A method and system for railway land development potential assessment and decision-making

Abstract

The present application provides a railway land development potential evaluation and decision-making method and system, which comprises the following steps: acquiring multi-source data of railway land and performing standardization processing to construct a railway land feature vector; inputting the railway land feature vector into a pre-trained gradient boosting decision tree model to output a development potential score corresponding to the railway land; using an explainability analysis algorithm to perform explainability analysis on the development potential score to calculate the contribution value of each dimension feature to the development potential score; based on the contribution value of each dimension feature, calculating the matching degree between the railway land feature vector and a plurality of preset development types, and generating a development type recommendation list in descending order of the matching degree; and generating a railway land comprehensive development strategy report based on the development potential score, the contribution value and the development type recommendation list. The present application fuses multi-source data for nonlinear modeling, provides an explainable decision basis, and effectively improves the accuracy, transparency and practicality of railway land development evaluation.

Description

Technical Field

[0001] This invention relates to the field of railway land development technology, and in particular to a method and system for assessing and making decisions on the development potential of railway land. Background Technology

[0002] With the continuous expansion of my country's railway network and the acceleration of urbanization, the vast amount of existing land resources along railway lines and around station areas are gradually becoming an important vehicle for railway transportation enterprises to realize asset appreciation, optimize resource allocation, and promote regional economic development. The comprehensive development of railway land can not only revitalize idle assets but also promote the integration of transportation and urban functions, improving land use efficiency. However, facing an increasingly complex market environment and diversified development demands, how to scientifically and accurately assess the development potential of railway land and match it with suitable business formats has become a key issue restricting the efficient use of railway land.

[0003] Current assessments of railway land development primarily rely on manual experience and single-indicator methods. Assessors typically make qualitative judgments based on a few explicit factors such as land area and location, or use simple linear weighted quantification. This results in highly subjective assessments lacking unified standards, and the limited data sources make it difficult to comprehensively reflect the combined impact of multiple dimensions of information, including accessibility, surrounding infrastructure, socio-economic factors, and policy guidance. Furthermore, traditional linear models cannot effectively capture the complex nonlinear relationships between various influencing factors, limiting predictive accuracy. The internal mechanisms of quantitative models are often considered black boxes, preventing decision-makers from understanding the specific contribution direction and intensity of each factor to the score. This lack of transparency and trust in sensitive areas such as asset disposal and cooperative development increases decision-making risks.

[0004] Therefore, there is an urgent need in this field for a comprehensive assessment method and device for railway land use that can integrate multi-source data, achieve nonlinear modeling, and provide interpretable decision-making basis, so as to improve the accuracy, transparency and practicality of the assessment and provide strong support for the efficient use of railway land resources and scientific decision-making. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method and system for assessing and making decisions on the development potential of railway land, in order to solve the problems of strong subjectivity, single data dimension, insufficient nonlinear modeling ability, and opaque decision-making process caused by reliance on human experience and single indicators in the existing railway land development assessment.

[0006] One aspect of the present invention provides a method for assessing and deciding on the development potential of railway land, the method comprising the following steps: Acquire multi-source data on railway land use, including basic land attribute data, location feature data, environmental feature data, and socio-economic feature data, and standardize the multi-source data to construct a railway land use feature vector. The railway land feature vector is input into a pre-trained gradient boosting decision tree model, which outputs the development potential score corresponding to the railway land. The gradient boosting decision tree model is trained based on historical railway land sample data and its corresponding development potential score labels. An interpretability analysis algorithm performs interpretability analysis on the development potential score output by the gradient boosting decision tree model, and calculates the contribution value of each dimension feature in the railway land feature vector to the development potential score. The contribution value includes the contribution direction and contribution intensity. Based on the contribution values ​​of each dimension of the railway land feature vector, the matching degree between the railway land feature vector and multiple preset development types is calculated; a recommended list of development types is generated in descending order of matching degree; the development types include one or more of commercial, residential, industrial, logistics, and park green space. Based on the development potential score, the contribution value, and the recommended list of development types, a comprehensive development strategy report for railway land is generated.

[0007] In some embodiments of the present invention, the pre-training step of the gradient boosting decision tree model includes: A training sample set is constructed, which includes multiple historical railway land use sample data. Each historical railway land use sample data contains the multi-source data as input features and the development potential score label corresponding to the multi-source data. The training sample set is preprocessed, including missing data imputation, outlier correction, normalization or standardization, and encoding conversion of categorical data. Set the model parameters of the gradient boosting decision tree model, wherein the model parameters include at least the maximum depth of the decision tree, the learning rate, the subsample sampling ratio, and the regularization parameter; The preprocessed training sample set is input into the gradient boosting decision tree model. Through multiple rounds of iterative training, an integrated model consisting of multiple decision trees is generated. In each round of iteration, a new decision tree is generated based on the difference between the current prediction result and the actual development potential score label to correct the prediction error, until the preset number of iterations is reached or the prediction error meets the preset conditions, thus obtaining the pre-trained gradient boosting decision tree model.

[0008] In some embodiments of the present invention, the multi-source data is collected from multiple data sources including railway land valuation reports, geographic information service platforms, meteorological data platforms and statistical data platforms, and the collected multi-source data is cleaned and normalized to construct data entries in a unified format. The processed multi-source data is stored in a structured database according to a preset data structure, and a unique identifier and a trusted timestamp are generated for each data entry. Based on the unique identifier and the trusted timestamp, the hash value of the data entry is uploaded to the blockchain network for evidence storage, so as to achieve traceability and immutability of the data source; The structured database supports dynamic updates and version management of the multi-source data, and sets corresponding data access permissions for different user roles; The location feature data includes distances to railway stations, commercial centers, airports, and public transportation stations. These data are acquired through the spatial distance measurement function of the geographic information service platform, and the measurement results are stored as standardized feature values ​​in the structured database.

[0009] In some embodiments of the present invention, the multi-source data is standardized, including: For numerical data, normalization or standardization methods are used to unify the dimensions. For categorical data, one-hot encoding or dummy variable transformation is used to convert it into a numerical vector. For distance-type data, spatial distance is measured through a geographic information system or map application interface, and the measurement results are used as standardized feature values. Missing value detection and imputation, as well as outlier identification and correction, are performed on the multi-source data.

[0010] In some embodiments of the present invention, the gradient boosting decision tree model is an XGBoost model; the XGBoost model introduces a regularization term during training to prevent overfitting.

[0011] In some embodiments of the present invention, the interpretability analysis algorithm is the SHAP algorithm; the contribution value is the Shapley value, and the formula for calculating the contribution value of each dimension feature to the development potential score is: In some embodiments, the SHAP value can be calculated as follows: ; in, Indicates the first Features The corresponding Shapley value; U represents the set of all the stated features, i.e. P represents the total number of input features; S represents that U does not contain any features. any subset of; This represents the number of features in subset S; This represents the predicted output value of the gradient boosting decision tree model when only a subset of features S is used; This indicates that the feature is added to the feature subset S. The predicted output value of the gradient boosting decision tree model described later.

[0012] In some embodiments of the present invention, the step of calculating the matching degree of multiple preset development types based on the contribution value and the railway land use feature vector includes: Based on the contribution direction and intensity of each dimension of features, identify the key driving factors that have a positive impact on the development potential score and the limiting factors that have a negative impact. Based on the key driving factors and limiting factors, the matching degree between each development type and the railway land is calculated according to the preset development type adaptation rules, which are constructed based on the domain knowledge base; A list of recommended development types is generated in descending order of matching degree.

[0013] In some embodiments of the present invention, the method further includes: Based on the dynamic updates of the multi-source data, the railway land use feature vector is periodically expanded and corrected, and the gradient boosting decision tree model is incrementally trained or retrained using newly added historical railway land use sample data to maintain the timeliness and accuracy of the model.

[0014] On the other hand, the present invention also provides a railway land development potential assessment and decision-making system, including a processor, a memory, and a computer program or instructions stored in the memory, wherein the processor is used to execute the computer program or instructions, and when the computer program or instructions are executed, the system implements the steps of the method described in any of the above.

[0015] On the other hand, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the method as described in any of the preceding claims.

[0016] The present invention provides a method and system for assessing and deciding on the development potential of railway land. This method involves acquiring multi-source data on railway land and standardizing it to construct feature vectors; inputting these feature vectors into a pre-trained gradient boosting decision tree model to output a development potential score; analyzing the score using an interpretability analysis algorithm to calculate the contribution value of each dimension; calculating the matching degree with multiple preset development types based on the contribution values ​​and generating a recommendation list in descending order; and generating a comprehensive development strategy report based on the score, contribution values, and recommendation list. This invention, through the organic combination of gradient boosting decision trees and SHAP interpretability analysis, achieves intelligent processing throughout the entire process from data fusion and nonlinear modeling to interpretable decision-making, significantly improving the accuracy, transparency, and scientific rigor of railway land development assessment.

[0017] Furthermore, by standardizing and storing multi-source data on the blockchain, this invention ensures the credibility and traceability of input data, providing a high-quality data foundation for subsequent evaluation.

[0018] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the description, or may be learned by practice of the invention. The objects and other advantages of the invention can be realized and obtained by means of the structures specifically pointed out in the description and drawings.

[0019] Those skilled in the art will understand that the objectives and advantages achievable with the present invention are not limited to those specifically described above, and that the above and other objectives achievable with the present invention will become clearer from the following detailed description. Attached Figure Description

[0020] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0021] Figure 1 This is a flowchart illustrating the railway land development potential assessment and decision-making method according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the architecture of a railway land development potential assessment and decision-making system according to an embodiment of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this invention are used to explain the invention, but are not intended to limit the invention.

[0024] It should also be noted that, in order to avoid obscuring the invention with unnecessary details, only the structures and / or processing steps closely related to the solution according to the invention are shown in the accompanying drawings, while other details that are not closely related to the invention are omitted.

[0025] As a crucial asset for railway transportation enterprises, the comprehensive development of railway land is of great significance for revitalizing existing resources, promoting asset appreciation, and driving regional economic development. However, current assessments of the development potential of railway land mainly rely on manual experience and single-indicator methods. Assessors typically make qualitative judgments or simple linear weighted quantifications based on a few explicit factors such as land area and location conditions. This results in highly subjective assessments lacking unified standards, and the limited data sources make it difficult to comprehensively reflect the combined impact of multi-dimensional information such as accessibility, surrounding infrastructure, socio-economic factors, and policy guidance. Furthermore, traditional linear models cannot effectively capture the complex nonlinear relationships between various influencing factors, limiting predictive accuracy. The internal mechanisms of quantitative models are often considered black boxes, preventing decision-makers from understanding the specific contribution direction and intensity of each factor to the score. This lack of transparency and trust in sensitive areas such as asset disposal and cooperative development increases decision-making risks and severely restricts the efficient utilization and scientific decision-making of railway land resources.

[0026] In view of this, one aspect of the present invention provides a method for assessing and deciding on the development potential of railway land, such as... Figure 1 As shown, the process includes the following steps S101~S105: Step S101: Obtain multi-source data on railway land use, including basic land attribute data, location feature data, environmental feature data, and socio-economic feature data. Standardize the multi-source data and construct a feature vector for railway land use.

[0027] Step S102: Input the railway land feature vector into the pre-trained gradient boosting decision tree model and output the development potential score corresponding to the railway land. The gradient boosting decision tree model is trained based on historical railway land sample data and its corresponding development potential score labels.

[0028] Step S103: The interpretability analysis algorithm performs interpretability analysis on the development potential score output by the gradient boosting decision tree model, and calculates the contribution value of each dimension feature in the railway land feature vector to the development potential score. The contribution value includes the contribution direction and contribution intensity.

[0029] Step S104: Based on the contribution values ​​of each dimension of the railway land feature vector, calculate the matching degree between the railway land feature vector and multiple preset development types; generate a recommended list of development types in descending order of matching degree; the development types include one or more of commercial, residential, industrial, logistics, and park green space.

[0030] Step S105: Generate a comprehensive development strategy report for railway land based on development potential score, contribution value, and recommended development type list.

[0031] In step S101, multi-source data on railway land is first acquired to provide a data foundation for subsequent development potential assessment. This multi-source data covers four core dimensions: basic land attribute data, location characteristic data, environmental characteristic data, and socio-economic characteristic data. Basic land attribute data includes land area, plot ratio, land rights attributes, land use term, and infrastructure construction status; location characteristic data includes the distance between the land and railway stations, commercial centers, airports, and public transportation stations; environmental characteristic data includes air pollution status and noise pollution status; and socio-economic characteristic data includes population, GDP, per capita disposable income, and policy orientation.

[0032] During the data acquisition phase, multi-source data on railway land use is obtained by integrating multiple data sources, including railway land valuation reports, geographic information service platforms, meteorological data platforms, and statistical data platforms. The collected raw multi-source data needs to be standardized to construct a unified railway land use feature vector, making it suitable for the input requirements of subsequent machine learning models.

[0033] In some embodiments, the specific operations of standardization processing include: for numerical data, such as land area, average annual temperature, and GDP, normalization or standardization methods can be used to unify the dimensions and eliminate the impact of different dimensions on model training; for categorical data, such as land rights attributes and plot conditions, one-hot encoding or dummy variable transformation can be used to convert them into numerical vectors; for ordered categorical data, ordered numerical assignment is used to maintain its hierarchical order information; for distance-based data, such as distances to railway stations, commercial centers, airports, and public transportation stations, spatial distance measurements are performed through geographic information systems or map application programming interfaces, and the measurement results are used as standardized feature values. Furthermore, missing value detection and imputation, as well as outlier identification and correction processing, are required for multi-source data to ensure data quality.

[0034] To further enhance the credibility and traceability of data, some embodiments also introduce data governance mechanisms. Specifically, after cleaning and normalizing the collected multi-source data, it is constructed into data entries in a unified format; the processed multi-source data is stored in a structured database according to a preset data structure, and a unique identifier and a trusted timestamp are generated for each data entry; based on the unique identifier and trusted timestamp, the hash value of the data entry is uploaded to a blockchain network for notarization, realizing the traceability and immutability of the data source; the structured database supports dynamic updates and version management of multi-source data, and sets corresponding data access permissions for different user roles.

[0035] Through the aforementioned standardization and data governance, a railway land use feature vector was finally constructed. This feature vector contains standardized values ​​for each dimension, providing unified, standardized, and reliable input data for subsequent development potential scoring and prediction. Simultaneously, the standardization process eliminated the influence of dimensional differences, missing values, and outliers in the original data, ensuring the quality and consistency of the feature vector, thereby improving the accuracy and stability of subsequent model predictions.

[0036] In step S102, the railway land feature vector obtained in step S101 is input into a pre-trained gradient boosting decision tree model, which outputs a development potential score corresponding to the railway land. The purpose of this step is to utilize the powerful nonlinear modeling capabilities of machine learning models to transform multi-dimensional feature vectors into a quantitative development potential score, providing objective and accurate numerical basis for subsequent business matching and decision-making.

[0037] Gradient boosting decision tree model is an ensemble learning-based algorithm that improves prediction accuracy by combining multiple decision trees. It can effectively capture the complex nonlinear mapping relationship between the features of each dimension in the railway land use feature vector and the development potential score. For example, the impact of land area and plot ratio on development potential may show different effects in different value ranges. Traditional linear models are difficult to characterize such relationships, while gradient boosting decision tree model can flexibly model these nonlinear interactions.

[0038] In some embodiments, the gradient boosting decision tree model needs to be pre-trained. The training process specifically includes sub-steps S1021~S1024: Sub-step S1021: Construct a training sample set, which contains multiple historical railway land use sample data. Each sample data includes multi-source data as input features and a corresponding development potential score label. The multi-source data includes basic land attribute data, location feature data, environmental feature data, and socio-economic feature data. These historical sample data can be derived from railway land parcels for which development potential assessments have been completed or have clear evaluation results by various railway bureaus in the past few years. For example, data on railway land development projects completed by railway bureaus such as Jinan Railway Bureau, Taiyuan Railway Bureau, Xi'an Railway Bureau, and Guangzhou Railway Bureau between 2016 and 2024 can be collected.

[0039] Sub-step S1022: Preprocess the training sample set, including missing data imputation, outlier correction, normalization or standardization, and encoding conversion of categorical data, to ensure data quality and format consistency.

[0040] Sub-step S1023: After completing the preprocessing, set the parameters of the gradient boosting decision tree model, including at least the maximum depth of the decision tree, the learning rate, the subsample sampling ratio, and the regularization parameter.

[0041] Sub-step S1024: Input the preprocessed training sample set into the model, and generate an ensemble model composed of multiple decision trees through multiple rounds of iterative training. In each round of iteration, the model generates a new decision tree based on the difference between the current prediction result and the actual development potential score label to correct the prediction error, until the preset number of iterations is reached or the prediction error meets the preset conditions, and finally the pre-trained gradient boosting decision tree model is obtained.

[0042] In some embodiments, the gradient boosting decision tree model can be implemented using the XGBoost model. The XGBoost model introduces a regularization term during training to control model complexity, prevent overfitting, and ensure the model's generalization ability on new data. The regularization term effectively suppresses the influence of noisy data and redundant features by penalizing the number of leaf nodes and leaf weights, enabling the trained model to maintain prediction accuracy while possessing good stability and adaptability.

[0043] Through the above process, step S102 inputs the railway land feature vector into the pre-trained gradient boosting decision tree model, outputting a quantitative development potential score. This score comprehensively reflects the impact of various factors such as basic land attributes, location conditions, environmental constraints, and socio-economic factors on development potential, providing a scientific and objective input basis for subsequent interpretability analysis and business matching. Furthermore, due to the regularization mechanism introduced during model training, the output development potential score exhibits good stability and generalization ability, enabling it to adapt to the railway land assessment needs under different regional and market environments.

[0044] In step S103, an interpretability analysis is performed on the development potential score output by the gradient boosting decision tree model. The contribution value of each dimension of the railway land use feature vector to the score is calculated, including the direction and intensity of the contribution. The purpose of this step is to break down the black-box nature of machine learning models, decomposing abstract prediction results into understandable quantitative indicators. This allows decision-makers not only to know the final development potential score but also to clearly grasp the specific role of each feature factor in the score formation process, thereby enhancing the transparency and credibility of the evaluation results.

[0045] To achieve this goal, this step employs an interpretability analysis algorithm to interpret the model output.

[0046] In some embodiments, the algorithm employs the SHAP algorithm, whose core idea originates from the Shapley value in cooperative game theory. By considering all possible combinations of feature subsets, it fairly allocates the marginal contribution of each feature to the prediction result.

[0047] Specifically, for each railway land parcel to be evaluated, the SHAP algorithm iterates through all feature subsets that do not contain a certain feature, calculates the change in the model's predicted values ​​before and after adding that feature, and then performs a weighted average of the changes across all subsets to obtain the Shapley value for that feature. This value represents the contribution of that feature to the current land parcel's development potential score; its positive or negative sign indicates the direction of contribution, such as positive promotion or negative inhibition, and the absolute value indicates the intensity of the contribution.

[0048] In some embodiments, the SHAP value can be calculated as follows: ; in, Indicates the first Features The corresponding Shapley value; U represents the set of all features, i.e. P represents the total number of input features; S represents that U does not contain any features. any subset of; This represents the number of features in subset S; This indicates the predicted output value of the gradient boosting decision tree model when only a subset of features S is used; This indicates adding features to the feature subset S. The predicted output value of the post-gradient boosting decision tree model. This formula iterates through all features not included. The feature subset is used to calculate the change in the model's predicted value before and after adding the feature, and the Shapley value of the feature is obtained by weighted averaging, thereby quantifying the marginal contribution of each feature to the development potential score.

[0049] The above calculations yield the contribution value of each feature to the development potential score. These contribution values ​​are presented in numerical form, intuitively revealing which features are key drivers of development potential, such as favorable location conditions and well-developed infrastructure; and which features are limiting factors that restrict development potential, such as environmental pollution and remote location.

[0050] For example, for a particular plot of land, SHAP analysis may show that "distance to subway station" has a significant positive contribution, while "noise pollution" has a negative contribution, thus providing precise guidance for subsequent business matching and optimization decisions.

[0051] The contribution value output in step S103 is not only the core output of the model interpretation, but also a bridge connecting the prediction results and business decisions. This quantified contribution information will be used in the next step of business matching calculation to ensure that development recommendations are not only based on the final score, but also on in-depth factor analysis, thereby significantly improving the scientific nature and rationality of decision-making.

[0052] In step S104, based on the feature contribution values ​​of each dimension calculated in step S103, the matching degree between the railway land feature vector and multiple preset development types is further calculated, and a recommended list of development types is generated in descending order of matching degree. The purpose of this step is to transform the feature contribution information revealed by the interpretability analysis into specific development direction suggestions, achieving a key leap from potential assessment to development decision-making, and making the assessment results more practical and instructive.

[0053] The core logic of matching degree calculation is to use the contribution value and contribution direction of each dimension feature to determine the degree of suitability of the railway land with different development types.

[0054] In some embodiments, this calculation process first identifies key driving factors that have a significant positive impact on the development potential score and limiting factors that have a significant negative impact, based on the direction and intensity of contribution of each dimension feature. For example, if the "distance to subway station" feature shows a large positive contribution, it indicates that the site has excellent accessibility; if the "noise pollution" feature shows a large negative contribution, it indicates that environmental factors may become a constraint on development.

[0055] Subsequently, the identified key driving and limiting factors are matched according to pre-defined development type adaptation rules. These rules are built upon a domain knowledge base, storing the requirements or preference patterns of different development types for various features. For example, the rules might specify: if the population density-related features of a plot contribute significantly, it matches well with residential or commercial / service development types; if the accessibility-related features contribute significantly, it matches well with commercial or integrated hub development types; and if environmental limiting factors contribute significantly, it matches well with park / green space or logistics / warehousing development types. By comparing the feature contribution patterns of actual plots with the standard patterns of various development types, the matching degree between the plot and each development type is quantitatively calculated.

[0056] In some embodiments, the development type specifically includes one or more of commercial, residential, industrial, logistics, and park green space. These types cover the main directions of integrated development of railway land, taking into account both economically oriented commercial and residential development, and functional and ecologically oriented industrial, logistics, and green space development.

[0057] After calculating the matching degree, each development type is sorted from highest to lowest matching degree, forming a recommended development type list. This list not only presents the most suitable development direction but also demonstrates the relative advantages and disadvantages between different directions, providing decision-makers with multi-dimensional selection references. Through this step, the originally abstract contribution value is transformed into concrete and actionable development suggestions, significantly improving the practical value of the evaluation results and laying the content foundation for the generation of subsequent strategy reports.

[0058] In step S105, based on the development potential score output in step S102, the feature contribution values ​​of each dimension calculated in step S103, and the development type recommendation list generated in step S104, a comprehensive development strategy report for railway land is generated. The purpose of this step is to structurally integrate all the aforementioned assessment and analysis results and present them to decision-makers in a standardized and intuitive document format, providing a comprehensive and auditable decision-making basis for actual business operations such as asset disposal, planning approval, and cooperative development of railway land.

[0059] To achieve this goal, this step uses a report generation module to summarize and organize the various output data.

[0060] In some embodiments, the report generation module supports selective configuration of report chapters, allowing users to select the chapters to be included based on their actual needs. The report chapters should at least include a project overview, development potential scoring results, explanations of key influencing factors, development type recommendations and comparisons, data sources, and appendices. The project overview mainly presents basic information about the land parcel, such as parcel name, location description, land area, and land use. The development potential scoring results section displays the comprehensive development potential score of the land parcel and the specific scores for each sub-dimension, such as location conditions, supporting facilities, transportation conditions, environmental score, policy support, and socio-economic factors. The explanations of key influencing factors section present the contribution values ​​of each dimension obtained from the SHAP analysis in the form of visual charts or text descriptions, clearly identifying positive driving factors and negative limiting factors. The development type recommendations and comparisons section lists development types sorted by matching degree, and may include the matching degree value for each type and a brief reason for its suitability. The data sources and appendices section records information such as the multi-source data sources used in this assessment, data collection time, and model version to enhance the report's traceability and credibility.

[0061] During report generation, the system automatically extracts assessment data for the target plot from the database and formats the data according to a preset template. The generated report can be exported in various electronic document formats.

[0062] In some embodiments, PDF and Word formats are supported to facilitate users' viewing, distribution, or further editing in different scenarios.

[0063] Step S105 culminates in a complete and standardized report on the integrated development strategy for railway land. This report not only summarizes the quantitative results of the development potential score but also reveals the underlying factors contributing to the score through interpretability analysis. Furthermore, it provides specific development direction recommendations, enabling decision-makers to fully understand the basis for the assessment conclusions and thus make more scientific and prudent development decisions. This structured and configurable report generation mechanism significantly enhances the operability and promotional value of this method in practical engineering applications.

[0064] In some embodiments, the method also includes a dynamic maintenance mechanism for the model and data.

[0065] Specifically, based on dynamic updates of multi-source data, the feature vector of railway land use is periodically expanded and corrected to ensure that the feature data reflects the latest land use status and market environment changes. Simultaneously, the gradient boosting decision tree model is incrementally trained or retrained using newly added historical railway land use sample data, enabling the model to continuously learn patterns and regularities in new data, thereby maintaining the timeliness and accuracy of the prediction results. Through this dynamic maintenance mechanism, this method can adapt to the needs of railway land use development assessment in different periods and regions, avoiding assessment biases caused by outdated data or rigid models, further improving the practicality and reliability of the method.

[0066] On the other hand, the present invention also provides a railway land development potential assessment and decision-making system, including a processor, a memory, and a computer program or instructions stored in the memory, wherein the processor is used to execute the computer program or instructions, and when the computer program or instructions are executed, the system implements the steps of the method described in any of the above.

[0067] On the other hand, the present invention also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the method as described in any of the preceding claims.

[0068] The present invention will now be described with reference to a specific embodiment: This embodiment will take a specific railway land parcel as an example to elaborate in detail on the complete technical solution of the railway land development potential assessment and decision-making method proposed in this invention. To facilitate understanding of the overall technical concept and implementation details of this invention, the technical solution will first be fully described below, specifically including the establishment of a railway land development characteristic database, development potential scoring prediction based on the XGBoost model, the interpretation and evaluation application method of prediction results based on SHAP, and the architecture and functional modules of the comprehensive railway land development assessment system.

[0069] I. Establishment of a Database of Railway Land Development Characteristics 1) Multi-source data acquisition and standardization This method integrates multiple data sources, including authoritative sources such as the "Railway Land Valuation Report," Gaode Maps, the official website of the China Meteorological Administration, and the National Bureau of Statistics, to collect multi-dimensional information related to railway land development. These data sources include various development characteristic data for railway land, such as land area, plot ratio, land rights attributes, land use term, and infrastructure construction status. Locational characteristics, such as the distance between the land and railway stations, commercial centers, airports, and public transportation stations, as well as environmental characteristics such as air pollution and noise pollution levels, are also incorporated into the database to ensure a comprehensive assessment of development potential.

[0070] 2) Data preprocessing and feature vector construction After data collection is complete, the data is preprocessed and transformed into feature vectors that can be used by machine learning models. Specific processing methods include: Numerical data, such as land area, average annual temperature, and GDP, are directly used as numerical inputs.

[0071] Categorical data, such as allocation or authorized operation in land rights attributes, is transformed using 0-1 dummy variables; ordered categorical data, such as excellent, relatively excellent, and average, are processed using ordered numerical assignment methods; and infrastructure scores are assigned numerical values ​​based on the specific conditions of each plot.

[0072] Distance-based features, such as distances to railway stations, business centers, airports, and public transportation stations, are measured using Amap and used as standardized features.

[0073] These feature vectors ensure the uniformity and consistency of the data, providing suitable input for subsequent model training and evaluation.

[0074] 3) Dynamic updates and expansion of the feature database This solution also incorporates a dynamic update mechanism to ensure the feature database can be continuously expanded and updated over time and in response to changes in the market environment. The feature information for each plot is updated based on new data and timelines to guarantee the database's timeliness and accuracy. Furthermore, by utilizing automated data acquisition and processing technologies, plot information can be updated and supplemented in real time from external data sources, enhancing the database's flexibility and scalability.

[0075] 4) Systematic and standardized data storage and management All collected and preprocessed data will be uniformly stored in a structured database to ensure efficient and secure data storage. The feature database not only stores the development characteristics of each plot but also supports rapid retrieval and analysis of multiple plots through efficient data indexing, association, and query mechanisms. The database design supports rapid processing of large-scale land development assessments, providing real-time data support and accurate analysis for development decisions.

[0076] II. Railway Land Development Potential Scoring and Prediction Model Based on XGBoost Model This solution provides a railway land development potential scoring and prediction model based on the XGBoost model, which uses advanced machine learning techniques to score and predict the development potential of railway land. This model can effectively process multi-dimensional data, capture complex nonlinear relationships, and provide a scientific and quantitative assessment of development potential.

[0077] Unlike traditional linear regression models, XGBoost can handle complex nonlinear relationships. The development potential of railway land is typically influenced by a combination of factors, and these factors may have complex nonlinear interactions. For example, the land area and plot ratio may have different effects on development potential within different ranges, and XGBoost can flexibly model these nonlinear effects to provide a more accurate development potential score for each plot.

[0078] Therefore, in this scheme, the XGBoost model is optimized to meet the needs of railway land development assessment, including specific adjustments for basic land attributes, location characteristics, environmental characteristics, and socio-economic characteristics. XGBoost's advantage lies in its ability to handle multiple data types and operate efficiently on large-scale datasets.

[0079] 1) Improved accuracy of tree generation and railway land development potential scoring In XGBoost, new decision trees are added iteratively to improve the estimation accuracy of railway land development potential. The sum of the predictions from all decision trees ultimately yields the predicted output of the railway land development potential score. The formulas for tree generation and accuracy improvement in railway development potential scoring are as follows: ; in, Indicates the first After the nth iteration Predicted development potential score for each railway land parcel; Indicates the first After the nth iteration Predicted development potential score for each railway land parcel; Indicates the first decision trees for the first Predicted output values ​​for each plot; Indicates the first The railway land feature vector of each plot.

[0080] This formula shows that the new decision tree generated in each iteration is used to correct the error of the previous prediction result, and gradually approaches the true value through multiple iterations.

[0081] 2) Regularization learning objectives In predicting the development potential of railway land, the dataset may contain a lot of noisy data or highly correlated features, which can easily lead to model overfitting. In the XGBoost model, a regularization term is introduced to control model complexity and prevent overfitting. The objective function for the k-th iteration is calculated as follows: ; in, This represents the total number of railway land parcels sampled during the training. Here is the loss function, used to measure the loss of the first... Actual development potential rating tags for individual plots With the Predicted value after the second iteration To address the differences between them, the mean square error function can be used in this solution; For the first After the nth iteration, the model... Predicted development potential score for each plot of land; This is a regularization term used to control the... The complexity of a decision tree.

[0082] Regularization term The specific calculation formula is as follows: ; in, Indicates the first The number of leaf nodes in each decision tree; For the first The weights of the leaf nodes; This is a regularization parameter used to penalize the number of leaf nodes and control the structural complexity of the tree; This is a regularization parameter used to penalize the L2 norm of leaf node weights, preventing excessively large weights.

[0083] By introducing a regularization term, the model can maintain low complexity while fitting the training data, effectively suppressing overfitting and improving its generalization ability on new data.

[0084] 3) Determine the optimal tree structure Identifying the optimal tree structure in XGBoost is crucial for obtaining the best model prediction results for railway land development potential scoring. Incremental training is employed to continuously optimize the objective until it is minimized. The loss function is Taylor-expanded to determine the specific optimization parameters. The optimized XGBoost objective function is calculated as follows: ; in, Indicates the first The objective function value of the XGBoost model at the next iteration; This represents the total number of railway land parcels sampled during the training. Represents the loss function, used to measure the loss of the first... Actual development potential rating tags for individual plots With the Predicted value after the second iteration The differences between them; Represents the loss function At the current forecast value The first derivative at that point; Indicates the first decision trees for the first Predicted output values ​​for each plot; Indicates the first The railway land use feature vector of each plot; Represents the loss function At the current forecast value The second derivative at point; Indicates the first The regularization term of the decision tree is used to control the complexity of the tree.

[0085] 4) Model Interpretation The importance of explanatory variables is the most important explanatory tool in tree models. A higher value indicates that the factor is used more frequently in constructing the decision tree, and thus its greater importance. Based on the final training results of the model, the relative importance of each explanatory variable is calculated as follows: The final formula for calculating feature importance is: ; in, Indicates the first Features Global importance in the entire XGBoost model; This represents the total number of decision trees in the model; Indicates the first Features In the Decision Tree Local importance in.

[0086] No. The formula for calculating the feature importance in a decision tree is: ; in, Indicates the first Features In the Local importance in a decision tree; Indicates the first The total number of leaf nodes in the decision tree; This represents the index of an internal node in the decision tree, with values ​​ranging from 0 to... , represents the split node of all non-leaf nodes; Represents a node The reduction in mean square error after splitting; For indicator functions, when node Splitting variables Equal to characteristics The value is 1 if the condition is met, and 0 otherwise.

[0087] This formula applies to all usage features. The node gains used as splitting variables are accumulated to obtain the importance of the feature in a single tree.

[0088] 5) Model Training In this scheme, the training process of the railway land development potential scoring prediction model based on the XGBoost model is based on historical railway land sample data. The historical railway land sample data comes from relevant land development projects from January 2016 to December 2024 provided by railway bureaus such as Jinan Railway Bureau, Taiyuan Railway Bureau, Xi'an Railway Bureau, and Guangzhou Railway Bureau. Each railway land sample corresponds to a railway land parcel that has completed development potential assessment or has a clear development potential evaluation result.

[0089] During model training, a railway land dataset is first constructed for training. Each sample in the dataset includes input data characterizing the development conditions of the railway land and a corresponding development potential score as output data. The input data consists of multi-dimensional feature data, including at least one or more of the following: basic land attribute data, location-related data, environmental constraint-related data, and socio-economic-related data, to comprehensively reflect the development conditions and environment of the railway land. The development potential score serves as the label data for model training, representing the development potential level of the railway land under a given evaluation system.

[0090] After constructing the training dataset, it is preprocessed to improve the stability and accuracy of model training. Preprocessing includes imputing or removing missing data, handling outliers, standardizing or normalizing data of different scales, and encoding categorical data to make various input data suitable for the training requirements of the XGBoost model.

[0091] After data preprocessing, the processed railway land sample data is input into the XGBoost model as training data. During model training, the parameters of the XGBoost model are set according to the characteristics of the railway land development potential scoring and prediction task, including but not limited to the maximum depth of the decision tree, learning rate, subsample sampling ratio, and regularization parameters, in order to control the complexity of the model and improve its generalization ability.

[0092] During model training, the XGBoost model gradually constructs an ensemble model composed of multiple decision trees through multiple rounds of iterative training. In each round of training, the model generates a new decision tree based on the difference between the current prediction result and the actual development potential score to correct the prediction error, thereby continuously optimizing the prediction results of the railway land development potential score. In this way, the model can gradually learn the nonlinear mapping relationship between the multidimensional features of railway land and the development potential score.

[0093] When the model training process reaches the preset number of iterations or the model prediction error meets the preset conditions, the model training is completed, forming the final railway land development potential scoring and prediction model. The trained model can receive multi-dimensional feature data of the railway land to be evaluated as input and output the corresponding railway land development potential score, which is used to assist in railway land development decision-making and planning assessment.

[0094] Through the above training process, this scheme can automatically construct a railway land development potential scoring and prediction model using historical railway land sample data, realize the quantitative assessment of railway land development potential, improve the objectivity and consistency of assessment results, and has good feasibility and promotional application value.

[0095] III. Interpretation and Evaluation Methods of Railway Land Development Potential Prediction Results Based on SHAP This paper proposes a method for interpreting and evaluating railway land development potential prediction results based on Shapley Additive Explanations (SHAP), addressing the problems of uninterpretable models and ineffective application of prediction results to development decisions in existing railway land development assessments. This method utilizes the SHAP algorithm to analyze the interpretability of railway land development potential scores predicted by the XGBoost model. By quantifying the local and overall contributions of different development influencing factors to the prediction results, it achieves transparent, structured, and engineering-oriented application of the development potential prediction results. The SHAP value calculation formula is as follows: ; in, Indicates the first Features The corresponding Shapley value; U represents the set of all features, i.e. P represents the total number of input features; S represents that U does not contain any features. any subset of; This represents the number of features in subset S; This indicates the predicted output value of the gradient boosting decision tree model when only a subset of features S is used; This indicates adding features to the feature subset S. The predicted output value of the post-gradient boosting decision tree model.

[0096] This formula iterates through all features that are not included. The feature subset is used to calculate the change in the model's predicted value before and after adding the feature, and the Shapley value of the feature is obtained by weighted averaging, thereby quantifying the marginal contribution of each feature to the development potential score.

[0097] 1) Full interpretation and contribution quantification of prediction results This scheme utilizes the SHAP algorithm to provide a fine-grained interpretation of the development potential score for each railway land parcel, calculating the Shapley value of each input feature that constitutes the prediction result, thereby quantifying the positive or negative contribution of each feature to the prediction result. Compared with traditional feature importance analysis, which only provides an overall contribution, this method can generate a detailed local interpretation down to the parcel level, enabling decision-makers not only to know which factors are important, but also to clarify the direction and intensity of each feature's effect on a specific parcel.

[0098] 2) Identify the key drivers and inhibitory factors of railway land development. By ranking the contributions of SHAP outputs, this scheme can automatically identify the key drivers promoting the development potential of railway land and the main inhibitory factors limiting development potential. This capability is particularly important for railway land development, as it can be used to determine the root causes of insufficient development potential for a particular plot, such as poor accessibility, limited land area, low socioeconomic level, and severe environmental pollution, and to provide a basis for subsequent optimization measures.

[0099] 3) Provide an explanatory basis for matching development business formats. This solution combines the interpreted output of SHAP with a development format recommendation system, so that format recommendations no longer rely solely on predictive scores, but are based on interpretable factors. For example, if SHAP shows that the population density of a plot contributes extremely high value, commercial and residential developments can be recommended; if SHAP shows that transportation accessibility contributes significantly, commercial and integrated hub developments are suitable; if SHAP shows strong environmental constraints, parkland or logistics warehousing formats can be recommended.

[0100] This decision-making mechanism based on scoring and interpretation enhances the rationality and transparency of business matching, and makes development recommendations auditable and technically sound.

[0101] 4) Comprehensive analysis of policy impacts and socioeconomic factors The SHAP (Self-Assessment and Evaluation) output of this scheme can effectively assess the impact of external factors such as policies and socio-economic factors on the development potential of railway land. For example, the economic development level of the surrounding area, policy changes, and environmental protection requirements will all reflect their specific impact in the SHAP analysis. Through these analyses, decision-makers can not only obtain accurate development potential scores, but also adjust development strategies according to changes in the external environment, optimize land resource allocation, and provide data support for development strategy reports.

[0102] 5) Structured output and engineering applications This solution standardizes the interpretation results of SHAP into outputs including: a development potential contribution report at the land parcel level; an overall influencing factor analysis at the regional level; interpretable supporting documentation that can be used for expert review and planning approval; and structured development potential data that can be used for asset valuation and operational management. Through the structured output described above, this solution transforms the model interpretation into a practical engineering tool that can be directly used in railway enterprise planning, development, investment, approval, and other processes, significantly improving the feasibility of intelligent evaluation methods.

[0103] IV. Comprehensive Evaluation System for Railway Land Use Development like Figure 2 As shown, the railway land comprehensive development assessment system of this scheme adopts a bottom-up layered architecture. The bottom layer of the system is the data acquisition layer, which includes modules for acquiring basic land information, surrounding facilities, transportation, environmental data, economic data, and policy data, used to obtain various multi-source data on railway land. The raw data collected is cleaned, standardized, and feature-constructed by the data cleaning, data standardization, and feature construction modules of the data preprocessing layer before being transmitted to the model service layer.

[0104] The model service layer includes an XGBoost development potential prediction module and a result interpretation and analysis module. The former predicts the development potential of railway land based on feature data, while the latter performs interpretability analysis on the prediction results. The development potential score and interpretation analysis results output by the model service layer are passed to the business application layer, where the development business type recommendation module and the development strategy report generation module respectively perform business type matching and strategy report generation.

[0105] The top layer of the system is the user interaction layer, which includes a spatial map display module, an assessment process control module, and a result display and report export module. It is used to provide users with spatial map visualization, assessment process operation, and display and export functions of assessment results and reports, thereby realizing a complete business closed loop for the comprehensive development assessment of railway land.

[0106] The system homepage uses a spatial map as its core display interface, integrating and presenting the spatial distribution of railway land, assessment task status, current land use, and development potential scoring results to achieve a holistic understanding and quick overview of railway land resources. The system supports loading the plot layer to be assessed into the spatial map, and by selecting the target plot, triggering the corresponding comprehensive development assessment process for railway land, it enters the development potential assessment and comprehensive development process for the plot to be assessed.

[0107] Basic Land Parcel Information Input: The system inputs basic land parcel information, including parcel name, location description, land certificate number, land use nature, land area, planned plot ratio, land use, city or administrative district, city level, and regional attributes. Based on the spatial data of the parcel, the system automatically obtains and scores surrounding supporting information, including information such as convenience stores within 500 meters, restaurants within 1 kilometer, large commercial complexes within 1 kilometer, distance to the nearest commercial complex, kindergartens within 1 kilometer, primary and secondary schools within 1 kilometer, distance to the nearest school, tertiary hospitals, community hospitals, basic outpatient clinics, distance to the nearest medical point, parks within 1 kilometer, green space coverage level, and distance to the nearest park. Transportation information includes the distance to the nearest subway station, bus station, high-speed rail station, and airport. Local economic and policy information includes population, GDP, per capita disposable income, overall or control plan name, industrial orientation or functional positioning, and policy-driven business types. After completing the collection, calculation, and verification of the basic land parcel information, the system saves the information and uses it as input parameters in the railway land development potential scoring process.

[0108] Development Potential Scoring and Spatial Result Display: This section presents the assessment results and spatial information of the target plot. Based on the entered basic information of the plot, the railway land development potential scoring module based on the XGBoost model is invoked to assess the development potential of the target plot. The assessment includes scoring multiple dimensions such as location conditions, surrounding supporting facilities conditions, transportation conditions, policy support conditions, and socio-economic conditions. The scores of different dimensions are weighted according to the relative importance of each influencing variable in the model to obtain the comprehensive development potential score of the target plot. Example data is shown in Table 1. Simultaneously, this section combines the spatial information of the target plot for spatial display, presenting the land use scope of the plot on a map and overlaying the spatial distribution information of surrounding supporting facilities. Supporting facilities include at least points of interest such as schools, commercial facilities, high-speed rail stations, and medical institutions to support intuitive analysis and verification of the development potential scoring results.

[0109] Table 1. Example data for development potential scoring. Results Interpretation and Analysis: The system invokes the results interpretation and analysis module to process the development potential score predicted by the XGBoost model. This includes visualizing the global feature importance, analyzing the correlation between features, displaying the weight of each feature in the development potential score, and visualizing the feedback of whether a feature has a positive or negative impact on the development potential score. Simultaneously, based on the interpretation and analysis results, the system generates a diagram illustrating the nonlinear relationship between a single influencing factor and the railway land development potential, revealing the changing trends of the impact of different features on development potential across different value ranges. Furthermore, based on the interpretability analysis results, the system identifies and displays key factors that have a major positive impact on development potential and limiting factors that have a major negative impact. After confirming the interpretation and analysis results, users can trigger the railway land development business type recommendation process through the system.

[0110] Development Format Recommendation and Strategy Generation: After interpreting and analyzing the results, the system, based on the target plot's basic information, development potential score, and influencing factor analysis, calls the development format recommendation module to match and rank various preset development formats. By comparing the plot's characteristic parameters with the adaptation rules of different development formats, a matching degree ranking for each development format is generated, and the recommended list is displayed in descending order of matching degree. Users can trigger the generation process of the comprehensive development strategy report for railway land after confirming or selecting the corresponding format.

[0111] Development Strategy Report Generation and Export: After recommending development formats, the system invokes the development strategy report generation module to structure the comprehensive development assessment results for railway land. The system provides a function for selecting and configuring report chapters, including project overview, development potential scoring results, explanations of key influencing factors, recommended and compared formats, data sources, and appendices. Based on the user-selected chapters, the system automatically summarizes and formats the corresponding assessment data, analysis results, and recommendation conclusions, generating a comprehensive development strategy report for railway land. The generated report can be exported as an electronic document, supporting PDF and Word formats.

[0112] In summary, the railway land development potential assessment and decision-making method and system provided by this invention acquires multi-source railway land data and performs standardized processing to construct feature vectors; inputs the feature vectors into a pre-trained gradient boosting decision tree model to output a development potential score; analyzes the score using an interpretability analysis algorithm to calculate the contribution value of each dimension; calculates the matching degree with multiple preset development types based on the contribution value and generates a recommendation list in descending order; and generates a comprehensive development strategy report based on the score, contribution value, and recommendation list. This invention, through the organic combination of gradient boosting decision tree and SHAP interpretability analysis, achieves intelligent processing throughout the entire process from data fusion and nonlinear modeling to interpretable decision-making, significantly improving the accuracy, transparency, and scientific rigor of railway land development assessment.

[0113] Furthermore, by standardizing and storing multi-source data on the blockchain, this invention ensures the credibility and traceability of input data, providing a high-quality data foundation for subsequent evaluation.

[0114] Those skilled in the art will understand that the exemplary components, systems, and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software, or a combination of both. Whether implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention. When implemented in hardware, it can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this invention are programs or code segments used to perform the desired tasks. The programs or code segments can be stored in a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried in a carrier wave.

[0115] It should be clarified that the present invention is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of the present invention.

[0116] In this invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with or in place of features of other embodiments.

[0117] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations of the embodiments of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for assessing and deciding on the development potential of railway land, characterized in that, The method includes the following steps: Acquire multi-source data on railway land use, including basic land attribute data, location feature data, environmental feature data, and socio-economic feature data, and standardize the multi-source data to construct a railway land use feature vector. The railway land feature vector is input into a pre-trained gradient boosting decision tree model, which outputs the development potential score corresponding to the railway land. The gradient boosting decision tree model is trained based on historical railway land sample data and its corresponding development potential score labels. An interpretability analysis algorithm performs interpretability analysis on the development potential score output by the gradient boosting decision tree model, and calculates the contribution value of each dimension feature in the railway land feature vector to the development potential score. The contribution value includes the contribution direction and contribution intensity. Based on the contribution values ​​of each dimension of the railway land feature vector, the matching degree between the railway land feature vector and multiple preset development types is calculated; a recommended list of development types is generated in descending order of matching degree; the development types include one or more of commercial, residential, industrial, logistics, and park green space. Based on the development potential score, the contribution value, and the recommended list of development types, a comprehensive development strategy report for railway land is generated.

2. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The pre-training steps of the gradient boosting decision tree model include: A training sample set is constructed, which includes multiple historical railway land use sample data. Each historical railway land use sample data contains the multi-source data as input features and the development potential score label corresponding to the multi-source data. The training sample set is preprocessed, including missing data imputation, outlier correction, normalization or standardization, and encoding conversion of categorical data. Set the model parameters of the gradient boosting decision tree model, wherein the model parameters include at least the maximum depth of the decision tree, the learning rate, the subsample sampling ratio, and the regularization parameter; The preprocessed training sample set is input into the gradient boosting decision tree model. Through multiple rounds of iterative training, an integrated model consisting of multiple decision trees is generated. In each round of iteration, a new decision tree is generated based on the difference between the current prediction result and the actual development potential score label to correct the prediction error, until the preset number of iterations is reached or the prediction error meets the preset conditions, thus obtaining the pre-trained gradient boosting decision tree model.

3. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The multi-source data is collected from multiple data sources, including railway land valuation reports, geographic information service platforms, meteorological data platforms, and statistical data platforms. The collected multi-source data is cleaned and normalized to construct data entries in a unified format. The processed multi-source data is stored in a structured database according to a preset data structure, and a unique identifier and a trusted timestamp are generated for each data entry. Based on the unique identifier and the trusted timestamp, the hash value of the data entry is uploaded to the blockchain network for evidence storage, so as to achieve traceability and immutability of the data source; The structured database supports dynamic updates and version management of the multi-source data, and sets corresponding data access permissions for different user roles; The location feature data includes distances to railway stations, commercial centers, airports, and public transportation stations. These data are acquired through the spatial distance measurement function of the geographic information service platform, and the measurement results are stored as standardized feature values ​​in the structured database.

4. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The multi-source data is standardized, including: For numerical data, normalization or standardization methods are used to unify the dimensions. For categorical data, one-hot encoding or dummy variable transformation is used to convert it into a numerical vector. For distance-type data, spatial distance is measured through a geographic information system or map application interface, and the measurement results are used as standardized feature values. Missing value detection and imputation, as well as outlier identification and correction, are performed on the multi-source data.

5. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The gradient boosting decision tree model is an XGBoost model; the XGBoost model introduces a regularization term during training to prevent overfitting.

6. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The interpretability analysis algorithm is the SHAP algorithm; the contribution value is the Shapley value, and the formula for calculating the contribution value of each dimension feature to the development potential score is: ； in, Indicates the first Features The corresponding Shapley value; U represents the set of all the stated features, i.e. P represents the total number of input features; S represents that U does not contain any features. any subset of; This represents the number of features in subset S; This represents the predicted output value of the gradient boosting decision tree model when only a subset of features S is used; This indicates that the feature is added to the feature subset S. The predicted output value of the gradient boosting decision tree model described later.

7. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The step of calculating the matching degree for multiple preset development types based on the contribution value and the railway land use feature vector includes: Based on the contribution direction and intensity of each dimension of features, identify the key driving factors that have a positive impact on the development potential score and the limiting factors that have a negative impact. Based on the key driving factors and limiting factors, the matching degree between each development type and the railway land is calculated according to the preset development type adaptation rules, which are constructed based on the domain knowledge base; A list of recommended development types is generated in descending order of matching degree.

8. The method for assessing and deciding on the development potential of railway land according to claim 1, characterized in that, The method further includes: Based on the dynamic updates of the multi-source data, the railway land use feature vector is periodically expanded and corrected, and the gradient boosting decision tree model is incrementally trained or retrained using newly added historical railway land use sample data to maintain the timeliness and accuracy of the model.

9. A railway land development potential assessment and decision-making system, comprising a processor, a memory, and a computer program or instructions stored in the memory, characterized in that, The processor is configured to execute the computer program or instructions, and when the computer program or instructions are executed, the system implements the steps of the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1 to 8.

Images