationships, thereby shaping the structure of social networks. Network structures at the social level act as "transmission vehicles," transmitting information, ideas, practices, or "cultural memes" from one node to another. Social relationships and network structures are the channels and media for cultural transmission. The "gene distribution" at the cultural level generates "demand feedback," which in turn influences spatial indicators at the physical level. This means that specific cultural preferences, lifestyles, or values ​​(i.e., the distribution of cultural memes) generate demand for physical space, which can lead to adjustments in spatial form, the construction or renovation of infrastructure, and changes in land use patterns.

[0263] The input of the coupling model is the output features of the three models, namely, physical space features such as the expansion speed of the heritage community, the contour fractal dimension, the synthesized aging effect image and the texture evolution prediction value, community network features (such as neighborhood density index, the length of the intangible cultural heritage inheritance chain, the proportion of local employment, etc.), cultural factor transmission features (such as the end time, the number of cultural factor adopters, innovation adoption coefficient, etc.).

[0264] Figure 1 The output of the coupling model is the prediction of the material spatial structure of the human settlement heritage community, such as the spatial expansion range and rate diagram, and the community integration trend diagram; the prediction of the network structure of the heritage community, such as resident migration or interaction, network topology evolution, and the prediction of cultural factor transmission, such as the cultural theme diffusion heat map.

[0265] Scenario simulations input different scenario variables (such as tourism population growth, policy intervention, and infrastructure development) and output the evolutionary paths of heritage systems under different scenarios. For example, they predict the spatial functional reconstruction trend of the core area over the next ten years under the conditions of "high tourism pressure and low governance response."

[0266] Step 5: Prediction and scenario simulation of human settlement heritage systems.

[0267] To achieve intelligent prediction and control simulation of human-inhabited heritage systems, this paper designs an integrated "prediction-feedback-visualization" mechanism based on a multimodal coupling model, constructs three functional modules including a scenario generation engine, an ABM multi-agent behavior simulation system, and a three-dimensional visualization platform, which collaboratively support the evolution trend analysis and situational response capability assessment of complex heritage systems.

[0268] 1. Scenario Generation Engine

[0269] This module is used to construct system disturbance scenarios under multiple external input conditions, and to predict the regulatory effects of macro variables such as policy intervention and environmental changes on the system evolution path. The specific design is as follows:

[0270] 1. Multi-strategy parameterized configuration: supports user-defined input variables, including but not limited to the following scenarios:

[0271] (1) Adjustment of the intensity of heritage protection policies, such as restrictions on building renovation and the scale of investment in cultural heritage protection funds;

[0272] (2) Changes in resident structure, such as migration rate and age structure;

[0273] (3) Tourism development pressures, such as annual growth rate of tourists and changes in the frequency of cultural activities;

[0274] (4) Promotion of cultural policies, such as the popularization of intangible cultural heritage education.

[0275] 2. Perturbation path injection mechanism: By adjusting the key parameters in the coupling model such as ( ) to make dynamic adjustments, simulate the impact of external inputs on the system's internal evolution mechanism, and drive the system's multi-path evolution under different constraints.

[0276] (2) ABM multi-agent behavior simulation linkage module

[0277] This module is used to simulate the behavioral decision-making and interaction processes of various micro-individuals in the heritage system (residents, tourists, intangible cultural heritage inheritors, policy implementers, etc.), revealing the social feedback mechanism and system emergence characteristics.

[0278] 1. Agent type classification:

[0279] (1) Resident Agent: has state variables such as spatial residential location, cultural identity level, and social circle;

[0280] (2) Intangible Cultural Heritage Inheritor Agent: including inheritance level, influence, and activity;

[0281] (3) Policy implementer Agent: can conduct regulatory intervention, such as capital injection and functional adjustment

[0282] (4) Tourist Agent: Adjusts behavior trajectory based on the cultural attraction heat map.

[0283] 2. Coupling model empowerment mechanism: The state variables of each agent (such as migration probability, cultural adoption willingness, and dissemination ability) are dynamically output by the ST-GNN model; the social network structure affects the behavior diffusion path, and the cultural heat field affects behavioral preferences; agent behavior in turn acts on the model input, forming a closed-loop feedback system.

[0284] 3. Example of rules of conduct:

[0285] (1) Residents' willingness to move out: influenced by both living comfort and cultural participation;

[0286] (2) The inheritor agent selects the recipient of the art: the neighbor nodes with high cultural acceptance are given priority;

[0287] (3) Tourist Agent Path Selection: It is guided by both the frequency of intangible cultural heritage activities and the popularity of spatial aggregation.

[0288] (3) 3D visualization platform

[0289] To enable intuitive presentation and interactive analysis of system evolution results, this paper constructs a web-based 3D GIS visualization platform. This platform supports interactive operations such as timeline scrolling, perspective rotation, and switching between multidimensional data layers. The backend, connected to the coupled model, refreshes the simulation scene in real time using GeoJSON format. Key features include:

[0290] 1. System evolution path animation: Dynamic demonstration process showing changes in building density, changes in social network structure and the propagation trajectory of cultural factors.

[0291] 2. Heat map: Calculate the "cultural attraction field" based on the density of cultural factors. High-heat areas are highlighted in red, and low-heat areas are faded. Supports simultaneous comparison with policy intervention events to facilitate analysis of response effects.

[0292] 3. Scenario comparison chart: Supports the "dual-window simultaneous playback" function, presenting the similarities and differences in system evolution under two strategy settings. It can dynamically compare indicators such as resident migration rate, cultural activity, and the spatial distribution density of intangible cultural heritage inheritors over time.

[0293] The above description is merely a preferred embodiment of the present invention and does not limit the present invention in any way. Any person skilled in the art who, without departing from the scope of the present invention, makes any equivalent substitution, modification, or other changes to the technical solution and technical content disclosed in the present invention shall be deemed to be within the scope of the present invention and still fall within the scope of protection of the present invention.

Claims

1. An intelligent prediction method for the dynamic evolution of human settlement heritage systems, characterized by: The following steps are involved: Establish a model for the evolution of heritage material space, a dynamic interaction model of multiple community agents, and a model for the dissemination and diffusion of cultural factors; Then, the heritage material spatial morphological evolution model, the community multi-agent dynamic interaction model, and the cultural factor diffusion model are coupled to form a multimodal coupling model. Finally, based on the multimodal coupling model, future scenario patterns are predicted; The future scenario models include: the predicted scenario of the physical space structure of the human settlement heritage community, the predicted scenario of the network structure of the heritage community, and the predicted scenario of the spread of cultural factors; The steps of forming a multimodal coupling model include: Step 4A: Coupling mechanism design; The coupling mechanisms include: physical space-social network coupling, social network-cultural communication coupling, and cultural communication-physical space coupling; Step 4B: Multimodal coupling model structure design, specifically including: Step 4B1, modeling framework: ST-GNN multimodal spatiotemporal network; Step 4B2, setting the coupled equation system; The material space M, social group structure S and cultural communication state C are regarded as dynamic state variables respectively, and a multi-modal coupled evolution equation system is constructed; ； ； ； in, It represents the reaction effect of the community network model and the cultural transmission factor diffusion model on the material space evolution network; Indicates the diffusion coefficient of the material space form, which is the set value; It represents the reconstruction effect of the material space evolution network model and the cultural factor diffusion model on the community network. represents the social network adjacency matrix; It shows the supporting effect of the material space evolution network model and the community network model on the cultural factor diffusion model; The Laplace operator representing the cultural factor is the set value; The dynamic state variable M refers to the model of the material spatial evolution of heritage; The dynamic state variable S refers to the dynamic interaction model of multiple agents in the community; The dynamic state variable C refers to the cultural factor diffusion model; Step 4C: Design cross-modal interaction mechanisms, specifically including: Physical Space - Social Network: Use the spatial accessibility matrix to set the initial social network edge weights and initialize the social network adjacency matrix : ； in, represents the geographical distance between nodes, is the attenuation coefficient; Social Network - Cultural Communication: Adjusting the rate of cultural dissemination based on the network centrality of nodes , Used to adjust the cultural transmission equation function: ； in, Central node centrality; Cultural Communication—Material Space: Increase the probability of building updates based on the activity of cultural crafts , Used to adjust the material space evolution equation function: ； in, Indicates the level of craft inheritance. The adaptability of building functions.

2. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 1 is characterized by: The specific steps to establish a model of the evolution of heritage material spatial form include: Step 1A: Collect multimodal data on human settlement heritage; Step 1B: Analyze the material spatial form of residential heritage; Step 1C: Construct a model of the evolution of the heritage material spatial form.

3. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 2 is characterized by: Step 1B specifically includes the following steps: Step 1B1, macroscopic external space morphology analysis; Step 1B2: Analysis of microscopic internal space morphology.

4. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 3 is characterized by: Step 1C specifically includes the following steps: Step 1C1, establishing a macroscopic spatial morphological evolution prediction model; Step 1C2: Establish a microscopic spatial morphological evolution prediction model; Step 1C3: Couple the macroscopic spatial morphology evolution prediction model and the microscopic spatial morphology evolution prediction model to form a heritage material spatial morphology evolution model.

5. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 1 is characterized by: The steps to establish a dynamic interaction model of multiple community agents include: Step 2A, design of specific indicator system; Step 2B: Multi-source heterogeneous data collection and fusion; Step 2C: Dynamic relationship network construction and modeling.

6. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 5 is characterized by: Step 2A includes indicators: neighborhood density index NDI, intangible cultural heritage chain length TCL and local employment ratio LER.

7. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 1 is characterized by: The specific steps to establish a cultural factor diffusion model include: Step 3A, node semantic encoding; Step 3B: Build a multicultural communication network; Step 3C: Establish a cultural factor diffusion model based on the multicultural communication network.

8. The intelligent prediction method for the dynamic evolution of human settlement heritage systems according to claim 7 is characterized by: The cultural communication network in step 3B includes: master-apprentice inheritance network, festival participation network and language diffusion network.

Images