Territorial transaction dynamic monitoring and analysis system based on multi-source data fusion

Abstract

The invention relates to the technical field of geographic information, in particular to a territorial transaction dynamic monitoring and analysis system based on multi-source data fusion, which comprises a data integration and storage unit, a knowledge graph construction unit, a real-time verification unit, a trusted transaction verification unit and a dynamic monitoring and decision support unit. According to the method, heterogeneous data are integrated through a multi-modal data lake architecture, space-time reference unification is carried out, a transaction element association network is constructed by utilizing a knowledge graph technology to generate a dynamic verification rule, and real-time compliance filtering of a transaction process is realized in combination with an Flink flow processing engine; and transaction price rationality verification and subject correlation analysis are completed by means of an improved isolated forest algorithm and a graph convolutional network, and evidence storage is performed based on a block chain technology, so that full-process intelligent management of land transaction data, accurate identification and prevention and control of risks, and efficient and datamation support of supervision decision are realized.

Description

Technical Field

[0001] The present invention relates to the field of geographic information technology, and in particular to a land transaction dynamic monitoring and analysis system based on multi-source data fusion. Background Art

[0002] In existing technologies, traditional land transaction supervision has problems such as scattered and isolated data, delayed risk identification, and insufficient decision-making support.

[0003] Based on this, the present invention provides a dynamic monitoring and analysis system for land transactions based on multi-source data fusion to solve the technical problems raised above. Summary of the Invention

[0004] The purpose of the present invention is to provide a dynamic monitoring and analysis system for land transactions based on multi-source data fusion. The multimodal data lake architecture of the present invention integrates heterogeneous data and unifies the spatiotemporal benchmarks. It uses knowledge graph technology to construct a transaction element association network to generate dynamic verification rules. It combines the Flink stream processing engine to realize real-time compliance filtering of the transaction process. With the help of the improved isolation forest algorithm and graph convolutional network, it completes the rationality verification of transaction prices and subject correlation analysis and stores evidence based on blockchain technology. Finally, through the heat index model and idle warning mechanism, it quantifies market fluctuations and outputs risk signals, thereby realizing the full-process intelligent governance of land transaction data, accurate identification and prevention of risks, and efficient and data-based support for regulatory decision-making.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention provides a dynamic monitoring and analysis system for land transactions based on multi-source data fusion, including a data integration and storage unit, a knowledge graph construction unit, a real-time verification unit, a trusted transaction verification unit, and a dynamic monitoring and decision support unit, wherein:

[0007] The data integration and storage unit is used to integrate natural resource data, transaction data, public opinion data, and macroeconomic indicators through a multimodal data lake architecture, and use spatiotemporal coding technology to perform unified spatiotemporal benchmark mapping of heterogeneous data;

[0008] The knowledge graph construction unit is used to use knowledge graph technology to construct a three-dimensional association network that includes the equity penetration relationship of transaction entities, land development constraints, and policy effectiveness time windows, and dynamically generate real-time verification rules through semantic analysis of policy terms;

[0009] The real-time verification unit: Based on the Flink stream processing engine, it establishes a transaction process rule base, executes the real-time verification rules generated by the knowledge graph construction unit, and filters the transaction process for compliance;

[0010] The trusted transaction verification unit is used to verify the rationality of transaction prices by using an improved isolation forest algorithm, verify the relevance of transaction entities in combination with a graph convolutional network, conduct secondary verification of transactions that pass the real-time verification unit, and generate trusted evidence of compliant transactions based on blockchain technology;

[0011] The dynamic monitoring and decision support unit is used to quantify market fluctuations using a heat index model and output idle land early warning signals.

[0012] The data integration and storage unit includes a multimodal data access module, a spatiotemporal coding mapping module, and a data lake management module, wherein:

[0013] The multimodal data access module is used to receive data from different sources and formats on natural resources, transactions, public opinion, and economic indicators;

[0014] The spatiotemporal coding mapping module is used to map heterogeneous data to a unified spatiotemporal reference using spatiotemporal coding technology;

[0015] The data lake management module is used to build and manage a multimodal data lake.

[0016] The spatiotemporal coding technology uses the GeoHash algorithm to encode the spatial location of land parcels, combines the ISO 8601 standard to format transaction timestamps, unifies the spatiotemporal benchmarks of data, and achieves a coding accuracy of 6 GeoHash characters.

[0017] The knowledge graph construction unit includes an entity relationship extraction module, a policy analysis module, and a graph storage module, wherein:

[0018] The entity relationship extraction module is used to extract the corporate equity hierarchy relationship from the industrial and commercial data using NLP technology;

[0019] The policy parsing module: identifies the constraints in the policy text based on the BERT model and converts them into executable rules;

[0020] The graph storage module is used to store the "enterprise-land-policy" association network using the Neo4j graph database and supports multi-hop queries.

[0021] The policy parsing module uses the BERT model to identify the constraints in the policy text and convert them into executable rules. The specific operations are as follows:

[0022] A1: Text vectorization: Convert the policy text into a token sequence X = [t1, t2, ..., t n ], the subword unit is generated by the WordPiece word segmentation algorithm, and its word vector is represented as:

[0023] ei =E·t i +P i +S i

[0024] Where E is the word embedding matrix, P i is the position code, S i Segment embedding for sentences;

[0025] A2: BERT semantic encoding: Generate context vector H = {h1,…,h n}, where the attention calculation of the lth layer is:

[0026]

[0027] Where Q, K, and V are query, key, and value matrices, and d k is the key vector dimension;

[0028] A3: Rule Conversion: The identified constraint clause "If condition C then action A" is converted into a Flink SQL rule, where the extraction probability of condition C satisfies:

[0029] p(C)=σ(W c h0+b c )

[0030] Where σ is the Sigmoid function, W c and b c is the classification layer parameter.

[0031] The real-time verification unit includes a rule base management module, a real-time stream processing module, and a compliance filtering module, wherein:

[0032] The rule base management module is used to establish and maintain the transaction process rule base and store various verification rules;

[0033] The real-time stream processing module: based on the Flink stream processing engine, processes transaction process data in real time;

[0034] The compliance filtering module is used to execute the verification rules generated by the knowledge graph, perform compliance filtering on the transaction process, generate log records for intercepted abnormal transactions, including transaction ID, violation type, triggering rules, and push warning information to the supervision platform.

[0035] The trusted transaction verification unit includes a price rationality verification module, a correlation analysis module, a secondary verification module, and a blockchain evidence storage module, wherein:

[0036] The price rationality verification module is used to identify abnormal transactions that deviate from the normal price range by ≥2σ using a weighted isolation forest algorithm;

[0037] The association analysis module is used to calculate the similarity of bids between enterprises through GCN, and mark bids with a similarity greater than 0.9 as suspected bid rigging;

[0038] The secondary verification module is used to perform secondary verification on transactions that pass the real-time verification;

[0039] The blockchain evidence storage module is used to write the hash of the bidding documents and key data of the deposit certificate of the compliant transaction into the node on the Hyperledger Fabric chain.

[0040] In the association analysis module, GCN is used to calculate the similarity of bids between companies. If the similarity is greater than 0.9, it is marked as suspected bid rigging. The specific operation is as follows:

[0041] B1. Constructing an enterprise association graph: Using bidding enterprises as nodes and equity relationships and legal representative overlap as edges, we construct the initial graph structure.

[0042] B2. Graph convolution feature extraction: Node features are iteratively updated through the GCN model. The formula for updating the k-th layer features is:

[0043]

[0044] in, is an adjacency matrix with self-connection, is the degree matrix, W ( k) is a trainable weight;

[0045] B3. Similarity matrix calculation: For the final layer node feature H ( K) Calculate the cosine similarity matrix S, where:

[0046]

[0047] B4. Determination of bid-rigging behavior: When S ij When >0.9, enterprise i and enterprise j are marked as suspected bid-rigging entities.

[0048] The blockchain evidence storage module writes the key data of the bidding document hash and the deposit certificate of the compliant transaction into the Hyperledger Fabric chain node. The specific operations are as follows:

[0049] C1: Preprocessing of evidence data: Extract key fields such as the bid document SHA-256 hash value, security deposit certificate number, and timestamp from verified transaction data, and generate an evidence storage request in JSON format;

[0050] C2: Smart contract trigger: Calls the Hyperledger Fabric chain code to verify whether the evidence storage request meets the following conditions:

[0051] a) The transaction has passed secondary verification by the trusted transaction verification unit;

[0052] b) The compliance status output by the real-time verification unit is "passed";

[0053] C3: Multi-node consensus: Based on the preset endorsement strategy, the evidence data is packaged to generate a new block;

[0054] C4: On-chain storage: Write the block to the LevelDB database of each node and return the evidence certificate including transaction ID, block height, and Merkle tree root hash.

[0055] The dynamic monitoring and decision support unit includes a market heat calculation module, an idle warning module, and a visualization module, wherein:

[0056] The market heat calculation module generates a regional heat index based on bidding participation × average premium rate, where bidding participation = number of actual bidding companies / number of potential bidding companies × 100%, and average premium rate = ∑ (land parcel transaction price - starting price) / ∑ starting price;

[0057] The idle warning module is used to compare the development time nodes agreed in the land transfer contract. If the construction fails to start within 6 months, a red warning will be triggered, and if the construction fails to start within 3 months, a yellow warning will be triggered.

[0058] The visualization module is used to display regional heat distribution and idle land locations through GIS maps, and supports drilling to view related enterprise information.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] This invention integrates heterogeneous data through a multimodal data lake architecture and unifies spatiotemporal benchmarks, uses knowledge graph technology to construct a transaction element association network to generate dynamic verification rules, combines the Flink stream processing engine to achieve real-time compliance filtering of transaction processes, and uses an improved isolation forest algorithm and graph convolutional network to complete transaction price rationality verification and subject correlation analysis and store evidence based on blockchain technology. Finally, through a heat index model and idle warning mechanism, market fluctuations are quantified and risk signals are output, thereby realizing full-process intelligent governance of land transaction data, accurate identification and prevention of risks, and efficient and data-based support for regulatory decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a system diagram of the land transaction dynamic monitoring and analysis system based on multi-source data fusion of the present invention.

[0062] Figure 2 This is a flow chart for constructing a knowledge graph in the dynamic monitoring and analysis system of land transactions based on multi-source data fusion in the present invention.

[0063] Figure 3 This is a real-time verification flow chart of the land transaction dynamic monitoring and analysis system based on multi-source data fusion in the present invention.

[0064] Description of Figure Numbers:

[0065] 100. Data integration and storage unit; 101. Multimodal data access module; 102. Spatiotemporal coding mapping module; 103. Data lake management module; 200. Knowledge graph construction unit; 201. Entity relationship extraction module; 202. Policy analysis module; 203. Graph storage module; 300. Real-time verification unit; 301. Rule base management module; 302. Real-time stream processing module; 303. Compliance filtering module; 400. Trusted transaction verification unit; 401. Price rationality verification module; 402. Correlation analysis module; 403. Secondary verification module; 404. Blockchain evidence storage module; 500. Dynamic monitoring and decision support unit; 501. Market heat calculation module; 502. Idle warning module; 503. Visualization module. DETAILED DESCRIPTION

[0066] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0067] Example:

[0068] like Figure 1-Figure 3As shown, this embodiment provides a dynamic monitoring and analysis system for land transactions based on multi-source data fusion, including a data integration and storage unit 100, a knowledge graph construction unit 200, a real-time verification unit 300, a trusted transaction verification unit 400, and a dynamic monitoring and decision support unit 500, wherein: the data integration and storage unit 100: is used to integrate natural resource data, transaction data, public opinion data, and macroeconomic indicators through a multimodal data lake architecture, and adopts spatiotemporal coding technology to perform unified spatiotemporal benchmark mapping of heterogeneous data; the knowledge graph construction unit 200: is used to adopt knowledge graph technology to construct a three-dimensional data system including the equity penetration relationship of the transaction subject, the land development constraint conditions, and the policy effectiveness time window. The system constructs a transaction flow chart based on the knowledge graph construction unit 200, establishes a transaction flow rule base, executes the real-time verification rules generated by the knowledge graph construction unit 200, and filters the transaction flow for compliance; the ... and generates a trustworthy evidence for the compliant transaction based on the blockchain technology; the system constructs a transaction flow chart based on the knowledge graph construction unit 200, and generates a trustworthy evidence

[0069] Among them, it should be noted that the data integration and storage unit 100 realizes the unified governance of multi-source heterogeneous data, establishes a deep correlation network of transaction elements through the knowledge graph construction unit 200, the real-time verification unit 300 performs instant compliance review of the transaction process based on dynamic rules, and the trusted transaction verification unit 400 uses intelligent algorithms to complete the dual verification of the authenticity of the transaction and blockchain evidence storage, and finally the dynamic monitoring and decision support unit 500 realizes the quantitative assessment of the market status and risk warning.

[0070] In this embodiment, it should also be noted that the data integration and storage unit 100 includes a multimodal data access module 101, a spatiotemporal coding mapping module 102, and a data lake management module 103. The multimodal data access module 101 is used to receive data from various sources and formats, including natural resources, transactions, public opinion, and economic indicators. The spatiotemporal coding mapping module 102 is used to map heterogeneous data to a unified spatiotemporal benchmark using spatiotemporal coding technology. This spatiotemporal coding technology uses the GeoHash algorithm to encode the spatial location of land parcels and formats transaction timestamps in accordance with the ISO 8601 standard, unifying the data in terms of spatiotemporal benchmarks with an encoding accuracy of 6 GeoHash characters. The data lake management module 103 is used to build and manage a multimodal data lake.

[0071] It should be noted that the multimodal data access module 101 realizes the unified collection of multi-source heterogeneous data, and the unified mapping of spatiotemporal benchmarks is completed by the spatiotemporal coding mapping module 102 using the GeoHash algorithm (6-bit precision) and the ISO 8601 standard. Finally, the data lake management module 103 constructs a standardized and traceable multimodal data lake.

[0072] Furthermore, it should be noted that the multimodal data access module 101 supports access to land parcel map data (e.g., in CAD format) from the natural resources department, listing and transaction data (in JSON / XML format) from trading platforms, online public opinion crawler data (in text / HTML format), and macroeconomic indicators from the National Bureau of Statistics (through an API interface). The data lake management module 103 builds a multimodal data lake based on Hadoop, Hive, and HBase: ① Structured data (transaction records, economic indicators) is stored in the Hive data warehouse, partitioned by "time + region" to support efficient SQL queries; ② Semi-structured data (policy texts, public opinion information) is stored in HBase, using "spatiotemporal code + data type" as row keys for fast retrieval; ③ Unstructured data (land parcel maps, remote sensing images) is stored in HDFS, with metadata (e.g., hash value, spatial extent) associated with the Hive table. Data lineage tracking, quality monitoring (verifying spatiotemporal code integrity and field null value rates), and lifecycle management (automatically archiving historical data for more than three years) are also provided to ensure data quality and availability.

[0073] In this embodiment, it should also be noted that the knowledge graph construction unit 200 includes an entity relationship extraction module 201, a policy analysis module 202, and a graph storage module 203, wherein: the entity relationship extraction module 201 is used to extract the corporate equity hierarchy relationship from industrial and commercial data using NLP technology; the policy analysis module 202 is used to identify the constraints in the policy text based on the BERT model and convert them into executable rules; the specific operations are as follows: A1: Text vectorization: convert the policy text into a token sequence X = [t1, t2, ..., t n ], the subword unit is generated by the WordPiece word segmentation algorithm, and its word vector is represented as:

[0074] e i =E·t i +P i +S i

[0075] Where E is the word embedding matrix, P i is the position code, S i Embed the sentence segments; A2: BERT semantic encoding: Generate context vector H={h1,…,h n}, where the attention calculation of the lth layer is:

[0076]

[0077] Where Q, K, and V are query, key, and value matrices, and d k is the key vector dimension; A3: Rule conversion: The identified constraint clause "If condition C then action A" is converted into a Flink SQL rule, where the extraction probability of condition C satisfies:

[0078] p(C)=σ(W c h0+b c )

[0079] Where σ is the Sigmoid function, W c and b c Graph storage module 203: used to store the "enterprise-land-policy" association network using the Neo4j graph database, supporting multi-hop queries.

[0080] Among them, it should be noted that the entity relationship extraction module 201 uses NLP technology to construct the enterprise equity relationship network, and the policy analysis module 202 adopts the BERT model (including WordPiece word segmentation, Transformer encoding and rule probability conversion three-layer processing architecture) to realize the intelligent analysis and rule conversion of policy terms, and finally constructs the "enterprise-land-policy" three-dimensional association network in the Neo4j graph database through the graph storage module 203.

[0081] Furthermore, it should be noted that the entity relationship extraction module 201 uses the BERT-NER model to identify entities such as company names, legal representatives, and plot codes based on industrial and commercial data (corporate annual reports, equity change records) (with an accuracy rate of ≥95%); extracts relationships such as "shareholding ratio" and "legal representative change" through dependency syntactic analysis (such as extracting triples (XX company, shareholding, YY company, 30%) from "XX company holds 30% equity in YY company"); uses depth-first search to traverse the equity hierarchy and generate a "ultimate beneficiary-subsidiary-grandson company" penetration relationship chain, supporting the parsing of nested equity structures with more than 5 layers, and accurately identifying related transaction entities. The graph storage module 203 supports multi-hop queries through Cypher statements (such as querying "plot transactions of subsidiaries of credit-level A companies subject to 2024 policy constraints"), with a response time of ≤500ms (million-level node scale), providing efficient support for complex association analysis.

[0082] In this embodiment, it should also be noted that the real-time verification unit 300 includes a rule base management module 301, a real-time stream processing module 302, and a compliance filtering module 303, wherein: the rule base management module 301: is used to establish and maintain a transaction process rule base and store various verification rules; the real-time stream processing module 302: is based on the Flink stream processing engine, and performs real-time processing of transaction process data; the compliance filtering module 303: is used to execute the verification rules generated by the knowledge graph, perform compliance filtering on the transaction process, generate log records for intercepted abnormal transactions, including transaction ID, violation type, triggering rules, and push warning information to the supervision platform.

[0083] It should be noted that the rule base management module 301 dynamically maintains the verification rule base generated by the knowledge graph, relies on the Flink engine of the real-time stream processing module 302 to realize millisecond-level processing of transaction data, and finally the compliance filtering module 303 performs multi-dimensional rule matching and real-time interception.

[0084] Furthermore, it should be noted that the rule base management module 301 stores static rules (policy constraints, transaction specifications) and dynamic rules (dynamically adjusted through reinforcement learning based on historical anomaly data). It provides rule versioning (recording iteration history), conflict detection (verifying rule compatibility), and phased release (new rules are first applied to 10% of transaction flows) to ensure the stability and reliability of rule updates. The real-time stream processing module 302, based on the Flink stream processing framework, uses Kafka as a message queue to access transaction data, addressing cross-time zone and delayed data timing issues through event-time processing. Combined with window calculations (such as counting the number of land parcel bids in a 1-minute rolling window) and state storage (recording historical corporate violations), it provides real-time context for compliance verification, with processing latency ≤ 200ms. The compliance filtering module 303 uses rule base SQL rules and knowledge graph Cypher queries to filter transaction data in real time. Logs are generated for abnormal transactions (including transaction ID, violation type, and triggering rules), stored in Elasticsearch, and pushed to the regulatory platform via WebSocket. Warning latency is ≤ 200ms, enabling "second-level discovery and real-time resolution" of violations.

[0085] In this embodiment, it should also be noted that the trusted transaction verification unit 400 includes a price rationality verification module 401, an association analysis module 402, a secondary verification module 403, and a blockchain evidence storage module 404, wherein: the price rationality verification module 401 is used to identify abnormal transactions that deviate from the normal price range by ≥2σ using the weighted isolation forest algorithm; the association analysis module 402 is used to calculate the similarity of bid quotations between enterprises through GCN, and mark them as suspected bid rigging when the similarity is >0.9; the specific operations are as follows: B1, constructing an enterprise association graph: using the bidding enterprises as nodes and the equity relationship and legal representative overlap as edges to construct an initial graph structure; B2, graph convolution feature extraction: iteratively updating node features through the GCN model, and the feature update formula of the kth layer is:

[0086]

[0087] in, is an adjacency matrix with self-connection, is the degree matrix, W ( k) is a trainable weight; B3, similarity matrix calculation: for the final layer node feature H ( K) Calculate the cosine similarity matrix S, where:

[0088]

[0089] B4. Determination of bid-rigging behavior: When S ij When the value is greater than 0.9, companies i and j are flagged as suspected bid-rigging entities. Secondary Verification Module 403: Performs secondary verification on transactions that have passed real-time verification. Blockchain Evidence Module 404: Writes the bid document hash and key data of the security deposit certificate to the Hyperledger Fabric on-chain nodes. The specific operations are as follows: C1: Preprocessing of Evidence Data: Extracts the key fields of the bid document SHA-256 hash value, security deposit certificate number, and timestamp from the verified transaction data to generate a JSON-formatted evidence request. C2: Smart Contract Triggering: Invokes the Hyperledger Fabric chaincode to verify whether the evidence request meets the following conditions: a) The transaction has passed secondary verification by the Trusted Transaction Verification Unit 400; b) The compliance status output by the Real-Time Verification Unit 300 is "Passed." C3: Multi-Node Consensus: Based on the preset endorsement policy, the evidence data is packaged and generated into a new block. C4: On-chain Storage: Writes the block to each node's LevelDB database and returns an evidence certificate containing the transaction ID, block height, and Merkle tree root hash.

[0090] Among them, it should be noted that the price rationality verification module 401 adopts the weighted isolation forest algorithm to realize the intelligent screening of transaction prices, and combines the GCN graph convolution network of the association analysis module 402 (including enterprise association graph construction, multi-layer feature extraction and cosine similarity calculation) to deeply mine clues of bid rigging, and completes the double verification of the transaction through the secondary verification module 403. Finally, the hash value, certificate and other key data of the compliant transaction are stored on the chain through the blockchain evidence module 404 after triggering the smart contract and multi-node consensus.

[0091] Furthermore, it should be noted that the price rationality verification module 401 adopts a weighted isolation forest algorithm, selects 8-dimensional features such as "regional benchmark land price, volume ratio, supporting maturity, and historical premium rate", and assigns high weights (such as 0.3) to core features (such as benchmark land price) to reduce the interference of secondary features. The path length from the sample to the root node of the isolation tree is calculated, and when the path length is lower than the 50% percentile, it is marked as an anomaly. The anomaly identification accuracy rate is ≥92%, and transactions with prices deviating from the reasonable market range are accurately discovered. The secondary verification module 403 executes the "price verification → association verification → 5% manual review" process for transactions that have passed the real-time verification. The price verification calls the price rationality verification module 401, and the association verification calls the association analysis module 402. The manual review uses the visual interface to view the anomaly details (such as the price deviation curve and the enterprise association path), and the misjudgment rate is ≤3%, ensuring the accuracy of the verification.

[0092] In this embodiment, it should also be noted that the dynamic monitoring and decision support unit 500 includes a market heat calculation module 501, an idle warning module 502, and a visualization module 503, wherein: the market heat calculation module 501 generates a regional heat index based on the bidding participation rate × the average premium rate, where the bidding participation rate = the number of actual participating enterprises / the number of potential participating enterprises × 100%.

[0093] Average premium rate = ∑(land transaction price - starting price) / ∑ starting price; Idle warning module 502: used to compare the development time nodes agreed in the land transfer contract. If construction fails to start within 6 months of the deadline, a red warning will be triggered, and if construction fails to start within 3 months of the deadline, a yellow warning will be triggered; Visualization module 503: used to display the regional heat distribution and the location of idle land through GIS maps, and supports drilling to view related enterprise information.

[0094] It should be noted that the market heat calculation module 501 realizes a quantitative assessment of the regional market situation based on core indicators such as bidding participation and premium rate, and combines the time-series monitoring function of the idle warning module 502 (3-month yellow warning / 6-month red warning) to accurately identify the risk of land idleness, and finally realizes the multi-dimensional data linkage display of "heat distribution-idle plots-enterprise associations" through the GIS platform of the visualization module 503.

[0095] Furthermore, it should be noted that the idle warning module 502 compares the development time agreed in the land transfer contract and determines the idle status by calculating the time difference: a yellow warning is triggered when it is overdue for 3 months (a rectification plan is pushed), and a red warning is triggered when it is overdue for 6 months (an idle investigation is initiated). The issuance time of the construction project planning license and the construction filing information are integrated as the "basis for commencement of construction" to avoid misjudgment and improve the accuracy of the warning. The visualization module 503 displays the regional heat distribution (heat map) and the location of idle plots (red dot marks) through a GIS map, and supports filtering by administrative district and time. Click on the plot / enterprise to drill down to related transaction records (transaction price, bidding company), policy constraints, and blockchain evidence details; provide dashboards such as market heat trend charts and idle land disposal schedules, support PDF export and email subscription, and assist regulatory authorities in making "data-based and proactive" decisions.

[0096] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0097] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. The land transaction dynamic monitoring and analysis system based on multi-source data fusion is characterized by: The system comprises a data integration and storage unit (100), a knowledge graph construction unit (200), a real-time verification unit (300), a trusted transaction verification unit (400), and a dynamic monitoring and decision support unit (500), wherein: The data integration and storage unit (100) is used to integrate natural resource data, transaction data, public opinion data, and macroeconomic indicators through a multimodal data lake architecture, and to use spatiotemporal coding technology to perform unified spatiotemporal benchmark mapping of heterogeneous data; The knowledge graph construction unit (200) is used to use knowledge graph technology to construct a three-dimensional association network including the equity penetration relationship of transaction entities, land development constraints, and policy effectiveness time window, and dynamically generate real-time verification rules through semantic analysis of policy terms; The real-time verification unit (300) is configured to establish a transaction process rule base based on the Flink stream processing engine, execute the real-time verification rules generated by the knowledge graph construction unit (200), and perform compliance filtering on the transaction process; The trusted transaction verification unit (400) is used to verify the rationality of the transaction price by using an improved isolation forest algorithm, verify the relevance of transaction entities in combination with a graph convolutional network, conduct a secondary verification of the transaction that passes the real-time verification unit (300), and generate a trusted certificate for the compliant transaction based on blockchain technology; The dynamic monitoring and decision support unit (500) is used to quantify market fluctuations using a heat index model and output idle land early warning signals.

2. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 1 is characterized in that: The data integration and storage unit (100) includes a multimodal data access module (101), a spatiotemporal coding mapping module (102), and a data lake management module (103), wherein: The multimodal data access module (101) is used to receive data of different sources and formats on natural resources, transactions, public opinion, and economic indicators; The spatiotemporal coding mapping module (102) is used to map heterogeneous data to a unified spatiotemporal reference using spatiotemporal coding technology; The data lake management module (103) is used to build and manage a multimodal data lake.

3. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 2 is characterized in that: The spatiotemporal coding technology uses the GeoHash algorithm to encode the spatial location of land parcels, combines the ISO 8601 standard to format transaction timestamps, unifies the spatiotemporal benchmarks of data, and achieves a coding accuracy of 6 GeoHash characters.

4. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 1 is characterized in that: The knowledge graph construction unit (200) includes an entity relationship extraction module (201), a policy analysis module (202), and a graph storage module (203), wherein: The entity relationship extraction module (201) is used to extract the corporate equity hierarchy relationship from the industrial and commercial data using NLP technology; The policy parsing module (202) identifies the constraint clauses in the policy text based on the BERT model and converts them into executable rules; The graph storage module (203) is used to store the "enterprise-land-policy" association network using the Neo4j graph database, and supports multi-hop query.

5. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 4 is characterized in that: The policy parsing module (202) identifies the constraints in the policy text based on the BERT model and converts them into executable rules. The specific operations are as follows: A1: Text vectorization: Convert the policy text into a token sequence X = [t1, t2, ..., t n ], the subword unit is generated by the WordPiece word segmentation algorithm, and its word vector is represented as: e i =E·t i +P i +S i Where E is the word embedding matrix, P i is the position code, S i Segment embedding for sentences; A2: BERT semantic encoding: Generate context vector H = {h1,…,h n }, where the attention calculation of the lth layer is: Where Q, K, and V are query, key, and value matrices, and d k is the key vector dimension; A3: Rule conversion: The identified constraint clause "If condition C then action A" is converted into a Flink SQL rule, where the extraction probability of condition C satisfies: p(C)=σ(W c ·h0+b c ) Where σ is the Sigmoid function, W c and b c is the classification layer parameter.

6. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 1 is characterized in that: The real-time verification unit (300) includes a rule base management module (301), a real-time stream processing module (302), and a compliance filtering module (303), wherein: The rule base management module (301) is used to establish and maintain a transaction process rule base and store various verification rules; The real-time stream processing module (302) processes transaction process data in real time based on the Flink stream processing engine; The compliance filtering module (303) is used to execute the verification rules generated by the knowledge graph, perform compliance filtering on the transaction process, generate log records for intercepted abnormal transactions, including transaction ID, violation type, triggering rules, and push warning information to the supervision platform.

7. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 1 is characterized in that: The trusted transaction verification unit (400) includes a price rationality verification module (401), a correlation analysis module (402), a secondary verification module (403), and a blockchain evidence storage module (404), wherein: The price rationality verification module (401) is used to identify abnormal transactions that deviate from the normal price range by ≥2σ using a weighted isolation forest algorithm; The association analysis module (402) is used to calculate the similarity of bids between enterprises through GCN, and mark the bid as suspected rigging when the similarity is greater than 0.9; The secondary verification module (403) is used to perform secondary verification on transactions that have passed the real-time verification; The blockchain evidence storage module (404) is used to write the hash of the bidding documents and key data of the deposit certificate of the compliant transaction into the node on the Hyperledger Fabric chain.

8. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 7 is characterized in that: The association analysis module (402) calculates the similarity of bids between enterprises through GCN. When the similarity is greater than 0.9, it is marked as suspected bid rigging. The specific operation is as follows: B1. Constructing an enterprise association graph: Using bidding enterprises as nodes and equity relationships and legal representative overlap as edges, we construct the initial graph structure. B2. Graph convolution feature extraction: Node features are iteratively updated through the GCN model. The formula for updating the k-th layer features is: in, is an adjacency matrix with self-connection, is the degree matrix, W (k) is the trainable weight; B3. Similarity matrix calculation: For the final layer node feature H (K) Calculate the cosine similarity matrix S, where: B4. Determination of bid-rigging behavior: When S ij When >0.9, enterprise i and enterprise j are marked as suspected bid-rigging entities.

9. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 7 is characterized in that: The blockchain evidence storage module (404) writes the key data of the bidding document hash and the deposit certificate of the compliant transaction into the Hyperledger Fabric chain node. The specific operation is as follows: C1: Preprocessing of evidence data: Extract key fields such as the bid document SHA-256 hash value, security deposit certificate number, and timestamp from verified transaction data, and generate an evidence storage request in JSON format; C2: Smart contract trigger: Calls the Hyperledger Fabric chain code to verify whether the evidence storage request meets the following conditions: a) The transaction has passed the secondary verification of the trusted transaction verification unit (400); b) the compliance status output by the real-time verification unit (300) is "passed"; C3: Multi-node consensus: Based on the preset endorsement strategy, the evidence data is packaged to generate a new block; C4: On-chain storage: Write the block to the LevelDB database of each node and return the evidence certificate including transaction ID, block height, and Merkle tree root hash.

10. The land transaction dynamic monitoring and analysis system based on multi-source data fusion according to claim 1 is characterized in that: The dynamic monitoring and decision support unit (500) includes a market heat calculation module (501), an idle warning module (502), and a visualization module (503), wherein: The market heat calculation module (501) generates a regional heat index based on the bidding participation rate × average premium rate, wherein the bidding participation rate = the number of actual bidding participating enterprises / the number of potential participating enterprises × 100%, Average premium rate = ∑(land transaction price - starting price) / ∑ starting price; The idle warning module (502) is used to compare the development time nodes agreed in the land transfer contract. If the construction fails to start within 6 months after the deadline, a red warning will be triggered. If the construction fails to start within 3 months after the deadline, a yellow warning will be triggered. The visualization module (503) is used to display regional heat distribution and idle land locations through GIS maps, and supports drilling to view related enterprise information.

Images