Industrial space suitability optimization method and system

Abstract

The invention relates to the technical field of urban planning, in particular to an industrial space suitability optimization method and system, and the method comprises the steps: collecting geographic information data, industrial data, industrial service related data, enterprise information and infrastructure data as a multi-source data set, associating the multi-source data set with geographic space coordinates, and inputting the geographic space coordinates into a database; the adjusted weight coefficient is applied to the LSTM-ARIMA hybrid model, and updated future industry prediction data is obtained; generating optimized industrial land layout scheme data based on a preset parametric modeling rule; determining planning scale data selected by a user, and obtaining a unified database associated with cross-level data; and comparing the land utilization change condition and the enterprise activity state data with data in the unified database, and when it is detected that data updating lag exists, updating the latest data after verification and prediction filling to the unified database. According to the invention, the adaptability and the intelligent level of urban industrial space planning can be improved.

Description

Technical Field

[0001] This application relates to the technical field of urban planning, and in particular, to an optimization method and system for industrial space adaptability. Background Art

[0002] Traditional urban industrial space planning relies on static parameters, manual experience, and offline models to delimit land use boundaries and development intensity indicators through preset rules. However, in the face of dynamic adjustment of released information, rapid industrial iteration, and emergencies, the existing methods lack the ability of real-time data feedback and dynamic adjustment, resulting in the rapid obsolescence of the planning scheme and difficulty in adapting to the complex and changeable real-world demands.

[0003] First, multi-objective optimization (such as balancing development intensity, functional matching balance, and cost control) relies on manual rules or offline algorithms and is difficult to quickly weigh conflicting objectives in dynamic scenarios; second, after an early warning is triggered (such as when the land development intensity exceeds the standard), parameter adjustment and scheme optimization require manual intervention and cannot form an autonomous closed-loop system of "perception - decision - execution"; finally, the functional area division is rigid and the transition area design is insufficient, resulting in fragmented land use, poor accessibility, exacerbating the separation of functional matching and resource waste.

[0004] As can be seen from the above, how to improve the adaptability and intelligence level of urban industrial space planning remains to be solved. Summary of the Invention

[0005] In order to improve the adaptability and intelligence level of urban industrial space planning, this application provides an optimization method and system for industrial space adaptability.

[0006] In a first aspect, this application provides an optimization method for industrial space adaptability, adopting the following technical solutions: An industrial space adaptability optimization method, which obtains corresponding geographic information data, industrial data, industrial service-related data, enterprise information, and infrastructure data; performs data cleaning and standardization processing on the geographic information data, industrial data, industrial service-related data, enterprise information, and infrastructure data, and collects them into corresponding multi-source datasets. The multi-source datasets are associated with geographic space coordinates and entered into a database. Among them, the geographic information data includes the geographical location, terrain, and surrounding environment of industrial parks, the industrial data includes industrial scale parameters, enterprise operation indicators, and industry classification data, the industrial service-related data includes industrial development-related information and resource planning information, the enterprise information includes industrial and commercial registration and migration records, and the infrastructure data includes transportation road networks and public facilities; calls a pre-trained LSTM-ARIMA hybrid model, inputs the standardized multi-source datasets into the LSTM-ARIMA hybrid model, and obtains corresponding future industrial prediction data. The future industrial prediction data includes the industrial growth probability distribution, industrial transformation probability distribution, and human resource demand prediction data for the next 3 to 5 years. The LSTM-ARIMA hybrid model is trained and converged by inputting historical industrial data, industrial service-related data, population migration data, and land transaction data. The historical industrial data includes macroeconomic indicators, output value, and labor market data. Among them, the LSTM-ARIMA hybrid model includes an LSTM layer and an ARIMA layer. The number of hidden layer neurons in the LSTM layer is ≥64 and is used to capture non-linear time series relationships. The ARIMA layer optimizes the parameters (p, d, q) through the AIC criterion and is used to correct the residual sequence of the LSTM layer; the loss function is , where λ is the weight coefficient, and its value range is [0, 1], RMSE is the root mean square error, and MAE is the mean absolute error; calls a pre-trained BERT model, inputs the standardized industrial service-related data into the BERT model for classification. The types of industrial service-related data include high-influence type, low-influence type, and no obvious influence type. Extracts the keyword data from the classified industrial service-related data through TF-IDF, and calculates the corresponding text influence coefficient based on the keyword data. The calculation formula is , where is the text influence coefficient, is the sum of all keywords in the industrial service-related data, is the influence polarity of the keyword, is the A value representing the importance of a keyword; adjusting the corresponding weight coefficient based on the keyword list and the corresponding TF-IDF value in the keyword data to obtain the corresponding adjusted weight coefficient, applying the adjusted weight coefficient to the LSTM-ARIMA hybrid model to obtain updated future industry prediction data; retrieving a pre-set parametric modeling rule, optimizing the geographic information data and the updated future industry prediction data based on the parametric modeling rule, and generating optimized industrial land layout plan data; determining the planning scale data selected by the user, where the planning scale data includes the municipal level corresponding to the macro level, the district level corresponding to the meso level, the park level corresponding to the micro level, and the plot level corresponding to the fine-grained level, screening the data sets at the corresponding level and adjusting the GIS resolution based on the planning scale data. Among them, the macro level integrates macroeconomic indicators, total population, and transportation framework, the meso level integrates industrial land distribution and road network density, the micro level integrates building height and open space ratio, and the fine-grained level integrates terrain slope and property boundary to obtain a unified database with cross-level data association, where the unified database includes data sets at different levels and their corresponding GIS resolutions; monitoring the land use change situation in a specified area based on satellite remote sensing image technology, where the land use change situation includes the construction and demolition of buildings and the change of green spaces, and obtaining the corresponding enterprise activity status data in real time. Comparing the land use change situation and the enterprise activity status data with the data in the unified database, and when it is detected that there is a data update lag, prompting the corresponding user verification operation; before or at the same time as the user verification, predicting and filling in the missing data based on the detected surrounding change trends, where the surrounding change trends include the development intensity of adjacent plots and the concentration of enterprise activities, and updating the latest data after verification and predictive filling into the unified database.

[0007] By adopting the above technical solution, through integrating multi-source heterogeneous data and combining the dynamic optimization technology of deep learning and reinforcement learning, an industrial space planning system with adaptive capabilities is constructed, forming an intelligent, flexible, and iterative urban industrial space layout solution.

[0008] Optionally, in the calculation process of the text influence coefficient, the method further includes: Constructing a corresponding influence dictionary, where the influence dictionary contains positive keywords and negative keywords; Performing word segmentation on the industrial service-related data to extract candidate keywords; calculating the TF-IDF weight of each keyword, and the formula is: , where N is the total number of published information documents, is the number of documents containing the keyword, and the published information is the content of the industrial service-related data; Calculating the text influence coefficient according to the positive and negative classifications of the influence dictionary: , where the text influence coefficient and the keyword weight are combined to generate a dynamic adjustment coefficient, and the formula is: , where is a preset basic weight coefficient, is the value of the keyword, is used to control the influence degree of the influence coefficient on the weight, and its value range is from 0.3 to 0.7.

[0009] By adopting the above technical solution, by constructing an influence dictionary and combining TF-IDF weights to analyze industry service-related data, the positive and negative influence tendencies of the released information are quantified into a dynamic adjustment coefficient, enabling the system to real-time perceive the "influence intensity" and "keyword importance" of industry service-related data, thereby dynamically adjusting the weight parameters of the industry prediction model.

[0010] Optionally, in the process of generating the optimized industrial land layout plan data, the method further includes: Retrieving geographic information data, which also includes road networks, digital elevation models, and industrial space contour vector data, performing five-meter by five-meter grid resampling on the digital elevation model, and filling in the missing segments of the road network; Dividing the industrial plots in the future industry prediction data based on the Delaunay triangulation algorithm, screening and eliminating plots with a slope greater than 5% or an area less than two hectares; if there is a spatial overlap between the planned land and the ecological protection area, triggering the ant colony algorithm to relocate; Setting corresponding functional transition periods, including industrial-residential transition areas and industrial-public service transition areas; and performing corresponding street network density grading and corresponding line-of-sight corridor analysis.

[0011] By adopting the above technical solution, through the combination of multi-source geographic data processing and intelligent algorithms, the automatic screening and dynamic optimization of industrial land layout are realized. At the same time, through the design of functional transition areas, street density grading, and line-of-sight corridor analysis, the balance of functional mixing, traffic efficiency, and spatial aesthetics is achieved, and finally an intelligent planning scheme with ecological compatibility, functional adaptability, and spatial accessibility is formed, significantly improving the scientificity and dynamic adaptability of the urban industrial land layout.

[0012] Optionally, in the process of performing the corresponding line-of-sight corridor analysis, the method further includes: Determining the geographical location of landmark buildings in the geographic information system ; Based on the geographical location calculating the corresponding visible area, where the calculation formula is , that is, the set of all observation points not blocked by obstacles.

[0013] Obtain the actual height corresponding to the landmark building, and determine the corresponding restricted building height based on the actual height. The calculation formula is , and obtain the corresponding visible area vector map and building height limit rule file.

[0014] By adopting the above technical solution, by accurately positioning the spatial position and height of the landmark building, combining with the three-dimensional space simulation technology to dynamically delimit the visible area and generate the building height limit rules, it ensures the visual dominance of the landmark building and the height coordination of the surrounding buildings, protects both the openness of the urban landscape and the landmark recognition, and realizes the balance of spatial aesthetics and functional layout through data-driven height limit constraints, provides a scientific basis for the urban skyline planning, and improves the visual coherence and ecological compatibility of the spatial design.

[0015] Optionally, the method further includes: monitoring whether the land development intensity exceeds 89% based on the land use change situation or detecting a sudden change in industrial demand caused by the change of the published information, and obtaining the corresponding monitoring result; Judge whether the monitoring result triggers an alarm. The alarm is divided into a red alarm and a yellow alarm; if it is a red alarm, perform the operation of re-running the parametric model to generate a new layout plan. If it is a yellow alarm, optimize the functional mixing ratio or increase the green area, generate the corresponding alarm report, and feedback the alarm report.

[0016] By adopting the above technical solution, by real-time monitoring the industrial demand fluctuations caused by the land development intensity and the change of the published information, a hierarchical early warning system is constructed to realize the dynamic response and flexible adjustment of the planning scheme: the red alarm triggers a global model recalculation to completely reconstruct the layout, and the yellow alarm quickly responds through local optimization, forming a closed loop of "monitoring - early warning - adaptive adjustment", significantly improving the agile response ability of the planning to sudden changes.

[0017] Optionally, during the optimization process of the yellow alarm, the method further includes: When the land development intensity exceeds the preset development threshold or the regional function coordination degree index exceeds the preset scheduling threshold, where the calculation formula of the regional function coordination degree index is ; optimize the proportion of commerce, green space, and public services through linear programming, and the corresponding optimization target calculation formula is }}, obtain the corresponding optimization result, and determine the corresponding new layout plan and alarm report based on the optimization result.

[0018] By adopting the above technical solutions, the land use function ratio is dynamically adjusted through linear programming technology. When the land development intensity or functional matching imbalance approaches the threshold, aiming at minimizing the development intensity and the regional function coordination degree index, a local optimization scheme is quickly generated through mathematical modeling. Without completely reconstructing the global layout, the resource tension or spatial imbalance problem is accurately alleviated, thereby enhancing the flexible regulation ability of the planning and the resource allocation efficiency, and ensuring the sustainability of urban development.

[0019] Optionally, the parametric modeling rule includes a dynamic space optimization module based on reinforcement learning, and the method further includes: Retrieve the predefined state space , where the state space includes the current land use layout, future industrial prediction data, environmental compatibility parameters, development intensity, and functional matching ratio; retrieve the predefined action space , and the action space can change the function type of the plot, adjust the plot area, and modify the building height limit; Retrieve the corresponding reward function , where , , is the weight coefficient, the ecological reserve overlap penalty term is negative and proportional to the overlap area, the development intensity target is 89%, the functional matching balance threshold is 0.8 - 1.2, the deep Q-network is used as the reinforcement learning algorithm, the number of neurons in the hidden layer is ≥128, the Adam optimizer is used, and the learning rate is 0.001; Perform the corresponding multi-objective optimization, and the optimization function: , obtain the dynamic adjustment of the reward function weight according to the real-time enterprise migration rate data and the real-time release information change frequency data, increases as the development intensity approaches the threshold; Generate the Pareto front through the NSGA-II algorithm, select the comprehensive optimal solution as the final layout plan, determine it as the industrial land layout plan data and output the decision report. The industrial land layout plan data includes the vector data of the multi-objective optimization results, and marks the development intensity, functional matching ratio, and ecological impact of each region. The decision report records the weight adjustment path, Pareto front graph, and Monte Carlo simulation prediction data during the reinforcement learning process.

[0020] By adopting the above technical solutions, the land function ratio is dynamically adjusted through linear programming technology. When the land development intensity or functional matching imbalance approaches the threshold, the local optimization plan is quickly generated through mathematical modeling with the goal of minimizing the development intensity and regional functional coordination indicators. Without completely reconstructing the global layout, the resource shortage or spatial imbalance problem can be accurately alleviated, thereby improving the planning's flexible regulation ability and resource allocation efficiency, and ensuring the sustainability of urban development.

[0021] In the second aspect, the present application provides an industrial space adaptability optimization system, which adopts the following technical solutions: An industrial space adaptability optimization system, comprising: A multi-source data set collection module obtains corresponding geographic information data, industrial data, industrial service-related data, enterprise information and infrastructure data; performs data cleaning and standardization on geographic information data, industrial data, industrial service-related data, enterprise information and infrastructure data to collect corresponding multi-source data sets, associate the multi-source data sets with geographic spatial coordinates and enter them into a database, wherein the geographic information data includes the geographical location, topography and surrounding environment of the industrial park, the industrial data includes industrial scale parameters, enterprise operating indicators and industry classification data, the industrial service-related data includes information related to industrial development and resource planning information, the enterprise information includes industrial and commercial registration and migration records, and the infrastructure data includes transportation network and public facilities; The future industry forecast data update acquisition module calls the pre-trained LSTM-ARIMA hybrid model, inputs the standardized multi-source data set into the LSTM-ARIMA hybrid model, and obtains the corresponding future industry forecast data. The future industry forecast data includes the probability distribution of industry growth, the probability distribution of industry transformation, and the forecast data of human resource demand in the next 3 to 5 years. The LSTM-ARIMA hybrid model is obtained by inputting historical industry data, industry service-related data, population migration data, and land transaction data for training convergence. The historical industry data includes macroeconomic indicators, output value, and labor market data. Among them, the LSTM-ARIMA hybrid model includes an LSTM layer and an ARIMA layer. The number of hidden layer neurons of the LSTM layer is ≥64 and is used to capture nonlinear time series relationships. The ARIMA layer optimizes the parameters (p, d, q) through the AIC criterion and is used to correct the residual sequence of the LSTM layer; the loss function is , where λ is the weight coefficient with a value range of [0, 1], RMSE is the root mean square error, and MAE is the mean absolute error; retrieve the pre-trained BERT model, input the standardized industry service-related data into the BERT model for classification. The types of industry service-related data include high-impact type, low-impact type, and no obvious impact type. Extract the keyword data from the classified industry service-related data through TF-IDF, and calculate the corresponding text influence coefficient based on the keyword data. The calculation formula is , where is the text influence coefficient, is the sum of all keywords in the industry service-related data, is the influence polarity of the keyword, is the value, indicating the importance of the keyword; adjust the corresponding weight coefficient based on the keyword list and the corresponding TF-IDF value in the keyword data to obtain the corresponding adjusted weight coefficient, and apply the adjusted weight coefficient to the LSTM-ARIMA hybrid model to obtain updated future industry prediction data; Industry land layout plan data generation module, retrieve the pre-set parametric modeling rules, optimize the geographic information data and the updated future industry prediction data based on the parametric modeling rules, and use them to generate optimized industry land layout plan data; Unified database acquisition module, determine the selected planning scale data by the user. The planning scale data includes the municipal level corresponding to the macro level, the district level corresponding to the meso level, the park level corresponding to the micro level, and the plot level corresponding to the fine-grained level. Filter the data sets at the corresponding levels and adjust the GIS resolution based on the planning scale data. Among them, the macro level integrates macroeconomic indicators, total population, and transportation framework, the meso level integrates industrial land distribution and road network density, the micro level integrates building height and open space ratio, and the fine-grained level integrates terrain slope and property boundary, for obtaining a unified database with cross-level data association. Among them, the unified database includes data sets at different levels and their corresponding GIS resolutions; Data lag comparison module, monitor the land use change situation in the designated area based on satellite remote sensing image technology. The land use change situation includes the construction and demolition of buildings, and the change of green space, and obtain the corresponding enterprise activity status data in real time, for comparing the land use change situation and the enterprise activity status data with the data in the unified database. When it is detected that there is a data update lag, prompt the corresponding user verification operation; before or at the same time as the user verification, predictively fill in the missing data based on the detected surrounding change trends. The surrounding change trends include the development intensity of adjacent plots and the concentration of enterprise activities, and update the latest data after verification and predictive filling to the unified database.

[0022] In a third aspect, the present application provides an optimization method for industrial space adaptability, adopting the following technical solution: An optimization method for industrial space adaptability, including a processor, in which a program of the optimization method for industrial space adaptability described in any one of the above is run.

[0023] In a fourth aspect, the present application provides a storage medium, adopting the following technical solution: A storage medium stores a program of the optimization method for industrial space adaptability described in any one of the above.

[0024] In summary, the present application includes at least one of the following beneficial technical effects: By integrating multi-source data and intelligent analysis technologies, such as LSTM-ARIMA and BERT models, accurately predict industrial trends and the impact of release information, enhancing the forward-looking and flexibility of planning. Adopt reinforcement learning and multi-objective optimization technologies, and combine with a hierarchical early warning mechanism to achieve dynamic monitoring and adaptive adjustment of key indicators such as land development intensity and functional matching balance, ensuring the balance between ecological constraints and development efficiency. Finally, using cross-level data association, real-time remote sensing monitoring and automated generation technologies, an intelligent industrial space layout from macro to fine-grained is constructed, providing scientific and iterative decision-making support, and promoting the sustainable development of the city and the ability to respond to future challenges. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a flowchart of an optimization method for industrial space adaptability shown according to an exemplary embodiment.

[0026] Figure 2 is a structural block diagram of an optimization system for industrial space adaptability shown according to an exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following details the embodiments of the present application, and the examples of the embodiments are shown in the drawings.

[0028] In the description of this specification, the description referring to the terms "certain embodiments", "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0029] The present application embodiment discloses an industrial space adaptability optimization method, referring to Figure 1 ,include: S100, obtaining corresponding geographic information data, industrial data, industrial service-related data, enterprise information and infrastructure data; performing data cleaning and standardization on the geographic information data, industrial data, industrial service-related data, enterprise information and infrastructure data, collecting them into corresponding multi-source data sets, associating the multi-source data sets with geographic spatial coordinates and entering them into a database, wherein the geographic information data includes the geographical location, topography and surrounding environment of the industrial park, the industrial data includes industrial scale parameters, enterprise operating indicators and industry classification data, the industrial service-related data includes information related to industrial development and resource planning information, the enterprise information includes industrial and commercial registration and migration records, and the infrastructure data includes transportation network and public facilities; Among them, for geographic information data, the geographical location of the industrial park is obtained through the GIS system to obtain the boundary coordinates of the industrial park; the terrain data uses digital elevation models or satellite topographic maps; the surrounding environment is integrated land use classification data, and the data format is spatial vector data or raster data.

[0030] As for industrial data, industrial data include industrial scale parameters, enterprise operating indicators and industry classification data. It should be pointed out here that industrial scale parameters can be obtained through regional industrial summary reports and industry analysis reports released by official statistical agencies. Enterprise operating indicators refer to the annual operating reports and financial release information disclosed by enterprises. Industry classification data is collected based on the standardized classification system released by industry associations and the segmented field research reports released by market research institutions. The above methods for obtaining industrial data are all obtained by technical personnel in this field on the corresponding official website through the Internet.

[0031] Obtain the corresponding industry service-related data through the official website. The industry service-related data includes information related to industrial development and land development restrictions. For example, you can obtain relevant release information about high-impact industries through the official website, and regularly determine whether the official website has updated relevant release information notifications to ensure that you can obtain the latest industry development-related information.

[0032] Enterprise information is obtained through the industrial and commercial registration system and the enterprise credit platform. Enterprise information includes registration time, industry classification, and migration records. Infrastructure data includes transportation network and public facilities. The transportation network includes road grades and intersection locations; public facilities include the coordinates and service radius of schools, hospitals, and parks.

[0033] It should be pointed out that the above-mentioned data need to be cleaned and standardized accordingly. Data cleaning and data standardization are common data preprocessing in this field, which will not be elaborated here.

[0034] For spatial association and geocoding, spatial association needs to be carried out first. The enterprise registration address is converted into coordinate points through GIS tools. It is also necessary to associate the applicable scope of the published information with geographical boundary data to ensure a one-to-one correspondence between the published information and the spatial area. Then, the integration of multi-source data is carried out. By associating industrial data with geographical regions, a "region-output value" spatial layer is generated. In addition, it is necessary to associate enterprise migration records with traffic road network data to analyze the relationship between migration paths and traffic accessibility.

[0035] Finally, the multi-source data sets are stored in the database and managed. It should be noted here that a spatial index needs to be created to accelerate regional queries. Through the above steps, the systematic integration of multi-source heterogeneous data is achieved, providing a high-quality and structured data foundation for subsequent prediction modeling and spatial optimization, and ensuring that the input data of the planning model has spatio-temporal consistency and business relevance.

[0036] S200, retrieve the pre-trained LSTM-ARIMA hybrid model, input the standardized multi-source data sets into the LSTM-ARIMA hybrid model to obtain the corresponding future industry prediction data. The future industry prediction data includes the industry growth probability distribution, industry transformation probability distribution, and human resource demand prediction data for the next 3 to 5 years. The LSTM-ARIMA hybrid model is obtained through training and convergence by inputting historical industry data, industry service-related data, population migration data, and land transaction data. The historical industry data includes macroeconomic indicators, output value, and labor market data. Among them, the LSTM-ARIMA hybrid model includes an LSTM layer and an ARIMA layer. The number of hidden layer neurons in the LSTM layer is ≥64 and is used to capture non-linear time series relationships. The ARIMA layer optimizes the parameters (p, d, q) through the AIC criterion and is used to correct the residual sequence of the LSTM layer; the loss function is , where λ is the weight coefficient, and its value range is [0, 1], RMSE is the root mean square error, and MAE is the mean absolute error; retrieve the pre-trained BERT model, input the standardized industry service-related data into the BERT model for classification. The types of industry service-related data include high-impact type, low-impact type, and no obvious impact type. Extract the keyword data from the classified industry service-related data through TF-IDF, and calculate the corresponding text influence coefficient based on the keyword data. The calculation formula is , where, is the text influence coefficient, is the sum of all keywords in the industry service-related data, is the influence polarity of the keyword, is the Value, representing the importance of keywords; based on the keyword list and corresponding TF-IDF values in the keyword data, adjust the corresponding weight coefficients to obtain the corresponding adjusted weight coefficients, and apply the adjusted weight coefficients to the LSTM-ARIMA hybrid model to obtain updated future industry prediction data.

[0037] Among them, for the LSTM-ARIMA hybrid model, the number of neurons in the LSTM layer is at least 64 neurons in the hidden layer to capture complex non-linear time series relationships, and the input data is multi-source data after standardization. The ARIMA layer selects the optimal (p, d, q) parameter combination (such as p = 2, d = 1, q = 1) through the AIC criterion to correct the residual sequence of the LSTM and improve the linear trend prediction accuracy; the input data is the residual sequence of the LSTM layer to ensure that the ARIMA focuses on the compensation of the linear trend.

[0038] That is to say, the LSTM layer captures the impact of non-linear factors such as sudden changes in released information and population migration on the industry, and the ARIMA corrects the long-term stable linear trend. The combination of the two significantly improves the accuracy and robustness of the prediction; by integrating multi-dimensional data such as industrial service-related data coding and population migration, it is ensured that the prediction results reflect the real market dynamics and the impact of released information.

[0039] For joint training and loss function optimization, the calculation formula of the loss function is as above. The role can balance the robustness and precision of the prediction error; for example, when = 0.6, the model pays more attention to reducing large errors.

[0040] For the training process, input historical industry data, industrial service-related data coding, population migration trends, and land transaction data, and then jointly optimize the LSTM and ARIMA parameters through backpropagation until the loss function converges. Through joint training, the parameters of the LSTM and ARIMA are jointly optimized, reducing the model's dependence on a single data pattern and improving the performance in complex scenarios. In addition, input the standardized industrial service-related data into the BERT model. The BERT captures the context relationship through the self-attention mechanism and outputs the classification probability, and the classification result assigns labels to each piece of released information.

[0041] At the same time, after segmenting the industrial service-related data, calculate the TF-IDF value of the keywords, and assign influence polarities to each keyword according to the pre-constructed influence dictionary. , calculate the text influence coefficient, and the calculation formula is shown above. The text influence coefficient converts the "influence tendency" of the industrial service-related data into a numerical value, providing a computable parameter for model weight adjustment; the TF-IDF weight highlights the core regulatory objects of the industrial service-related data.

[0042] For the embodiments of this application, the dynamic adjustment of weights and model update include the following steps: For the adjustment of basic weights, combine the text influence coefficient and the keyword weights to generate a dynamic adjustment coefficient: , where is an adjustment coefficient that controls the influence intensity of the influence coefficient on the weights; if the published information S = +2, then increases, enhancing the prediction weight of LSTM for industrial expansion, and it is necessary to ensure that ∈[0,1] to avoid weight imbalance.

[0043] By directly converting the semantic information of industrial service-related data into model parameter adjustment, the prediction result is closer to the content of the published information; the adjusted is re-input into the LSTM-ARIMA model to dynamically adjust the sensitivity of the LSTM layer to the features related to the published information.

[0044] For generating the updated future industrial prediction data, it includes: Industrial growth probability distribution: the expansion or contraction probability of each industry in the next 3-5 years; Industrial transformation probability distribution: the possibility of an industry becoming the regional leading industry; Human resource demand prediction data: predicting the difference between future employment demand and supply in each industry.

[0045] Through the above steps, the system realizes the semantic understanding of industrial service-related data, the dynamic fusion of multi-source data, and the adaptive optimization of the prediction model, and finally provides a scientific basis for industrial space planning with accuracy, real-time performance, and the ability to respond to published information. Through real-time analysis of published information and update of model parameters, a closed loop of "published information perception → weight adjustment → prediction optimization" is formed to ensure that the planning suggestions are always dynamically consistent with the latest published information; the output data such as industrial growth probability and labor gap directly provide a scientific basis for land development intensity control and functional area planning, reducing human judgment deviation.

[0046] S300, retrieve the pre-set parametric modeling rules, optimize the geographic information data and the updated future industrial prediction data based on the parametric modeling rules, and generate the optimized industrial land layout plan data.

[0047] Among them, retrieve the parametric rules matching the current planning goal from the pre-set rule library, for example: Development intensity: regional development intensity ≤ 89%; Ecological protection rule: the overlapping area with the ecological protection area ≤ 5%; Functional matching ratio threshold: 0.8 - 1.2; Road network density standard: the spacing between main roads ≤ 500 meters, the spacing between secondary roads ≤ 800 meters.

[0048] According to the data characteristics, an optimization algorithm is selected, and the parameters need to be initialized: set the number of neurons in the hidden layer and the learning rate; define the population size, crossover probability, and mutation probability.

[0049] Then input the geographical information data and the updated future industry prediction data; and the corresponding data formats are spatial data and prediction data. Among them, spatial data: GeoJSON format plot boundaries, raster terrain data; prediction data: industrial output value growth rate and labor gap value in the next 3-5 years. Associate the future industry prediction data with the geographical spatial unit; and perform attribute binding.

[0050] Then generate and output the optimization results, and output the optimized industrial land layout plan data, including: Vector data: plot boundary coordinates, function types, building height limits; Attribute data: development intensity (85%), function matching ratio (0.9), ecological overlap area (0.1%); Annotation information: key indicators. And convert the optimization results into a GIS-compatible format and generate metadata descriptions.

[0051] The system strictly follows the parameterization rules, inputs the geographical information data and the future industry prediction data into the optimization algorithm, and finally generates the industrial land layout plan data to provide a structured output for subsequent planning approval.

[0052] S400, determine the planning scale data selected by the user. The planning scale data includes the municipal level corresponding to the macro level, the district level corresponding to the meso level, the park level corresponding to the micro level, and the plot level corresponding to the fine-grained level. Based on the planning scale data, filter the corresponding-level data sets and adjust the GIS resolution. Among them, at the macro level, integrate macroeconomic indicators, total population, and transportation framework; at the meso level, integrate industrial land distribution and road network density; at the micro level, integrate building height and open space ratio; at the fine-grained level, integrate terrain slope and property boundary to obtain a unified database with cross-level data association. Among them, the unified database includes data sets at different levels and their corresponding GIS resolutions.

[0053] Among them, for the municipal level corresponding to the macro level, it is necessary to integrate macroeconomic indicators, total population, and transportation framework; the corresponding GIS resolution is that the vector data uses a scale of 1:50,000, and the raster data resolution is ≥100 m × 100 m; thus, it can provide a macro framework for the regional development strategy.

[0054] For the district level corresponding to the meso level, it is necessary to integrate industrial land distribution and road network density; the corresponding GIS resolution is that the vector data uses a scale of 1:10,000, and the raster data resolution is ≤20 m × 20 m; thus, it can support the sub-region planning.

[0055] For the park level corresponding to the micro level, it is necessary to integrate building height and the proportion of open space; the corresponding GIS resolution is that the vector data uses a scale of 1:500, and the raster data resolution ≤ 1 m × 1 m. The corresponding GIS resolution is used to guide the detailed design of the park.

[0056] For the plot level corresponding to the fine-grained level, it is necessary to integrate terrain slope and property rights boundary; the corresponding GIS resolution is that the vector data uses a scale of 1:200, and the raster data resolution ≤ 0.5 m × 0.5 m; thus, the implementation feasibility can be ensured.

[0057] It should be noted here that it is necessary to define the corresponding data association rules, that is, cross-level association fields: macro → medium is associated through administrative division codes; medium → micro is associated through the district code where the park is located; micro → fine-grained is associated through plot ID. By establishing a logical link between data levels, the coherence of multi-scale analysis is ensured.

[0058] It should be noted here that the execution process of GIS resolution adaptation and data resampling includes: for the macro level, it is necessary to simplify high-precision vector data to low-precision through topological aggregation; for the fine-grained level, it is necessary to subdivide low-precision vector data.

[0059] At the same time, raster data processing is also required, which can be achieved through bilinear interpolation; through super-resolution algorithms; data accuracy and computational efficiency can be balanced.

[0060] For building a unified database, it is necessary to design the hierarchical table structure, including macro table, medium table, micro table, and fine-grained table. Specifically, the macro table: fields include administrative division code, macroeconomic indicators, total population, and traffic skeleton coordinates; the medium table: fields include district code, proportion of industrial land, and road network density; the micro table: fields include park ID, building height, and proportion of open space; the fine-grained table: fields include plot ID, slope value, and property rights boundary coordinates.

[0061] At the same time, it is necessary to optimize the spatial index, create a spatial R-tree index for each level of data, and pre-compute common aggregation indicators, so as to improve the cross-level query efficiency. Then, data is imported and associated. Vector data and raster data can be batch imported through the use of ETL tools; resolution metadata is marked for each level of data. Different levels of data are associated through foreign keys, and a spatial relationship table is established; thus, seamless switching and joint analysis of multi-scale data can be realized.

[0062] The execution steps for cross-level data verification include: first, logical consistency verification, which requires checking whether the "municipal macroeconomic indicators" are equal to the sum of all district-level macroeconomic indicators, and verifying whether the "building height of a certain park" complies with the height control rules of the district to which it belongs; then, spatial integrity verification is performed, ensuring that all plot boundaries completely cover the park scope without overlap or gaps, and checking whether the traffic skeleton data is consistent with the spatial distribution of the road network density data, so as to ensure the logical and spatial consistency between data levels and avoid planning conflicts.

[0063] By standardizing and integrating multi-scale data and building a hierarchical database, the system achieves full-chain data support from macro-strategy to micro-implementation: resolution adaptation ensures that macro data supports strategic decisions with low precision, while fine-grained data ensures design feasibility with high precision; the hierarchical database improves the efficiency of macro analysis through hierarchical indexing, while isolating high-precision data to avoid computational redundancy; cross-level associations achieve top-down precise transmission of published information and bottom-up conflict feedback, ultimately forming a data-driven dynamic planning closed loop that takes into account both planning efficiency and implementation compliance.

[0064] S500 monitors the changes in land use in designated areas based on satellite remote sensing imaging technology. Land use changes include the construction and demolition of buildings and changes in green areas. It obtains corresponding enterprise activity status data in real time and compares the land use change and enterprise activity status data with the data in the unified database. When a data update lag is detected, it prompts the user to perform the corresponding verification operation. Before or at the same time as user verification, it predictively fills in the missing data based on the detected surrounding change trends. The surrounding change trends include the development intensity of adjacent plots and the concentration of enterprise activities. The latest data after verification and predictive filling is updated to the unified database.

[0065] Among them, multispectral / high-resolution images of designated areas are obtained from satellite service providers at a fixed frequency. High-frequency image acquisition can ensure timely capture of land use changes, and high-resolution images support plot-level change detection, and multispectral images identify green space changes.

[0066] The U-Net model can be used to segment the building boundaries of current and historical images and calculate boundary differences. For green space changes, time series NDVI curve analysis can be used to identify areas of significant decline or increase. This can reduce the cost of manual visual interpretation and support rapid scanning of large areas. In addition, deep learning models can distinguish between temporary covers and permanent buildings.

[0067] For the collection of enterprise activity status data, when choosing to integrate multi-source data, it is necessary to obtain enterprise activity data. Generally, enterprise activity data includes but is not limited to electricity consumption, logistics freight volume, and employee commuting heat; it is also necessary to obtain the number of vehicles in the parking lot and the lighting intensity in the factory area. The remote sensing detection results are cross-validated through data such as electricity consumption and logistics. It should be noted here that the data obtained in the embodiments of this application all need to be preprocessed, such as filtering outlier values and aligning timestamps.

[0068] Then, data comparison and lag detection are carried out. This process requires prior spatio-temporal matching and difference analysis, that is, spatially overlaying the change results detected by remote sensing with the land use status in the unified database, and comparing the attributes of the enterprise activity data with the enterprise registration information in the database.

[0069] When determining whether there is a lag, spatial and attribute determinations are required. For spatial inconsistencies, that is, a change area is detected but the database has not been updated; for attribute inconsistencies, that is, the enterprise activity data conflicts with the registration status. Thus, the disconnection between the plan and the reality can be discovered in a timely manner, avoiding decision-making mistakes caused by plan lag, so as to ensure that the database always reflects the real land use status.

[0070] When triggering user verification and notification, the triggering conditions are that the spatial / attribute difference exceeds the threshold or there is continuous abnormality; and the corresponding notification content is the spatial coordinates of the difference area, a screenshot of the remote sensing image, and a comparison chart of enterprise activity data; by excluding algorithm misjudgments through user verification, it can be ensured that illegal acts are reported and processed in a timely manner.

[0071] In the embodiments of this application, predictive filling and data pre-update also need to be carried out. The corresponding execution steps include: Peripheral trend analysis: It is necessary to detect the change in building density within 500 meters around the target plot in the recent 3 years; Enterprise activity concentration: It is necessary to count the number of newly added surrounding enterprises and the industry distribution.

[0072] Then use the time series ARIMA to predict the development intensity in the next 3 months, and predict the enterprise activity level of the target plot through the spatial regression model. Before user verification, based on the trend, missing fields are pre-filled, which can provide temporary data for areas that have not completed verification and maintain the real-time nature of the planning model.

[0073] Mark "prediction source" and confidence level in the predicted data; trigger additional verification prompts for high-risk areas. In this way, users can clearly know which data are predicted values and avoid misjudgment; at the same time, risk priority sorting is carried out, and resources can be concentrated to verify high-confidence abnormal areas.

[0074] After the user confirms "compliance change", update the fields such as land use type and enterprise status in the unified database; mark "illegal construction" as "pending processing status" and trigger the planning enforcement process. Here, the predicted data also needs to be marked as "temporary version", which is only used for planning simulation and not as the basis for formal approval; after verification, it is replaced with the official data and the temporary mark is deleted. By retaining historical versions and supporting difference analysis and auditing, it can ensure that the database always reflects the latest land use status and supports the real-time nature of the planning model.

[0075] In the embodiment of the present application, during the calculation process of the text influence coefficient, the method further includes: S110, construct a corresponding influence dictionary.

[0076] Among them, the influence dictionary includes positive keywords and negative keywords, and positive keywords are given an influence polarity value ; negative keywords are given an influence polarity value ; filtering of words with no obvious influence tendency: exclude words with no influence tendency; thus, it can provide standardized influence labels for keywords in industry service-related data and ensure the repeatability of subsequent calculations; optimize the dictionary according to the characteristics of industry service-related data to improve the accuracy of influence judgment.

[0077] S120, perform word segmentation on industry service-related data to extract candidate keywords; calculate the TF-IDF weight of each keyword, and the formula is: , where N is the total number of published information documents, is the number of documents containing the keyword, and the published information is the content of industry service-related data.

[0078] Among them, during the process of word segmentation, a Chinese word segmentation tool (such as jieba) can be used to perform word segmentation on industry service-related data; the corresponding filtering rule is to remove stop words and retain nouns, verbs, and industry terms. Thus, candidate keywords that may carry influence can be extracted from industry service-related data; by retaining industry terms, a professional feature space can be constructed to improve the representation ability of text analysis for industrial technical features.

[0079] TF (term frequency) is to calculate the frequency of the keyword appearing in a single published information document, and IDF (inverse document frequency): calculate the rarity of the keyword. A high TF-IDF value indicates that the keyword is key and domain-specific in the current published information. The weight of general vocabulary can be reduced through IDF to highlight the core regulatory object of the published information.

[0080] S130, according to the positive and negative classification of the influence dictionary, calculate the text influence coefficient: , where the text influence coefficient is combined with the keyword weight to generate a dynamic adjustment coefficient, and the formula is: , where is a preset basic weight coefficient, is the value of the keyword, is used to control the influence degree of the influence coefficient on the weight, and its common value range is from 0.3 to 0.7.

[0081] Among them, the positive and negative tendencies of the industrial service-related data are converted into numerical values, and the high TF-IDF keywords have a greater impact on the score, avoiding the interference of secondary words. In addition, by adjusting parameters, the influence degree of the influence coefficient on the weight is controlled to ensure the stability and flexibility of the model.

[0082] In the embodiment of the present application, in the process of generating the optimized industrial land layout plan data, the method further includes: S310, retrieving geographic information data, which also includes road networks, digital elevation models, and industrial space contour vector data, resampling the digital elevation model into a five-meter by five-meter grid, and filling in the missing sections of the road network.

[0083] Among them, the system extracts multi-source geographic information data from the geographic information system, including road network data, digital elevation models, and existing industrial land boundaries. The geographic information data forms the basis for subsequent analysis; in order to improve the accuracy and calculation efficiency of terrain analysis, the digital elevation model is resampled into a regular grid of five meters by five meters. During the resampling process, an interpolation algorithm is used to ensure the smooth transition and accuracy of terrain features; at the same time, for the missing parts in the road network data, spatial interpolation techniques are used for supplementation. For example, reasonable road network connection segments can be generated through the nearest neighbor method, buffer analysis, or by combining the topological relationships of surrounding roads to ensure the integrity of the road network.

[0084] Through the resampling process of the digital elevation model, more detailed surface information can be provided, which helps to accurately evaluate the impact of terrain undulations on the future industrial land layout, especially in slope analysis and drainage planning; the industrial space contour vector data is data that presents the spatial distribution range of industrial agglomeration areas in a digital graph, generated through spatial feature analysis and not involving specific enterprise ownership information; and complete and accurate road network data is the basis for ensuring the smoothness of logistics transportation, personnel flow, and infrastructure construction, providing reliable support for subsequent plot division and functional zoning; in addition, complete geographic information data can also provide real-time support for dynamic adjustment and optimization of the plan.

[0085] S320, divide the industrial plots in the future industrial prediction data based on the Delaunay triangulation algorithm, and screen and eliminate the plots with a slope greater than 5% or an area less than 2 hectares; if there is a spatial overlap between the planned land and the ecological protection area, trigger the ant colony algorithm to reselect the location.

[0086] Among them, after obtaining the geographical information data, the system uses the Delaunay triangulation algorithm to spatially divide the predicted industrial plots. It should be noted here that the Delaunay triangulation is a method for constructing an irregular triangular network based on a point set, which can maximize the minimum angle of each triangle, thus avoiding the appearance of long and narrow-shaped plots and ensuring the uniformity and rationality of the division results.

[0087] Then, according to the terrain analysis results, screen out the plots with a slope greater than 5% and eliminate them, because these areas may not be suitable for large-scale development, which is likely to cause soil erosion or too high construction costs; at the same time, the plots with an area less than 2 hectares will also be excluded, because small plots are difficult to meet the scale requirements of modern industrial development; if it is found that there is a spatial overlap between the planned industrial land and the ecological protection red line area, trigger the ant colony algorithm to reselect the location. The ant colony algorithm simulates the foraging behavior of ants and finds the optimal path or location through iterative optimization.

[0088] Since the Delaunay triangulation algorithm can effectively decompose complex geographical spaces, enabling the characteristics of each piece of land to be evaluated separately, ensuring the scientificity and rationality of the selection of industrial land; by setting the screening criteria for slope and area, it is possible to avoid development in unsuitable areas, reducing the risk of environmental damage and economic costs. The application of the ant colony algorithm can quickly find alternative solutions while protecting the ecological environment, ensuring the sustainability of the planning scheme.

[0089] S330, set the corresponding functional transition period, which includes the industrial-residential transition area and the industrial-public service transition area; and conduct the corresponding street network density grading and the corresponding line-of-sight corridor analysis.

[0090] Among them, after completing the division of industrial land, the system further sets up functional transition areas to promote the harmonious connection between different functional areas. Specifically, the functional transition areas are divided into two categories: the industrial-residential transition area and the industrial-public service transition area. The industrial-residential transition area usually arranges some small-scale service facilities with low noise and low pollution to alleviate the interference of industrial activities on residents' lives; the industrial-public service transition area is equipped with public service facilities such as education, medical care, and culture to serve industrial employees and their families.

[0091] On this basis, a hierarchical design of the street network density is carried out. For example, main roads with a lower density are adopted in the core industrial area, while the density of secondary roads and branch roads is increased in residential areas or public service areas to meet different traffic demands. In addition, a sight line corridor analysis is carried out. Through 3D modeling and field of view analysis, it is ensured that important landscape nodes remain visually connected and high-rise buildings are prevented from blocking key views.

[0092] The setting of the functional transition area can alleviate the possible conflicts between the old and new functional areas in the short term and lay a good foundation for the long-term urban development plan. The industrial-residential transition area can effectively reduce the impact of noise and pollution on residents' lives and improve the overall livability of the area.

[0093] The hierarchical design of the street network density helps to optimize the traffic flow, reduce congestion, and meet the needs of different functional areas at the same time; the sight line corridor analysis takes into account urban aesthetics and the mental health of residents, ensuring that the newly built area is both beautiful and livable, and also improving the overall quality and attractiveness of the city.

[0094] In the embodiment of the present application, during the corresponding sight line corridor analysis process, the method further includes: S331, determining the geographical location of the landmark building in the geographic information system and height .

[0095] Among them, first, the specific geographical location coordinates of the landmark building are obtained through the geographic information system (GIS) and its actual height . These data are usually stored in the GIS database and can be obtained by querying a specific layer or attribute table. Specifically, for the geographical location: use the spatial query function in the GIS software to locate the position of the landmark building; for the height: extract the height information of the landmark building from the attribute table or perform precise measurement through the digital elevation model.

[0096] It should be noted here that accurate geographical location and height information are the basis of the sight line corridor analysis, so as to ensure the accuracy of the position and height data of the landmark building.

[0097] S332, calculating the corresponding visible area based on the geographical location where the calculation formula is , that is, the set of all observation points not blocked by obstacles.

[0098] Among them, based on the geographical location of the landmark building , calculate the set of all observation points not blocked by obstacles . The specific steps are as follows: A reasonable field of view range needs to be set, for example, an area within a certain radius centered on a landmark building. For each potential observation point , calculate the line-of-sight path from this point to the landmark building.

[0099] Meanwhile, it is necessary to check whether there are any obstacles on the line-of-sight path. If the line-of-sight path intersects with an obstacle, this observation point is considered invisible; gather all the observation points that are not blocked by obstacles to form a visible area.

[0100] By calculating the visible area, the visibility of the landmark building in different directions can be determined, helping planners understand which areas can be clearly seen from the landmark building, thereby optimizing the urban landscape layout. In addition, determining the visible area helps protect the visual connectivity of important landscape nodes and enhances the overall beauty of the city.

[0101] S333. Obtain the actual height of the landmark building, and determine the corresponding restricted building height based on the actual height. The calculation formula is , and obtain the corresponding visible area vector map and building height limit rule file.

[0102] Among them, obtain the actual height of the landmark building , and determine the maximum allowable height of the surrounding buildings based on this height . The specific steps are as follows: Extract the actual height of the landmark building from the GIS database. Calculate the restricted building height: According to the formula , calculate the maximum allowable height of the surrounding buildings. This means that the height of the surrounding buildings cannot exceed 80% of the height of the landmark building and cannot exceed 60 meters; generate a visible area vector map and a building height limit rule file based on the calculation results for guiding subsequent urban planning and building design.

[0103] By determining the maximum allowable height of the surrounding buildings, the visual prominence and integrity of the landmark building can be effectively protected, avoiding new buildings from blocking it. The building height limit rule file provides specific guidance for planners to ensure that the design of new buildings meets the requirements of landscape protection while maintaining the harmony and unity of the city.

[0104] Through the above steps, not only can the visible area of the landmark building be accurately determined, but also the height of the surrounding buildings can be reasonably restricted to ensure the unique status and visual effect of the landmark building in the urban landscape.

[0105] In the embodiment of the present application, the method further includes: S610. Monitor whether the land development intensity exceeds 89% based on the land use change situation or detect a sudden change in industrial demand caused by a change in the published information, and obtain the corresponding monitoring result.

[0106] Among them, after the preliminary land use plan is completed, the system will regularly update and analyze the latest land use data to monitor changes in land development intensity. First, the development intensity of each piece of land in the region is accurately calculated through geographic information system technology. The development intensity is the ratio of the developed area to the total developable area.

[0107] If the development intensity of a certain area exceeds 89%, it is marked as a high-risk area and relevant data is recorded. At the same time, the system continuously monitors the dynamics of released information. Once it is found that there is a significant change in industrial demand due to changes in released information, a monitoring report is immediately generated, including the location, scale of the affected area, and the predicted degree of impact.

[0108] By monitoring the changes in land development intensity and released information in real time, potential risk points and development opportunities can be identified in a timely manner, which helps to take preventive measures in advance to avoid overdevelopment and ensure the sustainability and adaptability of urban development.

[0109] S620, determine whether the monitoring result triggers an early warning. The early warning is divided into a red early warning and a yellow early warning; if it is a red early warning, perform the operation of re-running the parametric model to generate a new layout plan. If it is a yellow early warning, take measures to optimize the function mixing ratio or increase the green area, generate the corresponding early warning report, and provide feedback on the early warning report.

[0110] Among them, according to the obtained monitoring results, the system will automatically evaluate whether an early warning needs to be triggered; when the monitoring results show that the development intensity of a certain area exceeds 89%, the system will activate the red early warning mechanism. At this time, it is necessary to immediately re-run the parametric model to generate a brand-new industrial land layout plan.

[0111] On the other hand, if the monitoring results indicate that although there are certain risks but the overall situation is controllable, for example, the development intensity in some local areas is close to but does not exceed the threshold, or based on the released information, it may only have a limited impact on specific industries, then a yellow early warning is triggered. For the yellow early warning, the system recommends a series of optimization measures, such as adjusting the ratio of different functional areas in the region to achieve a better balance, or appropriately increasing the green area to improve the environmental quality. Subsequently, the system automatically generates a detailed early warning report, outlining the current situation, the specific improvement measures proposed and their expected effects, and provides feedback to relevant departments and stakeholders.

[0112] The establishment of the early warning system can effectively distinguish risks at different levels and accordingly adopt corresponding coping strategies. The emergency response mechanism under the red warning ensures that the planning direction can be quickly adjusted in the face of serious challenges, reducing negative impacts; while the yellow warning provides a more flexible solution to help optimize the existing layout and improve the overall livability and attractiveness of the area. The timely feedback of the early warning report promotes information transparency and public participation in the decision-making process, enhancing the feasibility and social acceptance of the planning implementation.

[0113] In the embodiment of the present application, during the optimization of the yellow warning, the method further includes: S621, when the land development intensity exceeds the preset development threshold or the regional function coordination degree index exceeds the preset scheduling threshold, where the calculation formula of the regional function coordination degree index is .

[0114] Among them, the land development intensity of a specified area is regularly evaluated, that is, the ratio of the developed area to the total developable area. It should be noted here that the preset development threshold is 80%. If this ratio exceeds 80%, it is considered that the area is close to the critical point of overdevelopment; if the regional function coordination degree index is greater than 1.2, that is, in the embodiment of the present application, the preset scheduling threshold is 1.2, which means that the number of jobs per resident on average exceeds 1.2, which may imply that there are more job opportunities than the actual housing demand of the population, or the housing supply is insufficient, resulting in increased commuting pressure, traffic congestion and other problems.

[0115] When it is identified that the land development intensity of a specific area exceeds 80% or the regional function coordination degree index is greater than 1.2, the system starts the optimization process. First, in-depth analysis is carried out on these high-risk areas to clarify the specific reasons for the problems; then, the linear programming method is used to optimize the ratio of commercial facilities, green spaces and public service facilities in the area.

[0116] By setting specific thresholds, areas that need to be prioritized for attention and optimization can be accurately identified, ensuring that urban planning can respond to potential problems in a timely manner and avoiding waste of resources and environmental deterioration.

[0117] By collecting and sorting out data on existing commercial, green space and public service facilities in the area, including information such as their locations, scales and service capabilities; a linear programming model is established based on the collected data, and the linear programming model aims to maximize the quality of life of residents.

[0118] Solve the above linear programming model using an appropriate algorithm to obtain the optimal proportion allocation plan for commercial, green space, and public service facilities. This process also includes simulations under different scenarios to evaluate the actual effects of each plan. Based on the results output by the linear programming model, formulate specific optimization measures, such as increasing the green space area, adjusting the layout of commercial facilities, and expanding the service scope of public service facilities.

[0119] Through the above process, the demands of various types of land within the region can be effectively balanced, the pressure brought about by overdevelopment of land or imbalance in functional matching can be alleviated, and the quality of life of residents can be improved.

[0120] S622, optimize the proportion of commercial, green space, and public services through linear programming, and the corresponding optimization objective calculation formula is , obtain the corresponding optimization results, and determine the corresponding new layout plan and early warning report based on the optimization results.

[0121] Among them, collect and organize data on existing commercial, green space, and public service facilities within the region, including information such as their locations, scales, and service capabilities. Based on the collected data, establish a linear programming model with the goal of maximizing the quality of life of residents, such as minimizing the distance of residents to the nearest service facility and maximizing the green space coverage rate. Solve the above linear programming model using an appropriate algorithm to obtain the optimal proportion allocation plan for commercial, green space, and public service facilities. This process also includes simulations under different scenarios to evaluate the actual effects of each plan. Based on the results output by the linear programming model, formulate specific optimization measures, such as increasing the green space area, adjusting the layout of commercial facilities, and expanding the service scope of public service facilities.

[0122] According to the optimal proportion allocation plan obtained from the linear programming, design a new regional layout; set up functional transition areas, specifically divided into industrial-residential transition areas and industrial-public service transition areas. Arrange some small service facilities with low noise and low pollution in the industrial-residential transition area to reduce the interference of industrial activities on residents' lives; while configure public service facilities such as education, medical care, and culture in the industrial-public service transition area to serve industrial employees and their families.

[0123] Based on the new layout plan, conduct a hierarchical design of the street network density. Adopt main roads with a lower density in the core industrial area, while increase the density of secondary roads and branch roads in the residential area or public service area to meet different traffic demands. Through 3D modeling and visibility analysis, ensure the visual connectivity of important landscape nodes and avoid high-rise buildings blocking key views.

[0124] Early Warning Report Compilation and Feedback: Prepare a detailed early warning report covering existing problems, optimization measures taken and their expected effects, an overview of the new layout plan, etc., and provide feedback to relevant departments and stakeholders to ensure that all affected people can understand the situation and participate in subsequent discussions.

[0125] Through the above process, not only can the problems of excessive land development intensity and unbalanced functional matching be effectively solved, but also the harmonious connection between different functional areas can be promoted, and the overall quality and attractiveness of the city can be improved.

[0126] In the embodiment of the present application, the parametric modeling rules include a dynamic space optimization module based on reinforcement learning, and the method further includes: S710, retrieve the pre-defined state space , where the state space includes the current land use layout, future industry prediction data, environmental compatibility parameters, development intensity, and functional matching ratio; retrieve the pre-defined action space , and the action space can change the functional type of the plot, adjust the plot area, and modify the building height limit.

[0127] Among them, before performing dynamic space optimization, the system first retrieves the pre-defined state space, which contains multiple key elements: the current land use layout reflects the existing functions of each plot; the future industry prediction data provides an analysis of the industrial development trend in the region; the environmental compatibility parameters clarify the location and restrictive conditions of the ecological protection area; the development intensity refers to the ratio of the developed area to the total developable area; and the functional matching ratio is the ratio of the total number of jobs to the number of resident population. By integrating multiple data sources to form a comprehensive state space, it can be ensured that all important factors are considered in the optimization process, thus formulating a more scientific and reasonable planning scheme.

[0128] S720, retrieve the corresponding reward function , where , , is the weight coefficient, the ecological protection area overlap penalty term is negative and proportional to the overlap area, the development intensity target is 89%, the functional matching balance threshold is 0.8 - 1.2, the deep Q network is used as the reinforcement learning algorithm, the number of neurons in the hidden layer ≥ 128, the Adam optimizer is used, and the learning rate is 0.001.

[0129] Among them, the system retrieved a pre-defined action space that allows for changes in the functional type of plots, such as changing from industrial land to commercial land; adjusting the plot area to meet new demands; and modifying building height restrictions to promote more efficient land use. Each action is aimed at achieving specific goals, such as increasing land utilization or improving the quality of life of residents.

[0130] Defining a clear action space provides specific means to solve practical problems, making the optimization process not limited to the theoretical level but directly applicable to actual urban planning.

[0131] S730, perform corresponding multi-objective optimization, optimization function: , obtain the dynamic adjustment of the reward function weight according to the real-time enterprise migration rate data and the real-time release information change frequency data, Increase as the development intensity approaches the threshold.

[0132] Among them, at this stage, the system executed multi-objective optimization, aiming to balance the relationship between development intensity, functional matching balance and ecological protection; during the optimization process, the weight in the reward function was dynamically adjusted according to the real-time enterprise migration rate data and the release information change frequency data. Especially when the development intensity approaches the threshold of 89%, the relevant weight will be increased accordingly; this flexibility enables the plan to better respond to changes in the external environment and ensures long-term sustainability. The process of dynamically adjusting the reward function weight enhances the adaptability and response speed of the model, enabling it to remain effective in a rapidly changing environment.

[0133] S740, generate the Pareto front through the NSGA-II algorithm, select the comprehensive optimal solution as the final layout plan, determine it as the industrial land layout plan data and output the decision report. The industrial land layout plan data includes the vector data of the multi-objective optimization results, and marks the development intensity, functional matching ratio, and ecological impact of each region. The decision report records the weight adjustment path, Pareto front graph, and Monte Carlo simulation prediction data during the reinforcement learning process.

[0134] Finally, generate the Pareto front through the Non-dominated Sorting Genetic Algorithm II (NSGA-II), and select the comprehensive optimal solution from it as the final industrial land layout plan; the industrial land layout plan not only includes the results of multi-objective optimization, but also details the specific indicators of each region, such as development intensity, functional matching ratio, ecological impact, etc. At the same time, the decision report details the key information during the entire reinforcement learning process, including the weight adjustment path, Pareto front graph, and Monte Carlo simulation prediction data, providing comprehensive support for decision-makers.

[0135] The Pareto front generated by the NSGA-II algorithm helps identify multiple feasible solutions, while the detailed decision report improves transparency and decision-making quality, contributing to the implementation of more reasonable and effective urban planning strategies.

[0136] Through the deep integration and intelligent analysis of multi-source heterogeneous data, an industrial space planning system with high adaptability is constructed. By integrating multi-dimensional data such as geography, economy, and published information, and combining the LSTM-ARIMA hybrid model and BERT published information influence analysis technology, it can accurately predict industrial trends and the impact of published information, and dynamically adjust the parameters of the prediction model, thereby embedding the sensitivity to market fluctuations and changes in published information at the initial stage of planning, significantly enhancing the foresight and flexibility of the planning scheme.

[0137] By adopting reinforcement learning and multi-objective optimization techniques, the dynamic adaptive adjustment of the planning scheme is achieved. Through a hierarchical early warning mechanism, the system can quickly respond to sudden changes: a red warning triggers a recalculation of the global model to reconstruct the layout, and a yellow warning corrects the deviation precisely through local optimization, forming a closed loop of "monitoring - early warning - response".

[0138] Finally, through cross-level data association, real-time remote sensing monitoring, and automated generation technology, an intelligent and flexible industrial space layout solution is constructed; multi-scale data fusion and adaptive GIS resolution adjustment enable the planning to not only overall plan the whole but also be accurately implemented; and the optimization based on linear programming and NSGA-II algorithm ensures the scientific nature of the scheme in terms of spatial aesthetics, functional mixing, and ecological compatibility, providing iterative and verifiable intelligent decision-making support for the sustainable development of the urban industrial space.

[0139] An embodiment of the present application discloses an industrial space adaptability optimization system, referring to Figure 2 , including: A multi-source dataset collection module 001, which obtains corresponding geographic information data, industrial data, industrial service-related data, enterprise information, and infrastructure data; performs data cleaning and standardization processing on the geographic information data, industrial data, industrial service-related data, enterprise information, and infrastructure data for collecting the corresponding multi-source dataset, associates the multi-source dataset with geographic space coordinates and enters it into the database. Among them, the geographic information data includes the geographical location, terrain, and surrounding environment of the industrial park, the industrial data includes industrial scale parameters, enterprise operation indicators, and industry classification data, the industrial service-related data includes industrial development-related information and resource planning information, the enterprise information includes industrial and commercial registration and migration records, and the infrastructure data includes transportation road networks and public facilities; The future industry prediction data update and acquisition module 002 retrieves the pre-trained LSTM-ARIMA hybrid model, inputs the standardized multi-source data set into the LSTM-ARIMA hybrid model to obtain the corresponding future industry prediction data. The future industry prediction data includes the industrial growth probability distribution, industrial transformation probability distribution, and human resource demand prediction data for the next 3 to 5 years. The LSTM-ARIMA hybrid model is obtained through training and convergence by inputting historical industrial data, industrial service-related data, population migration data, and land transaction data. The historical industrial data includes macroeconomic indicators, output value, and labor market data. Among them, the LSTM-ARIMA hybrid model includes an LSTM layer and an ARIMA layer. The number of hidden layer neurons in the LSTM layer is ≥64 and is used to capture non-linear time series relationships. The ARIMA layer optimizes the parameters (p, d, q) through the AIC criterion and is used to correct the residual sequence of the LSTM layer; the loss function is , where λ is the weight coefficient, and its value range is [0, 1], RMSE is the root mean square error, and MAE is the mean absolute error; retrieve the pre-trained BERT model, input the standardized industrial service-related data into the BERT model for classification. The types of industrial service-related data include high-impact types, low-impact types, and no obvious impact types. Extract the keyword data from the classified industrial service-related data through TF-IDF, and calculate the corresponding text influence coefficient based on the keyword data. The calculation formula is , where is the text influence coefficient, is the sum of all keywords in the industrial service-related data, is the influence polarity of the keyword, is the value of the keyword, indicating the importance of the keyword; adjust the corresponding weight coefficient based on the keyword list and the corresponding TF-IDF value in the keyword data to obtain the corresponding adjusted weight coefficient, and apply the adjusted weight coefficient to the LSTM-ARIMA hybrid model to obtain the updated future industry prediction data; The industrial land layout plan data generation module 003 retrieves the pre-set parametric modeling rules, optimizes the geographic information data and the updated future industry prediction data based on the parametric modeling rules, and is used to generate the optimized industrial land layout plan data; The unified database acquisition module 004 determines the planned scale data selected by the user. The planned scale data includes the municipal level corresponding to the macro level, the district level corresponding to the meso level, the park level corresponding to the micro level, and the plot level corresponding to the fine-grained level. Based on the planned scale data, the corresponding hierarchical data sets are filtered and the GIS resolution is adjusted. Among them, the macro level integrates macroeconomic indicators, total population, and transportation framework; the meso level integrates industrial land distribution and road network density; the micro level integrates building height and open space ratio; the fine-grained level integrates terrain slope and property boundary, which is used to obtain a unified database for cross-hierarchical data association. Among them, the unified database includes data sets at different levels and their corresponding GIS resolutions; The data lag comparison module 005 monitors the land use change situation in the specified area based on satellite remote sensing image technology. The land use change situation includes the construction and demolition of buildings and the change of green spaces, and the corresponding enterprise activity status data is obtained in real time, which is used to compare the land use change situation and enterprise activity status data with the data in the unified database. When it is detected that there is a data update lag, a corresponding user verification operation is prompted; before or at the same time as the user verification, the missing data is predictively filled based on the detected surrounding change trends. The surrounding change trends include the development intensity of adjacent plots and the concentration of enterprise activities, and the latest data after verification and predictive filling is updated to the unified database.

[0140] The embodiment of the present application also discloses an industrial space adaptability optimization method, including a processor, and a program of the industrial space adaptability optimization method described in any one of the above is run in the processor.

[0141] The embodiment of the present application also discloses a storage medium storing a program of the industrial space adaptability optimization method described in any one of the above.

[0142] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for optimizing the adaptability of industrial space, characterized in that: include: Obtain corresponding geographic information data, industry data, industry service-related data, enterprise information and infrastructure data; Clean and standardize geographic information data, industrial data, industrial service-related data, enterprise information, and infrastructure data, collect them into corresponding multi-source data sets, associate the multi-source data sets with geographic spatial coordinates, and enter them into a database, wherein the geographic information data includes the geographical location, topography, and surrounding environment of the industrial park, the industrial data includes industrial scale parameters, enterprise operating indicators, and industry classification data, the industrial service-related data includes information related to industrial development and resource planning information, the enterprise information includes industrial and commercial registration and migration records, and the infrastructure data includes transportation network and public facilities; Retrieve the pre-trained LSTM-ARIMA hybrid model, input the standardized multi-source data set into the LSTM-ARIMA hybrid model, and obtain the corresponding future industry forecast data. The future industry forecast data includes the probability distribution of industry growth, the probability distribution of industry transformation, and the forecast data of human resource demand in the next 3 to 5 years. The LSTM-ARIMA hybrid model is obtained by inputting historical industry data, industry service-related data, population migration data, and land transaction data for training and convergence. The historical industry data includes macroeconomic indicators, output value, and labor market data. Among them, the LSTM-ARIMA hybrid model includes an LSTM layer and an ARIMA layer. The number of hidden layer neurons in the LSTM layer is ≥64 and is used to capture nonlinear time series relationships. The ARIMA layer optimizes the parameters (p, d, q) through the AIC criterion and is used to correct the residual sequence of the LSTM layer; the loss function is , where λ is the weight coefficient, ranging from [0, 1], RMSE is the root mean square error, and MAE is the mean absolute error; the pre-trained BERT model is retrieved, and the standardized industry service-related data is input into the BERT model for classification. The types of industry service-related data include high-impact, low-impact, and no-significant-impact. The keyword data in the classified industry service-related data is extracted through TF-IDF, and the corresponding text influence coefficient is calculated based on the keyword data. The calculation formula is: ,in, is the text influence coefficient, To sum all the keywords in the industry service related data, is the influence polarity of the keyword, For keywords value, indicating the importance of the keyword; adjusting the corresponding weight coefficient based on the keyword list in the keyword data and the corresponding TF-IDF value, obtaining the corresponding adjusted weight coefficient, applying the adjusted weight coefficient to the LSTM-ARIMA hybrid model, and obtaining updated future industry forecast data; Retrieve pre-set parametric modeling rules, optimize geographic information data and updated future industry forecast data based on the parametric modeling rules, and generate optimized industrial land layout plan data; Determine the planning scale data selected by the user. The planning scale data includes the city level corresponding to the macro level, the district level corresponding to the meso level, the park level corresponding to the micro level, and the plot level corresponding to the fine-grained level. Based on the planning scale data, select the corresponding level of data sets and adjust the GIS resolution. The macro level integrates macroeconomic indicators, total population, and transportation framework. The meso level integrates the distribution of industrial land and road network density. The micro level integrates building height and open space ratio. The fine-grained level integrates terrain slope and property boundaries. Obtain a unified database of cross-level data associations. The unified database includes data sets of different levels and their corresponding GIS resolutions. Based on satellite remote sensing imaging technology, the changes in land use in designated areas are monitored. The changes in land use include the construction and demolition of buildings, and the changes in green areas. The corresponding enterprise activity status data is acquired in real time, and the land use changes and the enterprise activity status data are compared with the data in the unified database. When a data update lag is detected, a prompt is given to perform the corresponding user verification operation. Before or at the same time as user verification, the missing data is predictively filled based on the detected surrounding change trends. The surrounding change trends include the development intensity of adjacent plots and the concentration of enterprise activities. The latest data after verification and predictive filling is updated to the unified database.

2. The industrial space adaptability optimization method according to claim 1 is characterized in that: During the calculation of the text influence coefficient, the method further includes: Construct a corresponding influence dictionary, where the influence dictionary contains positive keywords and negative keywords; Perform word segmentation on the industrial service related data to extract candidate keywords; calculate the TF-IDF weight of each keyword, the formula is: , where N is the total number of published information documents, is the number of documents containing the keyword, and the published information is the content of data related to industrial services; According to the positive and negative classification of the influence dictionary, the text influence coefficient is calculated: , where the text influence coefficient is combined with the keyword weight to generate a dynamic adjustment coefficient, the formula is: ,in, is the pre-set basic weight coefficient, For keywords value, It is used to control the influence of the influence coefficient on the weight, and its value range is 0.3 to 0.

7.

3. The industrial space adaptability optimization method according to claim 2 is characterized in that: In the process of generating optimized industrial land layout plan data, the method further includes: Retrieve geographic information data, which also includes road network, digital elevation model, and industrial space outline vector data, resample the digital elevation model to a five-meter by five-meter grid, and fill in the missing sections of the road network; The industrial plots in the future industrial forecast data are divided based on the Delaunay triangulation algorithm, and plots with a slope greater than 5% or an area less than two hectares are screened and eliminated; if the planned land overlaps with the ecological protection area, the ant colony algorithm is triggered to reselect the site; Set up corresponding functional transition periods, which include industry-residential transition zones and industry-public service transition zones; and conduct corresponding street network density grading and corresponding visual corridor analysis.

4. The industrial space adaptability optimization method according to claim 3 is characterized in that: In the process of performing the corresponding sight corridor analysis, the method also includes: Determine the geographical location of landmark buildings in the geographic information system and height ; Based on the geographical location Calculate the corresponding visible area, where the calculation formula is , that is, the set of all observation points that are not blocked by obstacles; Get the actual height of the landmark building, and determine the corresponding constrained building height based on the actual height. The calculation formula is: , obtain the corresponding visible area vector map and building height limit rule file.

5. The industrial space adaptability optimization method according to claim 4 is characterized in that: The method also includes: Based on the land use change, monitor whether the land development intensity exceeds 89% or detect a sudden change in industrial demand caused by a change in published information, and obtain corresponding monitoring results; Determine whether the monitoring results trigger an early warning, which is divided into red and yellow warnings. If it is a red warning, re-run the parameterized model to generate a new layout plan. If it is a yellow warning, optimize the functional mixing ratio or increase the green area, generate a corresponding early warning report, and provide feedback on the early warning report.

6. The industrial space adaptability optimization method according to claim 5 is characterized in that: During the optimization process of the yellow warning, the method further includes: When the land development intensity exceeds the preset development threshold or the regional function coordination index exceeds the preset scheduling threshold, the regional function coordination index calculation formula is: ; The proportion of commercial, green space and public services is optimized through linear programming, and the corresponding optimization target calculation formula is: , obtain corresponding optimization results, and determine corresponding new layout plans and early warning reports based on the optimization results.

7. The industrial space adaptability optimization method according to claim 6, characterized in that: The parameterized modeling rules include a dynamic spatial optimization module based on reinforcement learning, and the method further includes: Recalling a predefined state space , where the state space Contains current land use layout, future industry forecast data, environmental compatibility parameters, development intensity, and functional matching ratio; retrieves pre-defined action space , action space Ability to change the functional type of the plot, adjust the plot area, and modify the building height limit; Call the corresponding reward function ,in, , , is the weight coefficient, the ecological protection zone overlap penalty term is negative and proportional to the overlapping area, the development intensity target is 89%, the functional matching balance threshold is 0.8-1.2, the deep Q network is used as the reinforcement learning algorithm, the number of hidden layer neurons is ≥128, the Adam optimizer is used, and the learning rate is 0.001; Perform corresponding multi-objective optimization and optimize the function: , obtain the reward function weight dynamically adjusted according to the real-time enterprise migration rate data and the real-time release information change frequency data, It increases as the development intensity approaches the threshold; The Pareto frontier is generated through the NSGA-II algorithm, and the comprehensive optimal solution is selected as the final layout plan. The industrial land layout plan data is determined and the decision report is output. The industrial land layout plan data contains vector data of multi-objective optimization results, and the development intensity, functional matching ratio, and ecological impact of each region are marked. The decision report records the weight adjustment path, Pareto frontier diagram and Monte Carlo simulation prediction data in the reinforcement learning process.

8. An industrial space adaptability optimization system, characterized in that: include: Multi-source data set collection module, to obtain corresponding geographic information data, industry data, industry service related data, enterprise information and infrastructure data; Data cleaning and standardization processing is performed on geographic information data, industrial data, industrial service-related data, enterprise information, and infrastructure data to collect corresponding multi-source data sets, associate the multi-source data sets with geographic spatial coordinates, and enter them into a database, wherein the geographic information data includes the geographical location, topography, and surrounding environment of the industrial park, the industrial data includes industrial scale parameters, enterprise operating indicators, and industry classification data, the industrial service-related data includes information related to industrial development and resource planning information, the enterprise information includes industrial and commercial registration and migration records, and the infrastructure data includes transportation network and public facilities; The future industry forecast data update acquisition module calls the pre-trained LSTM-ARIMA hybrid model, inputs the standardized multi-source data set into the LSTM-ARIMA hybrid model, and obtains the corresponding future industry forecast data. The future industry forecast data includes the probability distribution of industry growth, the probability distribution of industry transformation, and the forecast data of human resource demand in the next 3 to 5 years. The LSTM-ARIMA hybrid model is obtained by inputting historical industry data, industry service-related data, population migration data, and land transaction data for training convergence. The historical industry data includes macroeconomic indicators, output value, and labor market data. Among them, the LSTM-ARIMA hybrid model includes an LSTM layer and an ARIMA layer. The number of hidden layer neurons of the LSTM layer is ≥64 and is used to capture nonlinear time series relationships. The ARIMA layer optimizes the parameters (p, d, q) through the AIC criterion and is used to correct the residual sequence of the LSTM layer; the loss function is , where λ is the weight coefficient, ranging from [0, 1], RMSE is the root mean square error, and MAE is the mean absolute error; the pre-trained BERT model is retrieved, and the standardized industry service-related data is input into the BERT model for classification. The types of industry service-related data include high-impact, low-impact, and no-significant-impact. The keyword data in the classified industry service-related data is extracted through TF-IDF, and the corresponding text influence coefficient is calculated based on the keyword data. The calculation formula is: ,in, is the text influence coefficient, To sum all the keywords in the industry service related data, is the influence polarity of the keyword, For keywords value, indicating the importance of the keyword; adjusting the corresponding weight coefficient based on the keyword list in the keyword data and the corresponding TF-IDF value, obtaining the corresponding adjusted weight coefficient, and applying the adjusted weight coefficient to the LSTM-ARIMA hybrid model to obtain updated future industry forecast data; The industrial land layout plan data generation module calls the pre-set parametric modeling rules, optimizes the geographic information data and the updated future industry forecast data based on the parametric modeling rules, and uses them to generate the optimized industrial land layout plan data; The unified database acquisition module determines the planning scale data selected by the user. The planning scale data includes the city level corresponding to the macro level, the district level corresponding to the meso level, the park level corresponding to the micro level, and the plot level corresponding to the fine-grained level. Based on the planning scale data, the corresponding level of data sets are selected and the GIS resolution is adjusted. The macro level integrates macroeconomic indicators, total population, and transportation framework. The meso level integrates the distribution of industrial land and road network density. The micro level integrates building height and open space ratio. The fine-grained level integrates terrain slope and property boundaries. It is used to obtain a unified database for cross-level data association. The unified database includes data sets of different levels and their corresponding GIS resolutions. The data lag comparison module monitors the land use changes in the designated area based on satellite remote sensing imaging technology. The land use changes include the construction and demolition of buildings and the changes in green spaces. The module obtains the corresponding enterprise activity status data in real time and compares the land use changes and the enterprise activity status data with the data in the unified database. When a data update lag is detected, the module prompts the user to perform the corresponding verification operation. Before or at the same time as the user verification, the module predictively fills in the missing data based on the detected surrounding change trends. The surrounding change trends include the development intensity of adjacent plots and the concentration of enterprise activities. The latest data after verification and predictive filling is updated to the unified database.

9. An industrial space adaptability optimization system, characterized in that: It comprises a processor, in which runs a program of the industrial space adaptability optimization method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: A program storing the industrial space adaptability optimization method as described in any one of claims 1 to 7.

Images