A campus surrounding vitality anomaly identification and strategy recommendation method and system based on multi-source heterogeneous spatial data fusion and a graph model

Abstract

A campus surrounding vitality anomaly recognition and strategy recommendation method and system based on multi-source heterogeneous spatial data fusion and graph model belong to the technical field of spatial information processing, geographic information system and spatio-temporal big data analysis. In order to solve the problems of difficult unification of multi-source heterogeneous spatial data, difficult cross-scale connection of spatial objects, reliance on experience for vitality anomaly, difficult conversion of interaction relationship into diagnostic rules and no closed loop of strategy recommendation, the method comprises the following steps: acquiring multi-source heterogeneous spatial data to generate a reliable data set; establishing a cross-scale spatial object code, constructing a spatial graph model and calculating graph topological features; calculating morphological features, accessibility features, interface features and industry features, and constructing a morphological-vitality feature matrix; identifying abnormal types based on the difference between predicted vitality values and observed vitality values; determining diagnostic problems, generating strategy trigger conditions, matching strategies and calculating strategy priority, and outputting a diagnostic report. The method is used for automatic diagnosis, abnormal type identification and hierarchical strategy recommendation of campus surrounding spatial vitality.

Description

Technical Field

[0001] This invention relates to a method and system for identifying and recommending strategies for anomalies in campus surroundings based on multi-source heterogeneous spatial data fusion and graph models, belonging to the fields of spatial information processing, geographic information systems, and spatiotemporal big data analysis. Background Technology

[0002] Against the backdrop of the deepening of territorial spatial planning and urban renewal initiatives, and the gradual implementation of policies promoting openness and sharing among universities, urban spatial renewal, campus openness and sharing, and the optimization of surrounding service facilities have become core issues for improving the refinement of urban governance and the efficiency of spatial resource allocation. In the process of urban spatial renewal, campus openness and sharing, and the optimization of surrounding service facilities, it is often necessary to determine whether there are problems such as insufficient spatial vitality, mismatched service facilities, inadequate transportation connections, or fragmented interfaces in the areas surrounding campuses. Traditional methods often rely on manual surveys, experience-based judgments, or single-layer statistical analysis, making it difficult to generate reproducible and updatable diagnostic results across different campuses, different spatial scales, and different time periods.

[0003] Existing spatial analysis systems can typically perform POI distribution statistics, heat map display, road network accessibility calculation, or measurement of single spatial morphological indicators. For example, existing technologies integrate multi-source geographic big data to construct comprehensive indicators and machine learning models for urban vitality evaluation; other technologies utilize multi-source urban geographic data to obtain evaluation indicators of urban street vitality and conduct vitality evaluation through multi-objective optimized spatial projection tracing methods; still other technologies collect multi-source heterogeneous data and align them spatiotemporally according to preset geographic grid units and time granularity to generate regional spatiotemporal feature tensors for regional economic vitality monitoring and assessment. However, the above-mentioned existing technologies still have the following technical shortcomings:

[0004] First, there is a lack of data heterogeneity and cross-scale mapping mechanisms. Data from different sources have inconsistent data structures, coordinate references, temporal granularities, and spatial units, making it difficult to automatically link them to a unified spatial object. POI point data, thermal gratings, road network vectors, and boundary surface data lack a unified spatiotemporal indexing and quality scoring mechanism; noise, duplicate points, coordinate offsets, and time lags reduce the reliability of calculation results. The lack of a unified coding and cross-scale mapping mechanism between regional, campus boundary, and plot scales makes it difficult to transfer diagnostic results and perform cross-scale backtracking.

[0005] Second, there is a lack of an automatic identification mechanism for the discrepancy between "theoretical vitality potential" and "actual observed vitality." Existing systems mostly directly display the level of vitality, only identifying the "amount" of vitality but not the "type" of discrepancy between potential and reality. High vitality may stem from exceeding morphological carrying capacity, while low vitality may stem from interface barriers or service mismatch—these two situations require completely different intervention strategies, but current technology cannot automatically distinguish between them.

[0006] Third, the topological relationships and interaction relationships of spatial maps have not been fully explored. Existing systems mostly treat each spatial indicator as an independent variable for statistical analysis, lacking an automated technical path that links spatial map topological relationships, indicator interaction enhancement relationships, and strategy recommendation rules. The synergistic or conflicting relationships between spatial elements—such as the superimposed effect of road accessibility and interface openness, and the coupling mechanism between POI supply and pedestrian flow heat—are difficult to transform into automated diagnostic rules.

[0007] Fourth, the confidence and traceability of diagnostic results are insufficient. Existing systems lack a comprehensive evaluation mechanism for data quality, model error, and sample coverage. Diagnostic results lack quantifiable confidence metrics, making it difficult to assess the reliability of the output. Furthermore, diagnostic results cannot be traced back to specific data sources, spatial locations, and triggering indicators, which is detrimental to subsequent planning decisions and effectiveness evaluation.

[0008] Fifth, there is a lack of strategy recommendation and closed-loop update mechanisms. Existing technologies are mostly one-off analyses, unable to automatically map abnormal diagnostic results into a list of actionable strategies, and lack feedback mechanisms for the effects of strategy implementation and model iteration updates, making it difficult to form a closed-loop process of "diagnosis-strategy-feedback-optimization".

[0009] Therefore, it is necessary to propose a multi-source heterogeneous data fusion diagnostic method and system for the space surrounding the campus, focusing on solving technical problems such as automatic access of multi-source spatial data, cross-scale object encoding, spatial graph topology modeling, morphological-vitality feature fusion, residual abnormality identification, interaction enhancement triggering, strategy library matching and dynamic updating. Summary of the Invention

[0010] The purpose of this invention is to address the problems in existing technologies, such as the difficulty in uniformly processing multi-source heterogeneous spatial data, the difficulty in cross-scale association of spatial objects around campuses, the reliance on manual experience to judge vitality anomalies, the difficulty in converting spatial element interaction relationships into automatic diagnostic rules, and the lack of a closed-loop update mechanism for strategy recommendation. This invention provides a method and system for identifying vitality anomalies and recommending strategies around campuses based on the fusion of multi-source heterogeneous spatial data and graph models.

[0011] The present invention discloses a method for identifying and recommending strategies for anomalies in campus surroundings based on multi-source heterogeneous spatial data fusion and graph models, comprising:

[0012] Step 1: Obtain multi-source heterogeneous spatial data through multi-source data interfaces, perform preprocessing and quality evaluation, and generate a reliable dataset;

[0013] Step 2: Based on the trusted dataset, establish cross-scale spatial object encoding, construct a spatial graph model, and calculate graph topological features;

[0014] Step 3: Based on the cross-scale spatial object encoding and the graph topology features, calculate the morphological features, accessibility features, interface features and business features of each spatial object unit, and construct a morphological-vitality feature matrix;

[0015] Step 4: Calculate the predicted vitality value based on the morphology-vitality feature matrix, and calculate the observed vitality value based on the observed vitality characterization data in the trusted dataset. Identify the anomaly type based on the difference between the predicted vitality value and the observed vitality value.

[0016] Step 5: Based on the interaction enhancement relationship between the indicators in each column of the morphology-vitality feature matrix, determine the diagnostic problem corresponding to the abnormality type described in Step 4, generate strategy triggering conditions, match strategies from the strategy library according to the strategy triggering conditions, and calculate the strategy priority.

[0017] Step 6: Output a diagnostic report, which includes a list of diagnostic issues, a list of strategies, and a priority ranking result.

[0018] Preferably, the multi-source heterogeneous data in step 1 includes point data, line data, area data, and raster data; the point data includes POIs, bus stops, and facility points; the line data includes road centerlines, walking paths, and bus routes; the area data includes campus boundaries, plot boundaries, building outlines, and green space areas; and the raster data includes heat maps, passenger flow raster, or nighttime activity raster.

[0019] The preprocessing of multi-source heterogeneous data includes: unified coordinate projection, time slice alignment, duplicate data removal, outlier handling, and spatial topology correction.

[0020] The quality assessment of multi-source heterogeneous data includes:

[0021] Calculate data quality score : ;

[0022] in, Indicates the coordinate offset score. Indicates the time-slice matching score. Indicates the confidence score of the classification. This indicates that you are resetting your credit score. Indicates the topological integrity score. , , , , Indicates configurable weights;

[0023] For data that is below the preset quality threshold, automatic rejection, manual verification and marking, or interpolation correction are performed.

[0024] Preferably, step 2, establishing cross-scale spatial object encoding, specifically includes:

[0025] Generate cross-scale codes;

[0026] The multi-source heterogeneous spatial data are linked to the same spatial object through a spatial connection method;

[0027] Establish R-tree spatial indexes and TimeCode temporal indexes to enable cross-scale data querying and result backtracking;

[0028] Specifically, the generation of cross-scale codes involves:

[0029] ;

[0030] in, Indicates the scale type, including regional scale, campus boundary scale, and plot scale; Indicates the campus object number; Indicates the specific spatial unit number; Indicates the time slice number;

[0031] The spatial connection methods include:

[0032] For point data, the system attaches it to spatial objects through point-surface overlay and proximity matching;

[0033] For line data, the system achieves connection through line-plane intersection, node extraction, and graph topology reconstruction;

[0034] For raster data, the system achieves the linking through raster clipping, resampling, and regional statistics.

[0035] Preferably, step 2, constructing the spatial graph model, specifically includes:

[0036] Campus boundary openings, road intersections, bus stops, plot entrances, and POI clusters are abstracted as graph nodes. The relationships between road segments, pedestrian walkways, and boundary accessibility and service reachability are abstracted as graph edges. Boundary weights are determined based on walking distance, road grade, boundary permeability, slope resistance, pedestrian crossing resistance, and traffic service intensity. ;

[0037] Constructing a spatial graph model : ;

[0038] The graph topology features include node connectivity, average shortest path, reachability time, betweenness centrality, service coverage strength, and boundary penetration strength.

[0039] Preferably, the morphological features described in step 3 include building coverage, green space ratio, planar compactness, planar regularity, and boundary permeability index, wherein the boundary permeability index is calculated jointly by the number of accessible entrances and exits, boundary length, and road connection strength; the accessibility features include road density, integration, selectivity, connectivity, and public transport station service intensity; the interface features include boundary openness and the number of accessible connections; and the business format features include the number of POIs, POI core density, business format hybrid entropy, and the proportion of production, living, and public service POIs.

[0040] Preferably, the predicted vitality value in step 4 is calculated using a weighted linear model, a random forest model, a gradient boosting tree model, a support vector regression model, a graph convolutional network, a graph attention network, or a combined model based on explanatory power weights; the model is stratified by campus, spatial scale, and time slice during training, and the model error is obtained through cross-validation;

[0041] The observed vitality value is calculated by weighting POI kernel density, thermal value, or passenger flow value.

[0042] Preferably, the specific method for identifying anomaly types based on the difference between predicted vitality values ​​and observed vitality values ​​in step 4 includes:

[0043] Calculate the standardized residuals:

[0044] ;

[0045] in, Represents the standardized residual. Indicates the predicted vitality value. Indicates the observed vitality value. Represents a spatial object unit. Indicates a time slice, This represents the standard deviation of residuals for samples of the same scale. This represents a very small constant to prevent the denominator from being zero;

[0046] according to value, and Based on the threshold relationship, the anomaly type is determined according to the following rules:

[0047] when , , When it is determined to be a high-potential-low-vitality type, it means that the spatial morphological conditions are met but the actual vitality is insufficient;

[0048] when , , When it is determined to be of the low potential-high vitality type, it means that the actual vitality is higher than the form's carrying capacity;

[0049] when When it is determined to be of a balanced development type, it means that the predictive potential and the observed vitality are basically consistent;

[0050] when and When it is determined to be of low potential-low vitality type, it means that both the spatial form and actual vitality are insufficient;

[0051] When a single indicator deviates significantly and the interaction relationship weakens, it is judged as a structural mismatch type, indicating that there is a lack of synergy between morphological elements.

[0052] in, This indicates a high threshold for predicted vitality. This indicates a low threshold for predicted vitality. It represents the standard deviation of residuals for samples of the same scale.

[0053] Preferably, the method for determining the interaction enhancement relationship between the indicators in each column of the morphology-vitality feature matrix described in step 5 includes:

[0054] Explanatory power based on single factor , With interactive explanatory power The comparison results determine the interaction type;

[0055] when When this occurs, it is determined to be nonlinear enhancement, indicating that the combined effect of the two indicators significantly amplifies the effect, triggering a combined strategy;

[0056] when When the condition is met, it is determined to be a two-factor enhancement, indicating that the two indicators have a synergistic effect and are included in the strategy matching.

[0057] when When the time is right, it is determined to be dominated by a single factor, indicating that the strategy corresponding to the dominant indicator is the main one;

[0058] when When this occurs, it is determined to be a weakened interaction, indicating that there is a spatial mismatch or element conflict, triggering a structural mismatch diagnosis.

[0059] Preferably, the specific method for calculating the strategy priority in step 5 includes:

[0060] ;

[0061] in, Indicates the severity of the residual. This indicates the explanatory power of the dominant indicator. Indicates the interaction enhancement coefficient. Indicates spatial scale weights, This represents the strategy fit coefficient. This represents the implementation cost coefficient.

[0062] , , , , , Indicates configurable weights;

[0063] according to The strategy is divided into four levels: Level 1, Level 2, Level 3, and Level 4, and output accordingly.

[0064] when At this time, it is a level one strategy;

[0065] when At that time, it is a secondary strategy;

[0066] when At that time, it is a three-level strategy;

[0067] when At that time, it was a level four strategy.

[0068] The present invention discloses a campus perimeter activity anomaly identification and strategy recommendation system based on multi-source heterogeneous spatial data fusion and graph model, comprising:

[0069] The data acquisition and preprocessing unit is used to acquire multi-source heterogeneous spatial data through multi-source data interfaces, perform preprocessing and quality evaluation, and generate a reliable dataset.

[0070] The spatial coding and graph modeling unit is used to establish cross-scale spatial object coding based on the trusted dataset, and to construct a spatial graph model and calculate graph topological features;

[0071] The feature matrix construction unit is used to calculate the morphological features, accessibility features, interface features, and business features of each spatial object unit based on the cross-scale spatial object encoding and the graph topology features, and to construct a morphological-vitality feature matrix.

[0072] The vitality prediction and anomaly identification unit is used to calculate the predicted vitality value based on the morphology-vitality feature matrix, and to calculate the observed vitality value based on the observed vitality characterization data in the trusted dataset, and to identify the anomaly type based on the difference between the predicted vitality value and the observed vitality value.

[0073] The strategy matching and recommendation unit is used to determine the diagnostic problem corresponding to the anomaly type based on the interaction enhancement relationship between the indicators in each column of the morphology-vitality feature matrix, generate strategy triggering conditions, match strategies from the strategy library based on the strategy triggering conditions, and calculate strategy priorities.

[0074] The report output unit is used to output a diagnostic report, which includes a list of diagnostic issues, a list of strategies, and a priority ranking result.

[0075] Advantages of this invention:

[0076] 1. Achieve automated processing and unified management of multi-source heterogeneous spatial data. Through unified coding, spatiotemporal indexing, and quality scoring mechanisms, multi-source data such as POI, heat map, road network, public transport, boundaries, buildings, and land use are automatically linked to the same spatial object, eliminating the need for manual sorting and comparison, thus improving data processing efficiency and result consistency.

[0077] 2. Achieve precise diagnosis of anomaly types. By using the standardized residuals between predicted and observed vitality values, the system automatically distinguishes five anomaly types: high potential-low vitality, low potential-high vitality, balanced development, low potential-low vitality, and structural mismatch, making the diagnosis of weaknesses more targeted.

[0078] 3. Incorporate spatial graph topology and indicator interaction relationships into automated diagnostic rules. Calculate topological features such as node connectivity, betweenness centrality, and boundary penetration strength using a spatial graph model. By comparing single-factor and interaction explanatory power, automatically identify nonlinear enhancement, two-factor enhancement, single-factor dominance, and interaction weakening relationships, uncovering combinatorial problems that are difficult to detect with single-indicator analysis.

[0079] 4. Establish a quantifiable strategy prioritization mechanism. Calculate strategy priorities based on a comprehensive analysis of residual severity, explanatory power of dominant indicators, interaction enhancement coefficient, spatial scale weight, strategy fit coefficient, and implementation cost coefficient, and output these priorities in a tiered manner for easy tiered processing.

[0080] 5. Supports confidence assessment and closed-loop dynamic updates. Diagnostic confidence is calculated by comprehensively considering data quality, model error, sample coverage, and residual stability; low-confidence results are automatically marked. It supports data feedback after strategy implementation, automatically correcting model parameters and strategy adaptation coefficients, forming a closed-loop process of "diagnosis—strategy—feedback—optimization". Attached Figure Description

[0081] Figure 1 This is the overall flowchart of the method for identifying abnormal activity around campuses and recommending strategies as described in this invention;

[0082] Figure 2 This is a structural block diagram of the campus perimeter activity anomaly identification and strategy recommendation system described in this invention;

[0083] Figure 3 This is the logic diagram for identifying abnormal vitality defects as described in this invention;

[0084] Figure 4 This is the logic diagram of the enhanced interaction recognition and problem triggering described in this invention;

[0085] Figure 5 This is the logic diagram of strategy recommendation and priority ranking described in this invention. Detailed Implementation

[0086] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0087] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0088] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0089] Example 1:

[0090] The following is combined Figures 1-5 This embodiment describes a method for identifying anomalies in campus surroundings and recommending strategies based on multi-source heterogeneous spatial data fusion and graph models, comprising the following steps:

[0091] Step 1: Acquisition and preprocessing of multi-source data.

[0092] The system automatically acquires heterogeneous spatial data from multiple sources, including POI data, thermal or passenger flow raster data, road network vector data, bus stop data, campus boundary data, building outline data, and land use facility data, through multi-source data interfaces, and writes them into the system's spatiotemporal index library.

[0093] The acquired multi-source heterogeneous spatial data are subjected to unified coordinate projection, time slice alignment, duplicate data removal, outlier processing, and spatial topology correction.

[0094] Specifically, for POI data, deduplication, classification correction, and spatial positioning are performed; for thermal raster data, resampling, noise filtering, and spatial clipping are performed; for road network data, breakpoint repair, connectivity checks, and topology node generation are performed; and for campus boundary and plot boundary data, self-intersection repair, hole handling, and boundary simplification are performed.

[0095] Then, the preprocessed data is evaluated for quality, and a data quality score is calculated:

[0096] ;

[0097] Among them, among them, Indicates the coordinate offset score. Indicates the time-slice matching score. Indicates the confidence score of the classification. This indicates that you are resetting your credit score. Indicates the topological integrity score. to This indicates configurable weights. For data below a preset quality threshold, automatic removal, manual review and labeling, or interpolation correction are performed to generate a reliable dataset.

[0098] Step 2: Encoding spatial objects across scales and constructing spatial graph models.

[0099] First, a cross-scale spatial object encoding is established based on a trusted dataset. The cross-scale spatial object encoding is then generated.

[0100] ;

[0101] in, This indicates the scale type, including regional scale, campus boundary scale, and plot scale. Indicates the campus object number. Indicates the specific spatial unit number, This represents the time slice number. Then, various data types from the trusted dataset are spatially linked to the same spatial object: for point data, they are linked to the spatial object through point-polygon overlay and nearest neighbor matching; for line data, they are linked through line-polygon intersection, node extraction, and graph topology reconstruction; for raster data, they are linked through raster pruning, resampling, and region statistics. Finally, an R-tree spatial index and a TimeCode temporal index are established to enable cross-scale data querying and result backtracking.

[0102] Simultaneously, a spatial graph model is constructed. Campus boundary openings, road intersections, bus stops, plot entrances, and POI clusters are abstracted as graph nodes. The relationships between road segments, pedestrian walkways, and boundary accessibility and service reachability are abstracted as graph edges. Boundary weights are determined based on walking distance, road grade, boundary permeability, slope resistance, pedestrian crossing resistance, and traffic service intensity. Construct and obtain a spatial graph model Then, the graph topological features are calculated, including node connectivity, average shortest path, reachability time, betweenness centrality, service coverage strength, and boundary penetration strength.

[0103] Step 3: Construction of the morphological-vitality feature matrix.

[0104] Based on cross-scale spatial object encoding and graph topology features, the morphological features, accessibility features, interface features, and business features of each spatial object unit are calculated to construct a morphological-vitality feature matrix.

[0105] Among them, morphological characteristics include building coverage, green space ratio, plan compactness, plan regularity, and boundary permeability index. The boundary permeability index is calculated by the number of accessible entrances and exits, boundary length, and road connection strength.

[0106] Accessibility characteristics include road density, integration, selectivity, connectivity, and transit station service intensity.

[0107] Interface features include boundary openness and the number of accessible connections.

[0108] Business characteristics include the number of POIs, POI core density, business mix entropy, and the proportion of production, living, and public service POIs.

[0109] Organize all the above indicators according to spatial object units to form a morphological-vitality characteristic matrix. When multiple time slices are included, a spatiotemporal feature tensor is formed. .

[0110] Step 4: Vitality prediction, vitality observation and anomaly identification.

[0111] Predicted vitality value based on morphological-vitality feature matrix The predicted vitality values ​​are calculated using a weighted linear model, a random forest model, a gradient boosting tree model, a support vector regression model, a graph convolutional network, a graph attention network, or a combined model based on explanatory power weights. During model training, stratified sampling is performed according to campus, spatial scale, and time slice, and model error is obtained through cross-validation.

[0112] Calculate the observation vitality value based on the observation vitality characterization data in the trusted dataset. The observed vitality value is calculated by weighting the POI kernel density, thermal value, or passenger flow value.

[0113] Then calculate the standardized residual. ,

[0114] in, Represents a spatial object unit. Indicates a time slice, This represents the standard deviation of residuals for samples of the same scale. To prevent extremely small constants with a denominator of zero.

[0115] according to Value and and The threshold relationship is used to determine the anomaly type according to the following rules:

[0116] when , ,and When it is determined to be of the high potential-low vitality type, it means that the spatial morphological conditions are met but the actual vitality is insufficient.

[0117] when , , When it is determined to be of the low potential-high vitality type, it means that the actual vitality is higher than the morphological carrying capacity.

[0118] when When the time frame is reached, it is determined to be a balanced development type, indicating that the predictive potential and the observational activity are basically consistent.

[0119] when and When the time is right, it is judged as low potential-low vitality type, indicating that both the spatial form and actual vitality are insufficient.

[0120] When a single indicator deviates significantly and the interaction relationship weakens, it is judged as a structural mismatch type, indicating that there is a lack of synergy between morphological elements.

[0121] in, To predict the high threshold of vitality value, To predict the low threshold of vitality value, The threshold for residual abnormalities.

[0122] Step 5: Enhanced Interaction Recognition and Policy Matching Recommendation.

[0123] Based on the interaction and enhancement relationship between the indicators in each column of the morphology-vitality feature matrix, the diagnostic problem corresponding to the abnormality type described in step 4 is determined, and the strategy triggering condition is generated. Based on the strategy triggering condition, a strategy is matched from the strategy library and the strategy priority is calculated.

[0124] The method for determining the interaction-enhancing relationship is as follows: based on the explanatory power of a single factor. , With interactive explanatory power The comparison results determine the interaction type.

[0125] when When this occurs, it is determined to be nonlinear enhancement, indicating that the combined effect of the two indicators significantly amplifies the effect, triggering a combined strategy.

[0126] when When the condition is met, it is determined to be a two-factor enhancement, indicating that the two indicators have a synergistic effect and are included in the strategy matching.

[0127] when When the situation is determined to be dominated by a single factor, the strategy corresponding to the dominant indicator should be the primary approach.

[0128] when When this occurs, it is determined to be a weakened interaction, indicating that there is a spatial mismatch or element conflict, triggering a structural mismatch diagnosis.

[0129] Strategy priority is calculated using the following formula: ,

[0130] in, For residual severity, As the dominant indicator of explanatory power, For interaction enhancement coefficient, Spatial scale weights, For strategy adaptation coefficients, To implement cost coefficients, , , , , , Configurable weights.

[0131] according to Values ​​divide strategies into:

[0132] Level 1 Strategy Secondary strategy Three-level strategy and Level 4 strategy Perform hierarchical output.

[0133] Step 6: Output the diagnostic report.

[0134] Output a diagnostic report, which includes a list of diagnostic issues, a list of strategies, and a priority ranking.

[0135] Furthermore, the method also includes diagnostic confidence calculation and closed-loop feedback update steps:

[0136] Based on data quality score Model error Sample coverage and residual stability Calculate the diagnostic confidence level:

[0137] ;

[0138] when When the exception type and strategy list are directly output, The diagnostic results are output in real time and a manual review is prompted. Provides suggestions for insufficient output data and additional data collection.

[0139] The changes in POI, thermal changes, and spatial transformation information after the strategy implementation are re-integrated as feedback data. The changes in residuals and vitality observations before and after implementation are compared to form the strategy effect feedback value. And used to correct the strategy adaptation coefficient. and implementation cost coefficient This enables closed-loop calculations involving data access, anomaly identification, strategy output, effect feedback, and parameter updates.

[0140] Example 2:

[0141] The following is combined Figure 2 This embodiment describes a campus perimeter activity anomaly identification and strategy recommendation system based on multi-source heterogeneous spatial data fusion and graph model, comprising the following units:

[0142] The data acquisition and preprocessing unit is used to acquire multi-source heterogeneous spatial data through a multi-source data interface, perform preprocessing and quality evaluation, and generate a reliable dataset. The multi-source heterogeneous spatial data includes POI point data, heat map or passenger flow raster data, road network vector data, bus stop data, campus boundary data, building outline data, and land use facility data. The preprocessing includes unified coordinate projection, time slice alignment, duplicate data removal, outlier handling, and spatial topology correction. The quality evaluation includes calculating a data quality score.

[0143] ,

[0144] Data below the preset quality threshold is automatically removed, manually reviewed and marked, or interpolated for correction.

[0145] The spatial coding and graph modeling unit is used to establish cross-scale spatial object coding based on the trusted dataset, and to construct a spatial graph model and compute graph topological features. The cross-scale spatial object coding is as follows:

[0146] ;

[0147] in This indicates the scale type, including regional scale, campus boundary scale, and plot scale. Indicates the campus object number. Indicates the specific spatial unit number, Indicates the time slice number. Spatial diagram model. The construction method is as follows: campus boundary openings, road intersections, bus stops, plot entrances, and POI clusters are abstracted as graph nodes. The relationships between road segments, pedestrian walkways, and boundary accessibility and service reachability are abstracted as graph edges. Boundary weights are determined based on walking distance, road grade, boundary permeability, slope resistance, pedestrian crossing resistance, and traffic service intensity. Graph topology features include node connectivity, average shortest path, reachability time, betweenness centrality, service coverage strength, and boundary penetration strength.

[0148] The feature matrix construction unit is used to calculate the morphological features, accessibility features, interface features, and business format features of each spatial object unit based on the cross-scale spatial object encoding and the graph topological features, and to construct a morphological-vitality feature matrix. Time and space feature tensor The morphological characteristics include building coverage, green space ratio, planar compactness, planar regularity, and boundary permeability index; the accessibility characteristics include road density, integration, selectivity, connectivity, and public transport station service intensity; the interface characteristics include boundary openness and the number of passable connections; and the business format characteristics include the number of POIs, POI core density, business format hybrid entropy, and the proportion of production, living, and public service POIs.

[0149] The vitality prediction and anomaly identification unit is used to calculate the predicted vitality value based on the morphology-vitality feature matrix. And calculate the observation vitality value based on the observation vitality characterization data in the trusted dataset. Anomaly types are identified based on the difference between predicted and observed vitality values. Predicted vitality values ​​are calculated using weighted linear models, random forest models, gradient boosting tree models, support vector regression models, graph convolutional networks, graph attention networks, or combined models based on explanatory power weights. Observed vitality values ​​are calculated by weighting POI kernel density, heat values, or passenger flow values. The anomaly type identification method involves calculating standardized residuals. ,according to Value and and The threshold relationship is determined as high potential-low vitality type, low potential-high vitality type, balanced development type, low potential-low vitality type or structural mismatch type.

[0150] The strategy matching and recommendation unit is used to determine the diagnostic question corresponding to the anomaly type based on the interaction enhancement relationships between the indicators in each column of the morphology-vitality feature matrix, generate strategy triggering conditions, match strategies from the strategy library based on the strategy triggering conditions, and calculate strategy priorities. The interaction enhancement relationships are based on single-factor explanatory power. , With interactive explanatory power The comparison results determine whether the strategy is nonlinear enhancement, two-factor enhancement, single-factor dominance, or interaction weakening. The strategy priority is calculated using the following formula: ,according to The strategy is divided into four levels for hierarchical output.

[0151] The report output unit is used to output a diagnostic report, which includes a list of diagnostic issues, a list of strategies, and a priority ranking result.

[0152] The data quality scoring unit is used to calculate scores for coordinate consistency, temporal consistency, type consistency, duplication probability, and topological integrity, and to perform removal, correction, or review marking on low-quality data.

[0153] The model training and confidence calibration unit is used to train the theoretical vitality potential prediction model. Calculate model error, sample coverage, and diagnostic confidence. And low-confidence results are marked.

[0154] Diagnostic confidence .

[0155] The feedback update unit is used to receive information on POI changes, thermal changes, and spatial transformations after the strategy is implemented, and to update the strategy adaptation coefficients and model parameters.

[0156] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for identifying anomalies in campus surroundings and recommending strategies based on multi-source heterogeneous spatial data fusion and graph models, characterized in that, It includes: Step 1: Obtain multi-source heterogeneous spatial data through multi-source data interfaces, perform preprocessing and quality evaluation, and generate a reliable dataset; Step 2: Based on the trusted dataset, establish cross-scale spatial object encoding, construct a spatial graph model, and calculate graph topological features; Step 3: Based on the cross-scale spatial object encoding and the graph topology features, calculate the morphological features, accessibility features, interface features and business features of each spatial object unit, and construct a morphological-vitality feature matrix; Step 4: Calculate the predicted vitality value based on the morphology-vitality feature matrix, and calculate the observed vitality value based on the observed vitality characterization data in the trusted dataset. Identify the anomaly type based on the difference between the predicted vitality value and the observed vitality value. Step 5: Based on the interaction enhancement relationship between the indicators in each column of the morphology-vitality feature matrix, determine the diagnostic problem corresponding to the abnormality type described in Step 4, generate strategy triggering conditions, match strategies from the strategy library according to the strategy triggering conditions, and calculate the strategy priority. Step 6: Output a diagnostic report, which includes a list of diagnostic issues, a list of strategies, and a priority ranking result.

2. The method for identifying and recommending strategies for campus perimeter activity anomalies based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, The multi-source heterogeneous data mentioned in step 1 includes point data, line data, area data, and raster data; the point data includes POIs, bus stops, and facility points; the line data includes road centerlines, walking paths, and bus routes; the area data includes campus boundaries, plot boundaries, building outlines, and green space areas; and the raster data includes heat maps, passenger flow raster, or nighttime activity raster. The preprocessing of multi-source heterogeneous data includes: unified coordinate projection, time slice alignment, duplicate data removal, outlier handling, and spatial topology correction. The quality assessment of multi-source heterogeneous data includes: Calculate data quality score : ; in, Indicates the coordinate offset score. Indicates the time-slice matching score. Indicates the confidence score of the classification. This indicates that you are resetting your credit score. Indicates topological integrity score. , , , , Indicates configurable weights; For data that is below the preset quality threshold, automatic rejection, manual verification and marking, or interpolation correction are performed.

3. The method for identifying and recommending strategies for campus perimeter activity anomalies based on multi-source heterogeneous spatial data fusion and graph models according to claim 2, characterized in that, Step 2, establishing cross-scale spatial object encoding, specifically includes: Generate cross-scale codes; The multi-source heterogeneous spatial data are linked to the same spatial object through a spatial connection method; Establish R-tree spatial indexes and TimeCode temporal indexes to enable cross-scale data querying and result backtracking; Specifically, the generation of cross-scale codes involves: ； in, Indicates the scale type, including regional scale, campus boundary scale, and plot scale; Indicates the campus object number; Indicates the specific spatial unit number; Indicates the time slice number; The spatial connection methods include: For point data, the system attaches it to spatial objects through point-surface overlay and proximity matching; For line data, the system achieves connection through line-plane intersection, node extraction, and graph topology reconstruction; For raster data, the system achieves the linking through raster clipping, resampling, and regional statistics.

4. The method for identifying and recommending strategies for campus perimeter activity anomalies based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, Step 2, which involves constructing the spatial graph model, specifically includes: Campus boundary openings, road intersections, bus stops, plot entrances, and POI clusters are abstracted as graph nodes. The relationships between road segments, pedestrian walkways, and boundary accessibility and service reachability are abstracted as graph edges. Boundary weights are determined based on walking distance, road grade, boundary permeability, slope resistance, pedestrian crossing resistance, and traffic service intensity. ; Constructing a spatial graph model : ; The graph topology features include node connectivity, average shortest path, reachability time, betweenness centrality, service coverage strength, and boundary penetration strength.

5. The method for identifying and recommending strategies for campus perimeter activity anomalies based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, The morphological features described in step 3 include building coverage, green space ratio, planar compactness, planar regularity, and boundary permeability index. The boundary permeability index is calculated by the number of accessible entrances and exits, boundary length, and road connectivity strength. The accessibility features include road density, integration, selectivity, connectivity, and public transport station service intensity. The interface features include boundary openness and the number of accessible connections. The business format features include the number of POIs, POI core density, business format mix entropy, and the proportion of production, living, and public service POIs.

6. The method for identifying and recommending strategies for anomalies in campus surroundings based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, The predicted vitality value in step 4 is calculated using a weighted linear model, a random forest model, a gradient boosting tree model, a support vector regression model, a graph convolutional network, a graph attention network, or a combined model based on explanatory power weights. The model is stratified and sampled according to campus, spatial scale, and time slice during training, and the model error is obtained through cross-validation. The observed vitality value is calculated by weighting POI kernel density, thermal value, or passenger flow value.

7. The method for identifying and recommending strategies for campus perimeter activity anomalies based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, Step 4 describes a specific method for identifying anomaly types based on the difference between predicted and observed vitality values, including: Calculate the standardized residuals: ； in, Represents the standardized residual. Indicates the predicted vitality value. Indicates the observed vitality value. Represents a spatial object unit. Indicates a time slice, This represents the standard deviation of residuals for samples of the same scale. This represents a very small constant to prevent the denominator from being zero; according to value, and Based on the threshold relationship, the anomaly type is determined according to the following rules: when , , When it is determined to be a high-potential-low-vitality type, it means that the spatial morphological conditions are met but the actual vitality is insufficient; when , , When it is determined to be of the low potential-high vitality type, it means that the actual vitality is higher than the form's carrying capacity; when When it is determined to be of a balanced development type, it means that the predictive potential and the observed vitality are basically consistent; when and When it is determined to be of low potential-low vitality type, it means that both the spatial form and actual vitality are insufficient; When a single indicator deviates significantly and the interaction relationship weakens, it is judged as a structural mismatch type, indicating that there is a lack of synergy between morphological elements. in, This indicates a high threshold for predicted vitality. This indicates a low threshold for predicted vitality. It represents the standard deviation of residuals for samples of the same scale.

8. The method for identifying and recommending strategies for campus perimeter activity anomalies based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, The method for determining the interaction enhancement relationship between the indicators in each column of the morphology-vitality feature matrix described in step 5 includes: Explanatory power based on single factor , With interactive explanatory power The comparison results determine the interaction type; when When this occurs, it is determined to be nonlinear enhancement, indicating that the combined effect of the two indicators significantly amplifies the effect, triggering a combined strategy; when When the condition is met, it is determined to be a two-factor enhancement, indicating that the two indicators have a synergistic effect and are included in the strategy matching. when When the time is right, it is determined to be dominated by a single factor, indicating that the strategy corresponding to the dominant indicator is the main one; when When this occurs, it is determined to be a weakened interaction, indicating that there is a spatial mismatch or element conflict, triggering a structural mismatch diagnosis.

9. The method for identifying and recommending strategies for anomalies in campus surroundings based on multi-source heterogeneous spatial data fusion and graph models according to claim 1, characterized in that, The specific method for calculating the strategy priority in step 5 includes: ； in, Indicates the severity of the residual. This indicates the explanatory power of the dominant indicator. Indicates the interaction enhancement coefficient. Indicates spatial scale weights, This represents the strategy fit coefficient. This represents the implementation cost coefficient. , , , , , Indicates configurable weights; according to The strategy is divided into four levels: Level 1, Level 2, Level 3, and Level 4, and output accordingly. when At this time, it is a level one strategy; when At that time, it is a secondary strategy; when At that time, it is a three-level strategy; when At that time, it was a level four strategy.

10. A campus perimeter activity anomaly identification and strategy recommendation system based on multi-source heterogeneous spatial data fusion and graph model, used to implement the method of any one of claims 1 to 9, characterized in that, It includes: The data acquisition and preprocessing unit is used to acquire multi-source heterogeneous spatial data through multi-source data interfaces, perform preprocessing and quality evaluation, and generate a reliable dataset. The spatial coding and graph modeling unit is used to establish cross-scale spatial object coding based on the trusted dataset, and to construct a spatial graph model and calculate graph topological features; The feature matrix construction unit is used to calculate the morphological features, accessibility features, interface features, and business features of each spatial object unit based on the cross-scale spatial object encoding and the graph topology features, and to construct a morphological-vitality feature matrix. The vitality prediction and anomaly identification unit is used to calculate the predicted vitality value based on the morphology-vitality feature matrix, and to calculate the observed vitality value based on the observed vitality characterization data in the trusted dataset, and to identify the anomaly type based on the difference between the predicted vitality value and the observed vitality value. The strategy matching and recommendation unit is used to determine the diagnostic problem corresponding to the anomaly type based on the interaction enhancement relationship between the indicators in each column of the morphology-vitality feature matrix, generate strategy triggering conditions, match strategies from the strategy library based on the strategy triggering conditions, and calculate strategy priorities. The report output unit is used to output a diagnostic report, which includes a list of diagnostic issues, a list of strategies, and a priority ranking result.

Images