Land registration data management methods, devices, equipment and storage media

Abstract

This invention relates to the field of cadastral data management technology, and discloses a cadastral data management method, apparatus, device, and storage medium. By acquiring multi-source operation data from the business windows of a cadastral data system, and utilizing a Long Short-Term Memory (LSTM) network based on hardware verification data and standardized temporal operation sequences, a temporal anomaly score is calculated. Considering the attention weight at each operation moment, an enhanced anomaly score with fused attention weights is obtained. Finally, XGBoost is used to consider the impact of sensitive operations while employing a dynamic threshold judgment method to achieve alarms for cadastral data system business window operation anomalies under different states in real-world scenarios. Therefore, this invention improves the accuracy and scenario adaptability of identifying business operation anomalies in cadastral data system business windows, ensuring the accuracy, integrity, and security of cadastral data lifecycle management activities, and supporting the compliant and efficient operation of core cadastral businesses such as land ownership registration, ownership change, and land planning.

Description

Technical Field

[0001] This invention relates to the field of cadastral data management technology, and in particular to a cadastral data management method, apparatus, equipment and storage medium. Background Technology

[0002] Cadastral data management refers to the full lifecycle management activities centered around cadastral data (covering core information such as land ownership, location, use, boundaries, area, and related business operation records). Its purpose is to ensure the accuracy, integrity, and security of cadastral data, and to support the compliant and efficient operation of core cadastral businesses such as land ownership registration, ownership change, and land planning.

[0003] As the core foundational data for land management, cadastral data's accuracy and security directly impact the compliance of government services such as land ownership registration, ownership transfer, and planning approval, as well as the public's land ownership rights. With the informatization upgrade of cadastral data system service windows, the frequency of business operations has significantly increased, and operational scenarios have become increasingly complex (such as multi-terminal login and batch work order processing). However, this also brings risks of abnormal operations such as unauthorized modifications, cross-authority access, and data entry errors. If such operations are not identified in a timely manner, they may lead to cadastral data distortion, leakage of sensitive information, and consequently, property rights disputes, administrative compliance risks, and even disruption of land management order.

[0004] To address the aforementioned risks, existing technologies have attempted to identify abnormal operations through operation log analysis and simple time-series models (such as ordinary LSTM). However, due to limitations such as single data sources, lack of hardware risk association in time-series analysis, lack of business-specific attention mechanisms, and static threshold fixation, these technologies struggle to cover complex scenarios such as illegal terminal operations, sensitive operations involving mismatched land parcels, and misjudgments during peak business periods. Consequently, the accuracy and scenario adaptability of business operation anomaly identification in the cadastral data system's business window are insufficient, failing to meet the cadastral data system's requirement for highly reliable anomaly monitoring.

[0005] Therefore, improving the accuracy and scenario adaptability of identifying abnormal business operations in the cadastral data system's business window is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] The present invention provides a cadastral data management method, apparatus, device and storage medium, which aims to solve at least one of the above-mentioned technical problems.

[0007] To achieve the above objectives, the present invention provides a cadastral data management method, comprising the following steps:

[0008] Obtain multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set;

[0009] Based on the standardized time-series operation sequences and hardware verification data in the aforementioned business operation feature set, the Long Short-Term Memory (LSTM) network is used to map and calculate the output time-series anomaly score according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequences.

[0010] Based on the business association features in the business operation feature set, the attention weight for each operation moment is calculated, and the attention weight is fused with the temporal anomaly score to obtain the enhanced anomaly score.

[0011] Using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, an XGBoost input feature is constructed. The XGBoost input feature is then input into an XGBoost binary classification model to obtain an XGBoost prediction score.

[0012] Based on the XGBoost prediction score and dynamic threshold, an alarm signal for abnormal business operation is generated in the business window of the cadastral data system, and abnormal business operation processing is performed.

[0013] Optionally, the steps of acquiring multi-source operation data from the cadastral data system's business window, preprocessing the multi-source operation data, and constructing a business operation feature set specifically include:

[0014] Obtain multi-source operation data from the business window of the land registration data system; wherein, the multi-source operation data includes operation log data, business work order data, facial recognition attendance data, and terminal registration data;

[0015] Extract the operational features of several operational data items from the operational log data, and construct a time-series feature sequence based on the operational features. And calculate the percentage of sensitive operations. Extract business association features from several business data items in the business work order data. The first hardware verification feature in facial recognition attendance data The second hardware verification feature in the terminal filing data and location compliance ;

[0016] Merging time series feature sequences Sensitive operation ratio Business-related characteristics First hardware verification feature Second hardware verification feature and location compliance Construct a business operation feature set .

[0017] Optionally, based on the standardized time-series operation sequences and hardware verification data in the business operation feature set, and utilizing a Long Short-Term Memory (LSTM) network, the step of mapping and calculating the output time-series anomaly score according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequences specifically includes:

[0018] Based on the aforementioned business operation feature set Hardware verification data and the proportion of computationally sensitive operations Calculate the initial cell state and initial hidden state ;

[0019] Among them, the initial cell state and initial hidden state The expression is as follows: ; ;

[0020] ;

[0021] ;

[0022] In the formula, Based on the initial cell state, Hardware risk labeling To preset the risk adjustment coefficient, These are the weighting coefficients. The threshold for the proportion of sensitive operations corresponding to hardware risks;

[0023] Based on the calculated initial cell state and initial hidden state Standardized time-series operation sequences in the aforementioned business operation feature set Perform forward propagation temporal feature extraction on a Long Short-Term Memory (LSTM) network, calculate the gating and state of the LSTM network step by step, and obtain the hidden states of the entire sequence. and final hidden state ;

[0024] Specifically, the gating and state of the Long Short-Term Memory (LSTM) network are calculated step-by-step: ;

[0025] ;

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] In the formula, , , , These are the gate weight parameters for the Long Short-Term Memory (LSTM) network. , , , These are the gating bias parameters for the Long Short-Term Memory (LSTM) network. It is the sigmoid activation function. Represents the forgetting gate, x t This represents the operational characteristics at time t. Indicates the input gate. Indicates the state of candidate cells. Indicates cell state update. This represents element-wise multiplication of vectors. Indicates the output gate. This represents the hidden state, and t represents time.

[0031] The temporal anomaly score is obtained by performing a linear transformation normalization calculation using the final hidden state. With time series anomaly level The expression is as follows: ;

[0032] ;

[0033] In the formula, The score is for timing anomalies. The time series anomaly level is: , This represents the time series anomaly score threshold corresponding to the time series anomaly level.

[0034] Optionally, based on the business association features in the business operation feature set, an attention weight is calculated for each operation moment, and the attention weight is fused with the temporal anomaly score to obtain an enhanced anomaly score. This step specifically includes:

[0035] Based on the aforementioned business operation feature set Business association characteristics Land parcel matching degree Sensitive weight matrix determined by the pre-set cadastral data business operation risk registration form Calculate the attention weights at each operation time t. The expression is as follows: ;

[0036] In the formula, X represents the operational characteristic at each operation time t. t In the sensitive weight matrix The corresponding sensitive weights, This indicates the penalty coefficient for business mismatch;

[0037] Attention weights obtained through calculation With the aforementioned time series anomaly score Fusion to obtain enhanced anomaly scores The expression is as follows:

[0038] ;

[0039] In the formula, Timing Anomaly Score The value in denoted by t represents the time series anomaly score at time t.

[0040] Optionally, the step of constructing XGBoost input features using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, and inputting the input features into the XGBoost binary classification model to obtain the XGBoost prediction score, specifically includes:

[0041] Using the enhanced anomaly score With the aforementioned business operation feature set Business Relationship Characteristics Sensitive operation ratio and location compliance Construct input features ;in, Indicates type matching degree, Indicates batch work order identifier;

[0042] Input features Input the data into the XGBoost binary classification model to obtain the initial XGBoost prediction score. Using enhanced anomaly scores Corresponding Enhanced Anomaly Level Preliminary prediction score Make corrections to obtain the final XGBoost prediction score. ;

[0043] Among them, the preliminary prediction score The expression to be corrected is as follows: ;

[0044] ;

[0045] In the formula, express Correction factor, Indicates enhanced anomaly level In the level of collaborative timing anomaly The threshold corresponding to the enhanced abnormal score A.

[0046] Optionally, based on the XGBoost prediction score and dynamic threshold, an alarm signal for abnormal business operations in the cadastral data system business window is generated, and abnormal business operation handling steps are executed, specifically including:

[0047] Query the business monitoring data of the cadastral data system, extract the business scenario parameters from the business monitoring data, normalize and encode them, and obtain the business scenario status. ;in, This indicates the parameters for determining peak periods. This represents the business volume parameter. This parameter represents the daily false positive rate.

[0048] Business scenario status Input a pre-trained DQN model, and use the basic threshold parameters and the DQN model to adjust the business scenario state. The output adjustment threshold parameter is used to calculate the dynamic threshold under the current business scenario. ;

[0049] Based on XGBoost predicted score Dynamic thresholds in the current business scenario The system generates an alarm signal for abnormal business operations in the cadastral data system's business window and performs abnormal business operation processing; wherein, the abnormal business operation processing includes emergency operations for the abnormal business window and manual review operations;

[0050] Obtain real-time manual review data from the cadastral data system and extract the hourly error rate from the manual review data. and hourly false negative rate Calculate the reward value R and feed the reward value R back to the DQN model;

[0051] The expression for calculating the reward value R is as follows:

[0052]

[0053] In the formula, and For penalty weights.

[0054] Optionally, based on XGBoost prediction score Dynamic thresholds in the current business scenario The steps for generating abnormal alarm signals for business operations in the cadastral data system business window include:

[0055] Based on XGBoost predicted score Dynamic thresholds in the current business scenario Calculate the threshold exceeding the limit. Based on the threshold exceeding the limit within a preset alarm level range, generate a business operation abnormality alarm signal for the cadastral data system business window. The specific expression is as follows:

[0056] ;

[0057] In the formula, For emergency alarms, the corresponding emergency actions for abnormal business windows include locking the abnormal business window; For regular alarms, the corresponding emergency actions for abnormal business windows include abnormal business window pop-up alarms; To alert users, the corresponding emergency actions for abnormal business windows include recording the business operation data of the abnormal business window.

[0058] Furthermore, to achieve the above objectives, the present invention also provides a cadastral data management device, comprising:

[0059] The construction module is used to acquire multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set.

[0060] The calculation module is used to calculate and output a timing anomaly score based on the standardized time-series operation sequence and hardware verification data in the business operation feature set, using a long short-term memory network (LSTM), and based on the initial cell state determined by the hardware verification data and the timing features extracted from the standardized time-series operation sequence.

[0061] The fusion module is used to calculate the attention weight for each operation moment based on the business association features in the business operation feature set, and fuse the attention weight with the time-series anomaly score to obtain the enhanced anomaly score.

[0062] The prediction module is used to construct XGBoost input features by utilizing the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, and input the XGBoost input features into the XGBoost binary classification model to obtain the XGBoost prediction score.

[0063] The processing module is used to generate an alarm signal for abnormal business operations in the cadastral data system business window based on the XGBoost prediction score and dynamic threshold, and to perform abnormal business operation processing.

[0064] In addition, to achieve the above objectives, the present invention also provides a cadastral data management device, the cadastral data management device comprising: a memory, a processor, and a cadastral data management program stored in the memory and executable on the processor, wherein the cadastral data management program, when executed by the processor, implements the steps of the cadastral data management method as described above.

[0065] In addition, to achieve the above objectives, the present invention also provides a storage medium storing a cadastral data management program, which, when executed by a processor, implements the steps of the above-described cadastral data management method.

[0066] The beneficial effects of this invention are as follows: It proposes a cadastral data management method, apparatus, device, and storage medium. By acquiring multi-source operation data of the cadastral data system's business window, it utilizes a Long Short-Term Memory (LSTM) network to calculate a timing anomaly score based on the initial cell state determined by hardware verification data and the timing features extracted from the standardized timing operation sequence. This score reflects the impact of hardware risks on the timing of operations. Considering the attention weight at each operation moment, it obtains an enhanced anomaly score that integrates the attention weight and the timing anomaly score, highlighting the impact of high-risk operation moments. Finally, using XGBoost and dynamic thresholds, while considering the impact of sensitive operations, it adopts a dynamic threshold judgment method to achieve alarms for cadastral data system business window operations under different states in actual scenarios. Therefore, this invention adopts a multi-source data fusion and hardware risk binding approach to achieve full coverage of abnormal feature extraction and analysis across risk dimensions. By associating business with attention and collaboratively modeling multi-dimensional features, it accurately calculates abnormal scores. Finally, it uses scenario-adaptive dynamic thresholds to achieve tiered alarms. This improves the accuracy and scenario adaptability of business operation anomaly identification in the cadastral data system's business window, ensuring the accuracy, integrity, and security of cadastral data, and supporting the compliant and efficient operation of core cadastral businesses such as land ownership registration, ownership change, and land planning. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention;

[0068] Figure 2 This is a flowchart illustrating an embodiment of the cadastral data management method of the present invention;

[0069] Figure 3 This is a structural block diagram of a cadastral data management device according to an embodiment of the present invention. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0072] like Figure 1 As shown, Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention.

[0073] like Figure 1 As shown, the device may include: a processor 1001, such as a CPU; a communication bus 1002; a user interface 1003; a network interface 1004; and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0074] Those skilled in the art will understand that Figure 1 The structure of the device shown does not constitute a limitation on the device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0075] like Figure 1 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a cadastral data management program.

[0076] exist Figure 1 In the terminal shown, network interface 1004 is mainly used to connect to the backend server and communicate with it; user interface 1003 is mainly used to connect to the client (user terminal) and communicate with it; while processor 1001 can be used to call the cadastral data management program stored in memory 1005 and perform the following operations:

[0077] Obtain multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set;

[0078] Based on the standardized time-series operation sequences and hardware verification data in the aforementioned business operation feature set, the Long Short-Term Memory (LSTM) network is used to map and calculate the output time-series anomaly score according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequences.

[0079] Based on the business association features in the business operation feature set, the attention weight for each operation moment is calculated, and the attention weight is fused with the temporal anomaly score to obtain the enhanced anomaly score.

[0080] Using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, an XGBoost input feature is constructed. The XGBoost input feature is then input into an XGBoost binary classification model to obtain an XGBoost prediction score.

[0081] Based on the XGBoost prediction score and dynamic threshold, an alarm signal for abnormal business operation is generated in the business window of the cadastral data system, and abnormal business operation processing is performed.

[0082] The specific embodiments of the present invention applied to the device are basically the same as the embodiments of the cadastral data management method described below, and will not be repeated here.

[0083] This invention provides a cadastral data management method, referring to... Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the cadastral data management method of the present invention.

[0084] In this embodiment, a cadastral data management method includes the following steps:

[0085] S100: Obtain multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set;

[0086] S200: Based on the standardized time-series operation sequence and hardware verification data in the business operation feature set, the Long Short-Term Memory (LSTM) network is used to map and calculate the output time-series anomaly score according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequence.

[0087] S300: Based on the business association features in the business operation feature set, calculate the attention weight for each operation moment, and fuse the attention weight with the time-series anomaly score to obtain the enhanced anomaly score;

[0088] S400: Using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, construct XGBoost input features, input the XGBoost input features into the XGBoost binary classification model, and obtain the XGBoost prediction score.

[0089] S500: Based on the XGBoost prediction score and dynamic threshold, generate an alarm signal for abnormal business operation in the cadastral data system business window and perform abnormal business operation processing.

[0090] It should be noted that cadastral data, as the core foundational data for land management, directly impacts the compliance of government services such as land ownership registration, ownership changes, and planning approvals, as well as the public's land ownership rights. With the informatization upgrade of cadastral data system business windows, the frequency of business operations has significantly increased, and operational scenarios have become increasingly complex (such as multi-terminal login and batch work order processing). However, this also brings risks of abnormal operations such as unauthorized modifications, cross-permission access, and data entry errors. If such operations are not identified in a timely manner, they may lead to cadastral data distortion, leakage of sensitive information, and consequently, property disputes, administrative compliance risks, and even disruption of land management order. To address these risks, existing technologies have attempted to identify abnormal operations through operation log analysis and simple time-series models (such as ordinary LSTM). However, limitations such as single data sources, lack of hardware risk correlation in time-series analysis, lack of business-specific attention mechanisms, and static threshold fixation make it difficult to cover complex scenarios such as illegal terminal operations, sensitive operations involving mismatched land parcels, and misjudgments during peak business periods. This results in insufficient accuracy and scenario adaptability in identifying abnormal business operations at the cadastral data system business windows, failing to meet the cadastral data system's requirement for highly reliable anomaly monitoring.

[0091] To address the aforementioned issues, this embodiment acquires multi-source operation data from the cadastral data system's business window. Utilizing a Long Short-Term Memory (LSTM) network, based on hardware verification data and standardized temporal operation sequences, it calculates temporal anomaly scores. Considering the attention weights at each operation moment, it obtains an enhanced anomaly score with fused attention weights. Finally, XGBoost is used to consider the impact of sensitive operations while employing a dynamic threshold judgment method to implement alarms for cadastral data system business window operation anomalies under different states in real-world scenarios. Therefore, this invention, based on a full-process design encompassing multi-source data preprocessing, temporal feature extraction, attention fusion, classification prediction, and dynamic alarms, achieves collaborative analysis of multi-dimensional risk information, avoiding the limitations of single models or single features. This improves the accuracy and scenario adaptability of identifying business operation anomalies in the cadastral data system's business window, ensuring the accuracy, integrity, and security of cadastral data, and supporting the compliant and efficient operation of core cadastral businesses such as land ownership registration, ownership changes, and land planning.

[0092] In a preferred embodiment, the steps of acquiring multi-source operation data from the cadastral data system's business window, preprocessing the multi-source operation data, and constructing a business operation feature set specifically include:

[0093] S110: Obtain multi-source operation data from the business window of the land registration data system; wherein, the multi-source operation data includes operation log data, business work order data, facial recognition attendance data, and terminal registration data;

[0094] S120: Extract operational features from several operational data items in the operational log data, and construct a time-series feature sequence based on the operational features. And calculate the percentage of sensitive operations. Extract business association features from several business data items in the business work order data. The first hardware verification feature in facial recognition attendance data The second hardware verification feature in the terminal filing data and location compliance ;

[0095] S130: Merging Temporal Feature Sequences Sensitive operation ratio Business-related characteristics First hardware verification feature Second hardware verification feature and location compliance Construct a business operation feature set .

[0096] In this embodiment, firstly, multi-source operation data, including operation log data, business work order data, facial recognition attendance data, and terminal registration data, is acquired. Differential features are then extracted from different data sources: a time-series feature sequence is extracted from the operation log data. The proportion of sensitive operations Extract business-related features from business work orders Extracting the first hardware verification feature from facial recognition attendance data Extracting the second hardware verification feature from the terminal registration Location compliance Finally, all extracted features are merged to form a complete set of business operation features. .

[0097] For example, in the preprocessing stage: ① Extract the "sequence of land parcel modification and query operations from 9:00 to 10:00 on May 20, 2024" from the operation log to form a time-series feature sequence. Simultaneously, the percentage of sensitive operations such as "modifying ownership information" was calculated to be 35% of all operations during this period; ② The "matching degree between land parcel number and work order filing number, matching degree between acceptance type and operation type, and batch work order identifier" were extracted from the business work orders to form the business association feature B=[0.8,0.9,1] (1 represents a batch work order); ③ The "matching similarity between the operator's face and the filed face" was extracted from the facial attendance data to obtain the first hardware verification feature. =0.98; ④ Extract the "consistency between the operating terminal number and the registration number" from the terminal registration data ( =1, representing consistency) and "consistency between the location of the operation terminal and the location of attendance" (L=1, representing consistency); ⑤ Finally, merge all features to obtain the business operation feature set X=[ [35%, B, 0.98, 1, 1] provides a complete data foundation for subsequent anomaly analysis.

[0098] Therefore, this embodiment breaks through the limitations of traditional single operation logs, integrates multi-dimensional data of business, hardware, and location, and ensures that the feature set can characterize the temporal continuity of operations, business compliance, hardware legality, and location rationality. It extracts exclusive features for the attributes of different data sources (such as the temporality of operation logs and the hardware attributes of terminal registration), avoids feature redundancy or missing features, and forms a unified feature set through structured merging, providing standardized input for subsequent LSTM time series analysis and XGBoost classification, and reducing the complexity of data preprocessing.

[0099] In a preferred embodiment, based on the standardized time-series operation sequences and hardware verification data in the business operation feature set, and utilizing a Long Short-Term Memory (LSTM) network, the step of mapping and calculating the output time-series anomaly score according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequences specifically includes:

[0100] S210: Based on the aforementioned business operation feature set Hardware verification data and the proportion of computationally sensitive operations Calculate the initial cell state and initial hidden state ;

[0101] Among them, the initial cell state and initial hidden state The expression is as follows: ;

[0102] ;

[0103] ;

[0104] ;

[0105] In the formula, Based on the initial cell state, Hardware risk labeling To preset the risk adjustment coefficient, These are the weighting coefficients. The threshold for the proportion of sensitive operations corresponding to hardware risks;

[0106] S220: Initial cell state obtained from calculation and initial hidden state Standardized time-series operation sequences in the aforementioned business operation feature set Perform forward propagation temporal feature extraction on a Long Short-Term Memory (LSTM) network, calculate the gating and state of the LSTM network step by step, and obtain the hidden states of the entire sequence. and final hidden state ;

[0107] Specifically, the gating and state of the Long Short-Term Memory (LSTM) network are calculated step-by-step: ;

[0108] ;

[0109] ;

[0110] ;

[0111] ;

[0112] ;

[0113] In the formula, , , , These are the gate weight parameters for the Long Short-Term Memory (LSTM) network. , , , These are the gating bias parameters for the Long Short-Term Memory (LSTM) network. It is the sigmoid activation function. Represents the forgetting gate, x t This represents the operational characteristics at time t. Indicates the input gate. Indicates the state of candidate cells. Indicates cell state update. This represents element-wise multiplication of vectors. Indicates the output gate. This represents the hidden state, and t represents time.

[0114] S230: Perform a linear transformation to normalize the score calculation using the final hidden state to obtain the temporal anomaly score. With time series anomaly level The expression is as follows:

[0115] ;

[0116] ;

[0117] In the formula, The score is for timing anomalies. The time series anomaly level is: , This represents the time series anomaly score threshold corresponding to the time series anomaly level.

[0118] In this embodiment, the initial state of the LSTM is adjusted by hardware verification data and the proportion of sensitive operations, so that the model can prioritize the detection of illegal terminal operations. =0), Location not compliant ( =0), High-sensitivity operation ratio ( To mitigate hardware-level risks, improve the targeting of time series analysis, and leverage the gating mechanism of LSTM (forget gate controls the retention of historical states, input gate controls the input of current features), effectively capture long-term dependencies in operation timing (such as the normal logic of "query first, then modify" and the abnormal logic of "direct modification without query"). At the same time, it can also classify abnormal levels by threshold, providing a clear risk gradient for subsequent attention fusion and alarm classification.

[0119] In a preferred embodiment, the step of calculating the attention weight for each operation moment based on the business association features in the business operation feature set, and fusing the attention weight with the temporal anomaly score to obtain the enhanced anomaly score, specifically includes:

[0120] S310: Based on the aforementioned business operation feature set Business association characteristics Land parcel matching degree Sensitive weight matrix determined by the pre-set cadastral data business operation risk registration form Calculate the attention weights at each operation time t. The expression is as follows:

[0121] ;

[0122] In the formula, X represents the operational characteristic at each operation time t. tIn the sensitive weight matrix The corresponding sensitive weights, This indicates the penalty coefficient for business mismatch;

[0123] S320: Attention weights obtained through calculation With the aforementioned time series anomaly score Fusion to obtain enhanced anomaly scores The expression is as follows:

[0124] ;

[0125] In the formula, Timing Anomaly Score The value in denoted by t represents the time series anomaly score at time t.

[0126] In this embodiment, the matching degree of land parcels is used. With penalty coefficient This configuration prioritizes attention weights for operations involving mismatched parcels (e.g., modifying parcel B data with a work order for parcel A), addressing the shortcomings of traditional time-series analysis that only considers the order of operations without considering business logic. It also allows for the configuration of high-risk moments (such as...) (low, high sensitivity weight) Larger, its corresponding It accounts for a higher proportion during fusion, thus enhancing the anomaly score. It can accurately reflect the combined risks of temporal anomalies and business anomalies. Simultaneously, it utilizes a pre-set sensitivity weight matrix based on the cadastral data business operation risk registration form. This ensures that attention calculations align with the actual business risk definition (e.g., pre-setting a higher sensitivity weight for "ownership modification" than for "normal queries").

[0127] In a preferred embodiment, the step of constructing XGBoost input features using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, and inputting the input features into the XGBoost binary classification model to obtain the XGBoost prediction score, specifically includes:

[0128] S410: Utilizing the enhanced anomaly score With the aforementioned business operation feature set Business Relationship Characteristics Sensitive operation ratio and location compliance Construct input features ;in, Indicates type matching degree, Indicates batch work order identifier;

[0129] S420: Input features Input the data into the XGBoost binary classification model to obtain the initial XGBoost prediction score. Using enhanced anomaly scores Corresponding Enhanced Anomaly Level Preliminary prediction score Make corrections to obtain the final XGBoost prediction score. ;

[0130] Among them, the preliminary prediction score The expression to be corrected is as follows:

[0131] ;

[0132] ;

[0133] In the formula, express Correction factor, Indicates enhanced anomaly level In the level of collaborative timing anomaly The threshold corresponding to the enhanced abnormal score A.

[0134] In this embodiment, input features covering business association (B), hardware location (L), enhanced anomaly score (A), and sensitivity percentage (S) are used. The construction of this system overcomes the limitations of traditional XGBoost, which relies solely on static features. It enables classification prediction to integrate multi-dimensional risks and corrects initial scores by configuring enhanced anomaly levels. This strengthens anomaly signals in high-risk scenarios while suppressing misjudgments of low-series anomalies and single-risk factors, thus improving the accuracy of anomaly classification. Furthermore, the introduction of batch work order identifiers in business-related features identifies situations in cadastral business where batch work orders easily trigger large-scale anomalies, making anomaly detection more accurate.

[0135] In a preferred embodiment, based on the XGBoost prediction score and dynamic threshold, an alarm signal for abnormal business operations in the cadastral data system business window is generated, and abnormal business operation processing steps are executed, specifically including:

[0136] S510: Query the business monitoring data of the cadastral data system, extract the business scenario parameters from the business monitoring data, normalize and encode them, and obtain the business scenario status. ;in, This indicates the parameters for determining peak periods. This represents the business volume parameter. This parameter represents the daily false positive rate.

[0137] S520: Transform the business scenario status Input a pre-trained DQN model, and use the basic threshold parameters and the DQN model to adjust the business scenario state. The output adjustment threshold parameter is used to calculate the dynamic threshold under the current business scenario. ;

[0138] S530: Score based on XGBoost prediction Dynamic thresholds in the current business scenario The system generates an alarm signal for abnormal business operations in the cadastral data system's business window and performs abnormal business operation processing; wherein, the abnormal business operation processing includes emergency operations for the abnormal business window and manual review operations;

[0139] S540: Obtain real-time manual review data from the cadastral data system and extract the hourly error rate from the manual review data. and hourly false negative rate Calculate the reward value R and feed the reward value R back to the DQN model;

[0140] The expression for calculating the reward value R is as follows:

[0141] ;

[0142] In the formula, and For penalty weights.

[0143] In this embodiment, business monitoring data from the cadastral data system is used to normalize and encode peak period determination parameters, business volume parameters, and daily misjudgment rate parameters to obtain the business scenario status. , business scenario status Input a pre-trained DQN (Deep Q-Network) model and combine it with basic threshold parameters to calculate the dynamic threshold for the current scene. By utilizing reinforcement learning with the DQN model, the accuracy of alarms in different scenarios is improved. Finally, the accuracy is determined based on the XGBoost prediction score and dynamic threshold. It generates alarm signals and performs anomaly handling (such as window locking, pop-up alarms, and data logging) to balance risk control and business efficiency. Simultaneously, for manually reviewed data, it calculates reward values ​​and feeds them back to the DQN model to optimize its threshold adjustment strategy, forming a closed loop of scene awareness, dynamic thresholds, alarm handling, and feedback optimization, thereby improving the adaptability of dynamic thresholds in the long term.

[0144] In a preferred embodiment, the prediction score is based on XGBoost. Dynamic thresholds in the current business scenario Generate alarm signals for abnormal business operations in the cadastral data system business window, specifically including:

[0145] S531: Score based on XGBoost prediction Dynamic thresholds in the current business scenario Calculate the threshold exceeding the limit. Based on the threshold exceeding the limit within a preset alarm level range, generate a business operation abnormality alarm signal for the cadastral data system business window. The specific expression is as follows:

[0146] ;

[0147] In the formula, For emergency alarms, the corresponding emergency actions for abnormal business windows include locking the abnormal business window; For regular alarms, the corresponding emergency actions for abnormal business windows include abnormal business window pop-up alarms; To alert users, the corresponding emergency actions for abnormal business windows include recording the business operation data of the abnormal business window.

[0148] In this embodiment, the XGBoost prediction score is used. Dynamic thresholds in the current business scenario The calculation threshold exceeds the limit, and provides different levels of alarms, including window locking, pop-up alarms, and data logging, to avoid response confusion caused by ambiguity in the degree of anomaly. This gives the alarm signal a clear risk gradient, which can effectively control risks and avoid affecting normal business due to excessive intervention.

[0149] Therefore, this embodiment adopts a multi-source data fusion and hardware risk binding approach to achieve full coverage of abnormal feature extraction and analysis across risk dimensions. By associating business with attention and collaboratively modeling multi-dimensional features, it accurately calculates abnormal scores. Finally, it uses scenario-adaptive dynamic thresholds to achieve tiered alarms. This approach can improve the accuracy and scenario adaptability of business operation anomaly identification in the cadastral data system's business window, ensuring the accuracy, integrity, and security of cadastral data, and supporting the compliant and efficient operation of core cadastral businesses such as land ownership registration, ownership change, and land planning.

[0150] Reference Figure 3 , Figure 3 This is a structural block diagram of an embodiment of the cadastral data management device of the present invention.

[0151] like Figure 3 As shown, the cadastral data management device proposed in this embodiment of the invention includes:

[0152] Module 10 is used to acquire multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set.

[0153] The calculation module 20 is used to calculate and output a timing anomaly score based on the standardized time-series operation sequence and hardware verification data in the business operation feature set, using a long short-term memory network LSTM, and based on the initial cell state determined by the hardware verification data and the timing features extracted from the standardized time-series operation sequence.

[0154] The fusion module 30 is used to calculate the attention weight for each operation moment based on the business association features in the business operation feature set, and fuse the attention weight with the time-series anomaly score to obtain the enhanced anomaly score.

[0155] Prediction module 40 is used to construct XGBoost input features by utilizing the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, and input the XGBoost input features into the XGBoost binary classification model to obtain the XGBoost prediction score.

[0156] The processing module 50 is used to generate an alarm signal for abnormal business operation of the cadastral data system business window based on the XGBoost prediction score and dynamic threshold, and to perform abnormal business operation processing.

[0157] Other embodiments or specific implementations of the cadastral data management device of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0158] Furthermore, the present invention also proposes a cadastral data management device, which includes: a memory, a processor, and a cadastral data management program stored in the memory and executable on the processor. When the cadastral data management program is executed by the processor, it implements the steps of the cadastral data management method as described above.

[0159] The specific implementation method of the cadastral data management device in this application is basically the same as the various embodiments of the above-mentioned cadastral data management method, and will not be repeated here.

[0160] Furthermore, this invention also proposes a readable storage medium, which includes a computer-readable storage medium storing a cadastral data management program thereon. The readable storage medium may be... Figure 1 The memory 1005 in the terminal can also be at least one of ROM (Read-Only Memory) / RAM (Random Access Memory), magnetic disk, optical disk, etc. The readable storage medium includes several instructions to cause a cadastral data management device with a processor to execute the cadastral data management method described in various embodiments of the present invention.

[0161] The specific implementation methods in the readable storage medium of this application are basically the same as those in the above-described embodiments of the cadastral data management method, and will not be repeated here.

[0162] It is understood that in the description of this specification, references to terms such as "one embodiment," "another embodiment," "other embodiments," or "first embodiment to Nth embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above 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 one or more embodiments or examples.

[0163] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0164] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0165] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0166] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for managing cadastral data, characterized in that, Includes the following steps: Acquire multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set; specifically including: Obtain multi-source operation data from the business window of the land registration data system; wherein, the multi-source operation data includes operation log data, business work order data, facial recognition attendance data, and terminal registration data; Extract the operational features of several operational data items from the operational log data, and construct a time-series feature sequence X based on the operational features. seq And calculate the proportion of sensitive operations S, extract the business association features B of several business data items in the business work order data, the first hardware verification feature H1 in the face attendance data, the second hardware verification feature H2 in the terminal filing data, and the location compliance L; Merging time series feature sequences X seq Based on the following factors: the proportion of sensitive operations (S), business-related features (B), first hardware verification feature (H1), second hardware verification feature (H2), and location compliance (L), a business operation feature set X = [X...] is constructed. seq [,S,B,H1,H2,L]; Based on the standardized time-series operation sequences and hardware verification data in the aforementioned business operation feature set, the Long Short-Term Memory (LSTM) network is used to map and calculate the output time-series anomaly score according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequences. Based on the business association features in the business operation feature set, the attention weight for each operation moment is calculated, and the attention weight is fused with the temporal anomaly score to obtain the enhanced anomaly score. Using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, an XGBoost input feature is constructed. The XGBoost input feature is then input into an XGBoost binary classification model to obtain an XGBoost prediction score. Based on the XGBoost prediction score and dynamic threshold, an alarm signal for abnormal business operation is generated in the cadastral data system business window, and abnormal business operation processing is performed.

2. The cadastral data management method as described in claim 1, characterized in that, Based on the standardized time-series operation sequences and hardware verification data in the aforementioned business operation feature set, and utilizing a Long Short-Term Memory (LSTM) network, the step of mapping and calculating the output time-series anomaly score is performed according to the initial cell state determined by the hardware verification data and the time-series features extracted from the standardized time-series operation sequences. Specifically, this includes: Based on the hardware verification data H = [H1, H2, L] in the business operation feature set X and the sensitive operation ratio S, calculate the initial cell state C0 and the initial hidden state h0. The expressions for the initial cell state C0 and the initial hidden state h0 are as follows: h0 = tanh C0 In the formula, Based on the initial cell state, Flag is the hardware risk marker, μ is the preset risk adjustment coefficient, α, β, and γ are weighting coefficients, and S0 is the threshold for the proportion of sensitive operations corresponding to hardware risks. Based on the calculated initial cell state C0 and initial hidden state h0, and the standardized temporal operation sequence X in the business operation feature set... seq =[x1,x2,···,x T The forward propagation temporal feature extraction of the Long Short-Term Memory (LSTM) network is performed, and the gating and state of the LSTM network are calculated step by step to obtain the hidden state H of the entire sequence. seq =[h1,···,h T ] and the final hidden state h T ; Specifically, the gating and state of the Long Short-Term Memory (LSTM) network are calculated step-by-step: f t =σ(W f ·[h t-1 ;x t ]+b f ) i t =σ(W i ·[h t-1 ;x t ]+b i ) the t =σ(W o ·[h t-1 ;x t ]+b o ) h t =o t ⊙tanhC t In the formula, W f W i W C W o b represents the gating weights of a Long Short-Term Memory (LSTM) network. f b i b C b o Here, σ represents the gating bias parameters of the Long Short-Term Memory (LSTM) network, σ is the sigmoid activation function, and f... t Represents the forgetting gate, x t Indicates the operational characteristics at time t, i t Indicates the input gate. Indicates the candidate cell state, C t Represents cell state update, ⊙ represents vector element multiplication, o t Indicates the output gate, h t This represents the hidden state, and t represents time. By performing a linear transformation to normalize the score using the final hidden state, the temporal anomaly score A0 and the temporal anomaly level Level are obtained. seq The expression is as follows: A0=sigmoid(W out ·h t +b out ) In the formula, A0 is the time series anomaly score, and Level seq A1 and A2 are the time series anomaly levels, and A1 and A2 are the time series anomaly score thresholds corresponding to the time series anomaly levels.

3. The cadastral data management method as described in claim 1, characterized in that, Based on the business association features in the business operation feature set, the attention weight for each operation moment is calculated, and the attention weight is fused with the temporal anomaly score to obtain the enhanced anomaly score. This process specifically includes: The sensitivity weight matrix W is determined based on the parcel matching degree b1 in the business association features B = [b1, b2, b3] of the business operation feature set X and the preset cadastral data business operation risk registration form. s =[w1,···,w k ] Calculate the attention weight α at each operation time t. t The expression is as follows: In the formula, The operational characteristic x represents the operation feature at each operation time t. t In the sensitive weight matrix W s The corresponding sensitive weights are defined in the text, where λ represents the penalty coefficient for business mismatch. The attention weight α obtained through calculation t The enhanced anomaly score A is obtained by fusing the time-series anomaly score A0 with the time-series anomaly score A, and the specific expression is as follows: In the formula, A 0,t The time series anomaly score A0 represents the time series anomaly score at time t.

4. The cadastral data management method as described in claim 1, characterized in that, Using the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, an XGBoost input feature is constructed. This input feature is then fed into an XGBoost binary classification model to obtain the XGBoost prediction score. The specific steps include: Using the enhanced anomaly score A and the business-related features B = [b1, b2, b3], the proportion of sensitive operations S, and the location compliance L in the business operation feature set X, input feature X is constructed. XGB = [B,L,A,S]; where b2 represents the type matching degree and b3 represents the batch work order identifier; Input features X XGB Inputting [B,L,A,S] into the XGBoost binary classification model yields the initial XGBoost prediction score. Using the enhanced anomaly score A and the corresponding enhanced anomaly level Level enhance Preliminary prediction score Make corrections to obtain the final XGBoost prediction score. Among them, the preliminary prediction score The expression to be corrected is as follows: In the formula, μ y express The correction factor, μ A Indicates the enhanced anomaly level. enhance In the level of coordinating timing anomalies seq The threshold corresponding to the enhanced abnormal score A.

5. The cadastral data management method as described in claim 1, characterized in that, Based on the XGBoost prediction score and dynamic threshold, an alarm signal for abnormal business operations is generated for the cadastral data system business window, and abnormal business operation handling steps are executed, specifically including: Query the business monitoring data of the cadastral data system, extract the business scenario parameters from the business monitoring data, normalize and encode them, and obtain the business scenario status S. dqn = [S1,S2,S3]; where S1 represents the peak period determination parameter, S2 represents the business volume parameter, and S3 represents the daily misjudgment rate parameter; The business scenario state S dqn =[S1,S2,S3] Input a pre-trained DQN model, and use the basic threshold parameters and the DQN model according to the business scenario state S dqn The output adjustment threshold parameter is used to calculate the dynamic threshold τ under the current business scenario state. Based on XGBoost predicted score Based on the dynamic threshold τ under the current business scenario, an alarm signal for abnormal business operation is generated for the business window of the cadastral data system, and abnormal business operation processing is performed; wherein, the abnormal business operation processing includes emergency operation of abnormal business window and manual review operation; Obtain real-time manual review data from the cadastral data system and extract the hourly error rate η from the manual review data. 误判 and hourly missed detection rate η 漏判 Calculate the reward value R and feed the reward value R back to the DQN model; The expression for calculating the reward value R is as follows: R=1-μ1·h 误判 -m2·h 漏判 In the formula, μ1 and μ2 are penalty weights.

6. The cadastral data management method as described in claim 5, characterized in that, Based on XGBoost predicted score The steps for generating an alarm signal for abnormal business operations in the cadastral data system business window, based on a dynamic threshold τ under the current business scenario, specifically include: Based on XGBoost predicted score Based on the dynamic threshold τ under the current business scenario, calculate the threshold exceedance. If the threshold exceedance falls within a preset alarm level range, generate a business operation anomaly alarm signal for the cadastral data system business window. The specific expression is as follows: In the formula, c1 is an emergency alarm, and the corresponding emergency operation for the abnormal business window includes locking the abnormal business window; c2 is a normal alarm, and the corresponding emergency operation for the abnormal business window includes pop-up alarm for the abnormal business window; c3 is a prompt alarm, and the corresponding emergency operation for the abnormal business window includes recording the business operation data of the abnormal business window.

7. A cadastral data management device, characterized in that, include: The construction module is used to acquire multi-source operation data from the business window of the cadastral data system, preprocess the multi-source operation data, and construct a business operation feature set. Specifically, it includes: Obtain multi-source operation data from the business window of the land registration data system; wherein, the multi-source operation data includes operation log data, business work order data, facial recognition attendance data, and terminal registration data; Extract the operational features of several operational data items from the operational log data, and construct a time-series feature sequence X based on the operational features. seq And calculate the proportion of sensitive operations S, extract the business association features B of several business data items in the business work order data, the first hardware verification feature H1 in the face attendance data, the second hardware verification feature H2 in the terminal filing data, and the location compliance L; Merging time series feature sequences X seq Based on the following factors: the proportion of sensitive operations (S), business-related features (B), first hardware verification feature (H1), second hardware verification feature (H2), and location compliance (L), a business operation feature set X = [X...] is constructed. seq [,S,B,H1,H2,L]; The calculation module is used to calculate and output a timing anomaly score based on the standardized time-series operation sequence and hardware verification data in the business operation feature set, using a long short-term memory network (LSTM), and based on the initial cell state determined by the hardware verification data and the timing features extracted from the standardized time-series operation sequence. The fusion module is used to calculate the attention weight for each operation moment based on the business association features in the business operation feature set, and fuse the attention weight with the time-series anomaly score to obtain the enhanced anomaly score. The prediction module is used to construct XGBoost input features by utilizing the enhanced anomaly score and the proportion of sensitive operations in the business operation feature set, and input the XGBoost input features into the XGBoost binary classification model to obtain the XGBoost prediction score. The processing module is used to generate an alarm signal for abnormal business operations in the cadastral data system business window based on the XGBoost prediction score and dynamic threshold, and to perform abnormal business operation processing.

8. A cadastral data management device, characterized in that, The cadastral data management device includes: a memory, a processor, and a cadastral data management program stored in the memory and executable on the processor. When the cadastral data management program is executed by the processor, it implements the steps of the cadastral data management method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores a cadastral data management program, which, when executed by a processor, implements the steps of the cadastral data management method as described in any one of claims 1 to 6.

Images