Intelligent collection compensation measuring and calculating method based on multi-source heterogeneous data fusion

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent collection and compensation measuring and calculating method based on multi-source heterogeneous data fusion, which comprises the following steps: cutting and extracting house data according to house compensation requirements, house characteristics, location characteristics and historical compensation data, and determining a multi-source fused house characteristic sequence; based on retrieval entities of the house feature sequence in each feature dimension, constructing a house topology tree according to an association relationship of the retrieval entities; performing feature expansion on each branch node of a house topological tree based on house topological transformation of house data retrieval times, judging the branch nodes separated during each round of retrieval, and determining compensation weights of the branch nodes during each round of retrieval; determining a priority sequence under each round of compensation according to the time dimension of retrieval by aiming at the compensation weight of the branch node during each round of retrieval; and based on the priority sequence under each round of compensation, performing regular matching on the priority sequence, and configuring a compensation strategy. And the pertinence and the processing efficiency of house compensation calculation are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically a method for intelligent calculation of expropriation compensation based on the fusion of multi-source heterogeneous data. Background Technology

[0002] Traditional expropriation compensation management relies on manual calculation and verification, which is not only inefficient but also prone to overlooking the impact of different housing characteristics, resulting in problems such as inaccurate scenario delineation and unreasonable compensation in compensation plans, making it difficult to fit the specific scenario of housing expropriation compensation.

[0003] For example, Chinese Patent Publication No. CN120373803A discloses a method, device, equipment, and storage medium for allocating and managing engineering cost estimation tasks. By analyzing engineering cost estimation tasks, the method performs the first update action on the engineering cost estimation task list. It considers the calculation time intervals of different combinations of engineering cost estimation personnel in the execution of different types of calculation tasks. With calculation quality and calculation period as the constraint set, and the lowest handover cost of the calculation stage of engineering cost estimation tasks as the optimization objective, it optimizes the calculation strategy, realizes the second update of the engineering cost estimation task list and executes the calculation tasks. Then, through its own delay risk elimination capability or replanning the calculation strategy, it updates the engineering cost estimation task list for the third time.

[0004] For example, Chinese Patent Publication No. CN120410009A discloses a method for calculating the radius of a public service based on adaptive network evolution. The method includes: dividing the public service into multiple functional units based on the service content provided by the public service, with each functional unit including at least one service point; constructing a public service network that matches the public service; setting a network evolution mechanism for the public service network to achieve adaptive evolution of its capacity, connectivity, and modular growth; setting a dynamic adjustment strategy to achieve elastic resource scheduling and optimization of service coverage of the public service; obtaining the service coverage, and based on a multi-layer network model, obtaining the service radius of each service point within the service coverage, and combining the number of demand points covered by each service point and the resource consumption under each service radius to form an optimal service radius vector set; and determining the optimal coverage radius and optimal service radius of each service point.

[0005] In existing technologies, the statistical processing of cost estimation is completed through measurement tasks; or the radius of each processing is evolved through the update of demand nodes. However, existing technologies tend to be single-dimensional measurement and analysis. When the input data is multi-source heterogeneous data, its measurement and update tends to be content matching of hard-coded patterns, which leads to the lack of scene differentiation recognition and reduces the dynamic adaptability of the measurement results. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a smart calculation method for expropriation compensation based on multi-source heterogeneous data fusion, including: S1, according to the housing compensation demand, the housing data is cropped and extracted according to the housing characteristics, location characteristics and historical compensation data to determine the multi-source fused housing characteristic sequence.

[0007] S2, based on the house feature sequence in each feature dimension, retrieves the entities and constructs a house topology tree containing logical branches and conditional branches according to the association relationship of the retrieved entities.

[0008] S3, based on the number of house data retrievals, performs feature expansion on each branch node of the house topology tree, determines the branch node that is detached in each round of retrieval, and determines the compensation weight of the branch node in each round of retrieval.

[0009] S4 determines the priority sequence under each round of compensation based on the time dimension of the retrieval, according to the compensation weight of the branch nodes in each round of retrieval.

[0010] S5 performs regular expression matching on the priority sequence based on the priority sequence under each round of compensation, and configures the compensation strategy under the global index.

[0011] The beneficial effects of this invention are as follows: First, this invention determines a multi-source fusion housing feature sequence by cropping and extracting housing data according to housing compensation needs, housing characteristics, location characteristics, and historical compensation data. Simultaneously, based on the retrieval entities of the housing feature sequence in each feature dimension, and according to the association relationships of the retrieval entities, a housing topology tree containing logical branches and conditional branches is constructed. By transforming the abstract housing feature sequence into a structured housing topology tree, and by locking the differences in legal housing calculation rules and conditional branches according to logical branches, the calculation and processing of the current scenario is made closer to the dynamic scenario of housing compensation; the backtracking capability and global processing efficiency of housing calculation are improved.

[0012] Second, this invention performs feature expansion on each branch node of the house topology tree by transforming the house topology based on the number of house data retrievals, determines the branch nodes that are detached in each retrieval round, and determines the compensation weight of the branch nodes in each retrieval round. By using the retrieval round as the driving force to perform dynamic incremental updates to the topology tree, and determining the compensation weight of the branch nodes based on the data of each retrieval round, the timeliness and accuracy of the calculation results are ensured.

[0013] Third, this invention determines the priority sequence for each round of compensation by assigning compensation weights to branch nodes during each round of retrieval, based on the time dimension of the retrieval. Through time-dimensional data management, data updates can avoid the risk of market data overwriting the calculation rules, and the execution order under the calculation is clearly defined, making the calculation process more targeted and adaptable to the current housing expropriation scenario.

[0014] Fourth, this invention configures the compensation strategy under a global index by performing regular expression matching on the priority sequence under each round of compensation. Regular expression matching achieves the limitation of the same scenario and the same batch of scenarios, ensuring differentiated strategy configuration and further guaranteeing the fairness, compliance, and traceability of expropriation compensation. Attached Figure Description

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] Figure 1 This is a flowchart illustrating an intelligent calculation method for expropriation compensation based on the fusion of multi-source heterogeneous data.

[0017] Figure 2 This is a flowchart illustrating step S1 of an intelligent calculation method for expropriation compensation based on the fusion of multi-source heterogeneous data.

[0018] Figure 3 This is a flowchart illustrating step S2 of an intelligent calculation method for expropriation compensation based on the fusion of multi-source heterogeneous data.

[0019] Figure 4 This is a flowchart illustrating step S3 of an intelligent calculation method for expropriation compensation based on the fusion of multi-source heterogeneous data.

[0020] Figure 5 This is a flowchart illustrating step S4 of an intelligent calculation method for expropriation compensation based on the fusion of multi-source heterogeneous data.

[0021] Figure 6 This is a flowchart illustrating step S5 of an intelligent calculation method for expropriation compensation based on the fusion of multi-source heterogeneous data. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0023] See Figure 1 A method for intelligent calculation of expropriation compensation based on multi-source heterogeneous data fusion includes: S1, according to housing compensation needs, housing data is cropped and extracted according to housing characteristics, location characteristics and historical compensation data to determine the multi-source fused housing characteristic sequence.

[0024] S2, based on the house feature sequence in each feature dimension, retrieves the entities and constructs a house topology tree containing logical branches and conditional branches according to the association relationship of the retrieved entities.

[0025] S3, based on the number of house data retrievals, performs feature expansion on each branch node of the house topology tree, determines the branch node that is detached in each round of retrieval, and determines the compensation weight of the branch node in each round of retrieval.

[0026] S4 determines the priority sequence under each round of compensation based on the time dimension of the retrieval, according to the compensation weight of the branch nodes in each round of retrieval.

[0027] S5 performs regular expression matching on the priority sequence based on the priority sequence under each round of compensation, and configures the compensation strategy under the global index.

[0028] The current scheme is used for monetary compensation for the area of ​​demolished houses after the house is measured, demolition compensation for the number of households to be demolished, temporary resettlement fees, etc.; in essence, it is a guiding process described in text or specific rules, and the measured amount is configured according to the guiding result.

[0029] Specifically, the housing feature sequence takes housing-related attributes as input, including but not limited to land type, structural type, housing use, housing nature, housing type, area type, apportionment type, housing unit type and housing apartment type, as well as location features such as number of floors, orientation and geographical location. These features are used as search content to determine the historical compensation data for housing search and to form a housing feature sequence under multi-dimensional fusion based on the data search content.

[0030] Among them, land type indicates the different forms of land use such as commercial and financial land, industrial and warehousing land, and municipal land; structure type indicates the description of building structure such as frame, frame-shear wall, reinforced concrete, and brick-wood; building use describes the actual use of commercial buildings, office buildings, factories, office buildings, etc.; building nature describes the different types of buildings such as resettlement housing, low-rent housing, housing reform housing, and market-oriented commercial housing; building type describes the types of buildings such as industrial buildings, residential buildings, and commercial buildings; area type is divided into internal building area and shared building area; apportionment type is divided into whole building apportionment and floor apportionment; house type indicates one-bedroom, two-bedroom, etc.; house type describes single-level, split-level, duplex, and split-level.

[0031] Furthermore, when retrieving historical compensation data from the database, the characteristics of the houses are mapped to the floor plan to determine the specific structural style of each house in its location.

[0032] like Figure 2 As shown, one implementation of step S1 includes: S11, determining the target house to be searched according to the feature dimensions corresponding to the house characteristics and location characteristics, based on the housing compensation requirements.

[0033] S12 involves associating the target house with historical compensation data by feature to obtain associated data across multiple feature dimensions.

[0034] S13 cross-validates the associated data across multiple feature dimensions, performs attribute association aggregation on each feature dimension, and considers the aggregated data as the validated data.

[0035] S14: Sort the verified data according to the update time to obtain the output house feature sequence.

[0036] Specifically, when performing feature association, the historical compensation data is retrieved from the feature dimensions included in the currently entered house features and location features, and the data corresponding to the current feature dimensions is used as the associated data. For example, if the current house is a residential house, a three-bedroom flat, and a certain floor, it is only necessary to search according to the feature dimensions that correspond to residential houses, and the retrieved data is regarded as the set associated data.

[0037] Furthermore, the aforementioned cross-validation includes, but is not limited to, utilizing the relationships between house features across multiple dimensions to supplement the description of feature dimensions in order to address the limitations of numerical features; and generating feature sequences rich in semantic associations by capturing indirect relationships between different dimensions in order to improve the accuracy of subsequent compensation calculations.

[0038] Since location is directly related to housing characteristics, it directly determines the compensation price and depreciation period of the house; the supporting facilities of the location, such as school district, subway, commerce, and medical care, will indirectly amplify the value differences of houses with different uses; here, based on the direct and indirect correlation between location characteristics and housing characteristics in the dimensions, the relationship of the housing characteristic sequence is constructed.

[0039] Specifically, the implementation methods for attribute association aggregation of each feature dimension include: aggregating the input feature dimensions according to preset association rules to form directly associated feature groups; wherein, the preset association rules are rule descriptions divided according to the attributes of the house itself and associated attributes; the division forms include: regarding all features in the house features and the orientation / number of floors in the location features as directly associated features; and regarding the geographical location of the location features and the historical compensation data of similar transactions in the surrounding area as indirectly associated features.

[0040] Furthermore, the direct feature association groups are further grouped according to feature dimensions, forming property rights association groups such as land type → building use → building nature → building type, which locks in the statutory compensation category of the building and determines the applicable standard for the compensation benchmark price; area type → apportionment type → building unit type → building apartment area association group, which locks in the statutory compensation area of ​​the building and solves the compliance issue of area calculation; structure type → total number of floors → location association group, which locks in the reset benchmark of the building and determines the structural compensation standard; and building apartment type → building unit type → building orientation attribute association group, which locks in the benchmark for adjusting the building's attribute, providing a basic query anchor point for housing allocation.

[0041] For each directly related feature group, its geographical location and historical compensation data of similar transactions in the surrounding area are used as extended content to determine the aggregated indirectly related feature groups.

[0042] The aggregated indirect and direct related feature groups are output as a house feature sequence according to the data update time.

[0043] Specifically, the aforementioned indirectly related feature groups will extend the surrounding historical data according to dimensions such as house use, location, area type, and structure type. Furthermore, using the processing logic of location location → surrounding house characteristics → transaction price difference → location correction, the feature dimensions of each directly related feature group will be extended and configured to determine the output house feature sequence under each dimension.

[0044] Furthermore, the regional correction process will be reflected in the pruning and updating of the house topology tree to determine the feature weights of different house features under the data update and the compensation process, ultimately affecting the specific output of the compensation scheme.

[0045] In one embodiment of the present invention, the logical branch represents a necessary process that strictly follows the statutory regulatory requirements for compensation for the expropriation of houses on state-owned / collective land, which is a necessary process that cannot be pruned and is a mandatory condition for housing compensation verification; its conditional branch is required to cover the retrieval scenario of specific elements and adapt to the differentiated expropriation scenarios under the same main logical branch, such as the existence of different situations such as street-facing residential conversion to commercial use, shared ownership, and historical legacy area, even if they are state-owned residential properties; its conditional branch serves as an optional condition for housing compensation verification.

[0046] The implementation of the house topology tree is as follows: First, according to the direct and indirect associations corresponding to the house feature sequence, the content corresponding to the logical branches and conditional branches is divided, and the topology tree positions of the logical branches and conditional branches are set according to the basic weights configured in the sequence; then, according to the business constraints corresponding to the basic weights, pruning is performed on the logical branches and conditional branches; the update content when the house feature sequence is entered is determined, and thus the house topology tree is set.

[0047] The root node of the final housing topology tree represents the main body of the currently calculated housing, and is further divided into three layers: Layer 1 nodes, Layer 2 nodes, and Layer 3 nodes. Layer 1 nodes represent the directly related parts, directly determining the calculated housing price. Based on the four sets of data corresponding to the directly related feature groups, Layer 1 nodes are divided into four data points, and logical branches are set downwards according to the specific descriptions in each data group, explaining the specific values ​​of the nodes. Conditional branches are also introduced as supplementary content under the normal logic. Layer 2 nodes represent the data parts that cannot be directly bound to the housing, belonging to the data elements contained in the indirectly related feature groups. Logical branches are set for these elements based on the location, supporting facilities, and compensation data of surrounding areas with similar conditions contained in the indirectly related feature groups. If a conditional branch exists in a Layer 1 node, a corresponding conditional branch or logical branch is added. Layer 3 nodes represent the legal policies and project plans for the entire calculation process, constraining the entire calculation process. Their logical branches are set according to policy requirements and business constraints, and node descriptions corresponding to the upper-level conditional branches are introduced. Finally, the root node represents the compensation plan that satisfies the constraints, explaining the allocation of different housing plans in the current scenario.

[0048] Specifically, such as Figure 3 As shown, the implementation of step S2 includes: S21, extracting the retrieval entity from the house feature sequence, and constructing the logical branch corresponding to the retrieval entity based on the feature dimension of the retrieval entity; wherein, the retrieval entity is a lower-level description of the house attribute, specifically represented as static text or numerical value of a single house, to characterize the specific parameters that need to be retrieved for the current house, such as the retrieval entity of land type = residential land, the retrieval entity of house use = residential / apartment, and taking the specific description under the corresponding feature dimension as the retrieval entity of the current scene.

[0049] Furthermore, the logical branches are set based on the feature dimensions corresponding to the housing data, so that branches corresponding to different entities are constructed in sequence according to feature dimensions such as land type, structure type, housing use and housing nature.

[0050] Specifically, the logical branches are divided into four branches based on four types of direct relationships: ownership association, area association, location association, and ontological attribute association. The subsequent branch nodes are selected based on the actual value of the current feature dimension under the corresponding logical branch.

[0051] S22, under the same logical branch, add conditional branches to the logical branch based on the feature differences of the retrieved entities.

[0052] Furthermore, the differences in features represent different values ​​under the feature dimension. For example, when the property use is residential and the owner holds a legal business license, conditional branches for commercial and residential properties are set at the level 1 node, respectively. For the current logical branch, conditional branches for commercial and residential properties in the same location are set at the level 2 node. In the current scenario, the logical branch is further described as an edge directly connected to the root node of the property topology tree, and the conditional branch serves as an extension of the logical branch to differentiate it.

[0053] At the same time, since the searched entities correspond to specific directly related feature groups and indirectly related feature groups, the logical branches and conditional branches formed here will be further eliminated according to the aggregation form of their directly related feature groups and indirectly related feature groups to determine the data boundaries of the current search.

[0054] The implementation of adding conditional branches to the logical branches in step S22 includes: activating multiple retrieval paths in parallel, with each retrieval path using different feature dimensions and fuzzy rules to perform the retrieval operation.

[0055] Retrieve the search results corresponding to the current logical branch, and divide the search results according to the location of the logical branch to obtain the set condition branches.

[0056] Specifically, fuzzy rules represent the text content that matches the description of the entity being searched. For example, if the current search content is "residential to commercial conversion", the database can be used to view the optional condition branches related to it based on the fuzzy rules set for "residential to commercial conversion". The condition branches can then be processed according to their positions in the logical branches.

[0057] S23. Based on the association relationship of the retrieved entities, perform node partitioning operation according to the direct and indirect associations corresponding to the retrieved entities, and determine the branch nodes corresponding to the conditional branches and logical branches.

[0058] S24. Under the same logical branch, the basic weights of the branch nodes are configured by using the ratio of the feature value of each branch node to the feature value of the total branch nodes.

[0059] In this process, the feature value of each branch node represents the data under the current dimension. After standardization, the attribute of the current house is transformed into a numerical feature, and a cross-correlation matrix is ​​constructed. By extracting feature values ​​from the cross-correlation matrix, the feature value of each branch node is obtained. Then, the ratio of its feature value to the feature value of the total branch nodes is used to set the priority of each branch node's association under the same logical branch. It should be noted that the feature value calculated here needs to be bound to a specific logical branch and a specific feature dimension to complete the feature value configuration under the same conditions.

[0060] Specifically, the input to the cross-correlation matrix is ​​the standardized feature vector of each branch node under the same logical branch, and the matrix dimension is N×N (N is the number of nodes under the logical branch). Feature value extraction adopts the feature decomposition algorithm. The feature value of the branch node is the variance contribution of the corresponding feature root, and the feature value of the total branch node is the sum of the feature values ​​of all nodes under the logical branch.

[0061] The basic weights for branch nodes are set based on their characteristic values ​​to determine the differentiated processing of different input data formats in housing compensation demand, so as to determine the priority housing cases to be processed.

[0062] S25 introduces the association priority corresponding to the basic weight and the business constraints corresponding to the branch node, performs pruning on the branch node of the conditional branch, and sets the output house topology tree according to the level of the branch node after pruning.

[0063] When performing pruning, the branch nodes under each condition branch are first checked according to business constraints. For example, if the collective homestead branch is hit, entities such as the benchmark land price of state-owned land and commercial housing transaction cases are automatically removed; if the commercial housing branch is hit, entities such as school district facilities and residential unit premiums are automatically removed. Connections that do not conform to business constraints are pruned, and the corresponding data is synchronized to the parent branch node to redetermine the subdivided lower branch nodes.

[0064] Association priority represents the data priority form under a specific feature dimension, which is expressed as the value of the basic weight. For example, under the housing use, the priority of residential compensation is greater than that of warehouse and factory.

[0065] The final output house topology tree represents a comprehensive solution under a specific dimension. The compensation scheme can be limited based on its data intersection, or it can be output in full as a data reference for subsequent use.

[0066] As for feature expansion in step S3, it follows the logic of incremental cross-validation, performs hierarchical mapping on the new housing compensation requirements entered in each round, and resets the compensation weight of each topology tree node through topology tree pruning and hierarchical remapping, thereby determining the priority processing part in each round of retrieval.

[0067] like Figure 4 As shown, the implementation method for determining the branch nodes that are detached during each round of retrieval in step S3 includes: S31, based on the triggering scenario of each round of retrieval, identifying the newly added branch nodes of each layer of the house topology tree.

[0068] S32, based on the newly added branch nodes in each round of retrieval, perform incremental cross-validation on all nodes. Newly added branch nodes that fail the validation are directly marked as nodes to be removed; newly added branch nodes that pass the validation proceed directly to step S33.

[0069] S33, perform hierarchical mapping on the newly added branch node according to its type to determine the branch to which the newly added branch node belongs.

[0070] S34, based on the list of nodes to be removed, perform branch pruning on the house topology tree and update the house topology tree structure synchronously.

[0071] Specifically, the incremental cross-validation process includes: incremental cross-validation encompasses data compliance checks, logical branch matching checks, and correlation checks; for nodes that fail the data compliance check, the corresponding data is directly cut from the house topology tree; for nodes that fail the logical branch matching and correlation checks, the position of the corresponding new branch node in the house topology tree is retained, marked as inactive, and incremental cross-validation is re-executed when newly entered data is related to the new branch node. If all new branch nodes under the entire conditional branch are nodes to be detached, the conditional branch is detached, and the general compensation method at the logical branch is reverted.

[0072] It should be noted that the current house topology tree uses an incremental node update method, without deleting existing data nodes, thus preserving the data update process across different versions. For data that has been pruned from a branch and is illogical to the current house topology tree structure, the pruned data will be stored separately with a reason for pruning, rather than simply being marked as inactive.

[0073] Among them, data compliance verification is used to verify whether newly added data complies with legal policies and project plans. Newly added data that complies with legal policies and project plans is considered to be compliant. Logical branch matching verification is used to determine whether the attributes of newly added data are consistent with the scenarios represented by logical branches and whether they correspond to their business logic. If the business logic passes, it is considered to be logical branch matching.

[0074] The correlation verification is performed by performing vector calculations on the new data and the corresponding parent node. After calculating the similarity value using cosine similarity according to the same data dimension, the similarity value is multiplied by the normalized time interval weight to obtain the correlation value of the current new branch node. When the correlation value is greater than the preset correlation value, it is considered a correlation match. For each new branch node, the timeliness of the new data needs to be verified to determine the availability of the compensation scheme in the new data scenario.

[0075] Furthermore, the weight of the time interval decreases as the data time span increases, with 1 for the first 6 months, linearly decreasing to 0.3 for the last 6 months to 2 years, and 0 for more than 2 years.

[0076] Furthermore, the preset correlation value is set based on the average correlation value of historical compensation data of the same type to determine the data distribution under normal circumstances.

[0077] S35. Based on the updated house topology tree, calculate the node contribution of each branch node in the house topology tree, and set the compensation weight according to the node contribution.

[0078] The node contribution is linked to the retrieval round and the incremental cross-validation results, and the node contribution is set according to the level of the branch node.

[0079] For example, the contribution of a branch node that fails incremental cross-validation is 0, while the contribution of a branch node that passes all cross-validations is set to 1.2. If the branch node added in this round is a new feature dimension, the contribution of that branch node is multiplied by 1.05, so that its contribution does not exceed 1.5. Finally, the obtained node contribution is multiplied by the configured base weight to obtain the output compensation weight.

[0080] The method for setting the basic weights here is the same as above. When the newly added branch node updates an existing feature dimension, the weight value of that feature dimension under the corresponding logical branch is retrieved, and its weight value is used as the basic weight of the newly added branch node, so that each data update can be identified based on the change of the basic weight. As for when the newly added branch node represents a newly added feature dimension, the basic weight is calculated in the same way as in the above process, and the compensation weight is set in combination with the node contribution.

[0081] It should be noted that the compensation weight is used to emphasize the relative situation under data updates. Its value does not directly participate in the calculation of compensation price. Instead, it is used to verify differentiated data in the global scenario. Its value is used to prioritize the retrieval of differentiated data to obtain the subsequently configured calculated price. Since housing calculation is directly defined and calculated through parameters such as property rights and area, the current processing is used to highlight the differences in demand caused by different regions or different housing attributes. By setting weights, priority is given to data with frequent changes in demand, thereby improving the rationality and relative fairness of housing compensation calculation.

[0082] In one embodiment of the present invention, the retrieval rounds are used to record the new compensation demands of the currently expropriated houses after each full business data retrieval is completed; the rounds are represented as follows: Round 1: Initial data retrieval, determining the initial state of the topology tree, which is the preliminary construction of the house topology tree in step S2; Round 2: Surrounding comparable cases / supporting facilities data retrieval; Round 3: Data retrieval of houses already signed in the same batch of the same project; Round 4: Retrieval of new compensation demands / supplementary materials from property owners; Round N: Data retrieval of policy adjustments / project plan changes.

[0083] The system updates the data round by round based on the entered housing compensation requirements. Based on the search results of each round, the system performs structured adjustments to the topology tree, including adding / cutting nodes, activating / detaching branches, and remapping hierarchical affiliations, without changing the hierarchical structure defined by the logical branches of the topology tree.

[0084] In one embodiment of the present invention, the compensation weight is set according to the update situation of each round. Specifically, the update probability of the feature dimension where the compensation weight is located is determined. Here, the compensation weight is used to record the ratio of the number of times the compensation weight is updated under a single feature dimension to the number of times all dimensions are updated, so as to quantify the data iteration weight under the update of housing compensation demand, so as to characterize the frequency of demand changes and updates.

[0085] like Figure 5 As shown, one implementation of step S4 includes: S41, converting the time dimension of the retrieval into the time window of the expropriation compensation business cycle, and obtaining the update interval corresponding to each round of retrieval.

[0086] S42, based on the update interval corresponding to each round of retrieval, repeatedly obtain the change path of the branch node in continuous time, combine the relative flow direction of the change path, lock the feature dimension corresponding to the branch node, and determine the update probability of the feature dimension where the compensation weight is located.

[0087] S43. According to the update probability of the feature dimension where the compensation weight is located, the feature dimension is decomposed into multiple compensation sub-items, and the compensation sub-items are sorted in descending order according to the value of the update probability to obtain the output priority sequence.

[0088] Furthermore, the feature sub-items represent specific compensation items decomposed from housing characteristics and location characteristics, such as location, building age, building structure, unit type, and property rights. Each compensation sub-item specifies the classification number under the corresponding feature dimension. For example, under the housing use, number 10 is commercial and number 12 is office. Here, feature dimensions that frequently have data updates are extracted first and used as the input content for subsequent regular expression matching in the manner of feature sub-items, so as to quickly generate the compensation plan for the current housing.

[0089] Furthermore, the change path of the branch node in continuous time represents the path of the newly added branch node in the housing topology tree. Its relative flow direction determines the currently frequently updated feature dimensions, thereby determining the multi-dimensional feature update situation to match the priority considerations under the corresponding area's housing expropriation.

[0090] Specifically, the change path starts with the node at the level where the logical branch is located and ends with the newly added branch node, connecting the branch nodes at the corresponding levels in sequence. The connection direction of the newly added branch node at different levels is regarded as its relative flow direction. The logical branch corresponding to each change path and the corresponding feature dimension are recorded.

[0091] In one embodiment of the present invention, regular expression matching is performed sequentially through the compensation sub-items of the priority sequence to finally determine the output compensation strategy.

[0092] like Figure 6 As shown, one implementation of configuring the compensation strategy under the global index in step S5 includes: S51, determining the logical branch and conditional branch corresponding to the input priority sequence. Here, it is necessary to determine the logical branch corresponding to each feature item in the priority sequence, that is, to determine the mandatory and optional values ​​under housing compensation, thereby completing the matching of the current data with relevant historical data and completing the regular expression matching of the strategy.

[0093] S52, the logical branch and conditional branch are sequentially matched with historical compensation data using regular expressions to obtain multiple regular expression matching results. The union of these results is considered the output compensation strategy. Regular expression matching introduces the computational content required for compensation calculation, searches for compensation strategies sequentially according to the content corresponding to the compensation sub-items, obtains multiple retrieved regular expression matching results, and finally selects their union to obtain a compensation scheme that satisfies various housing characteristics and location constraints.

[0094] Furthermore, if we consider the data variation patterns under housing compensation demand, then when performing regular expression matching between the logical branch and the conditional branch and the historical compensation data in step S5, the implementation method includes: dividing the priority sequence of the same scenario and the same batch into the same comparison group according to the rule entries of the global index; wherein, the same scenario represents housing data of the same project, the same land nature, the same housing use and the same housing structure.

[0095] For the same comparison group, a global benchmark strategy is set based on the characteristics of logical branch mapping; this global benchmark strategy represents the project's statutory compensation benchmark price, area determination rules, and compensation sub-items, and represents the general compensation under branches with no optional conditions.

[0096] For the same comparison group, based on the characteristics of the conditional branch mapping, a household adjustment strategy is set, and the global benchmark strategy and the household adjustment strategy are regarded as the output compensation strategy. The household adjustment strategy means that based on the value difference on the conditional branch, the housing compensation strategy is set for each household in the same batch according to the relocation fee, temporary resettlement fee, and loss of production and business interruption.

[0097] As for the comparison groups of different batches, the compensation strategies for their specific scenarios will be improved according to the global benchmark strategy and household adjustment strategy divided within the current batch.

[0098] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A method for intelligent calculation of expropriation compensation based on multi-source heterogeneous data fusion, characterized in that, include: S1. Based on housing compensation needs, housing data is cropped and extracted according to housing characteristics, location characteristics and historical compensation data to determine the multi-source fusion housing feature sequence; S2, based on the retrieval entities of the house feature sequence in each feature dimension, and according to the association relationship of the retrieval entities, construct a house topology tree containing logical branches and conditional branches; S3, based on the number of house data retrievals, performs feature expansion on each branch node of the house topology tree, determines the branch node that is detached in each round of retrieval, and determines the compensation weight of the branch node in each round of retrieval. S4, for the compensation weight of the branch nodes in each round of retrieval, determine the priority sequence under each round of compensation according to the time dimension of retrieval; S5 performs regular expression matching on the priority sequence based on the priority sequence under each round of compensation, and configures the compensation strategy under the global index.

2. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The implementation methods for determining the multi-source fusion house feature sequence in step S1 include: S11. Based on the housing compensation needs, determine the target housing to be searched according to the feature dimensions corresponding to housing characteristics and location characteristics respectively; S12, perform feature association between the target house and historical compensation data to obtain associated data with multiple feature dimensions; S13, cross-validate the associated data of multiple feature dimensions, perform attribute association aggregation on each feature dimension, and regard the aggregated data as the data that has passed the validation; S14: Sort the verified data according to the update time to obtain the output house feature sequence.

3. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 2, characterized in that, The implementation methods for attribute association and aggregation across feature dimensions include: The entered feature dimensions are aggregated according to preset association rules to form directly associated feature groups; For each directly related feature group, its geographical location and historical compensation data of similar transactions in the surrounding area are used as extended content to determine the aggregated indirectly related feature groups. The aggregated indirect and direct related feature groups are output as a house feature sequence according to the data update time.

4. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The implementation of the house topology tree in step S2 includes: S21, extract the retrieval entity from the house feature sequence, and construct the logical branch corresponding to the retrieval entity based on the feature dimension of the retrieval entity; S22, Under the same logical branch, add conditional branches to the logical branch based on the feature differences of the retrieved entities; S23, Based on the association relationship of the retrieved entities, perform node division operation according to the direct and indirect associations corresponding to the retrieved entities, and determine the branch nodes corresponding to the conditional branches and logical branches; S24. Under the same logical branch, the basic weights of the branch nodes are configured by using the ratio of the feature value of each branch node to the feature value of the total branch nodes. S25 introduces the association priority corresponding to the basic weight and the business constraints corresponding to the branch node, performs pruning on the branch node of the conditional branch, and sets the output house topology tree according to the level of the branch node after pruning.

5. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 4, characterized in that, The implementation methods for adding conditional branches to the logical branches in step S22 include: Multiple retrieval paths are activated in parallel, and each retrieval path performs retrieval operations using different feature dimensions and fuzzy rules. Retrieve the search results corresponding to the current logical branch, and divide the search results according to the location of the logical branch to obtain the set condition branches.

6. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The implementation methods for determining the branch node detached during each round of retrieval in step S3 include: S31, based on the triggering scenario of each round of retrieval, identify the newly added branch nodes of each layer of the house topology tree; S32, Based on the newly added branch nodes in each round of retrieval, perform incremental cross-validation on all nodes. Newly added branch nodes that fail the validation are directly marked as nodes to be removed; newly added branch nodes that pass the validation proceed directly to step S33. S33, perform hierarchical mapping on the newly added branch nodes according to their type to determine the branch to which the newly added branch nodes belong; S34, Based on the list of nodes to be removed, perform branch pruning on the house topology tree and update the house topology tree structure synchronously; S35. Based on the updated house topology tree, calculate the node contribution of each branch node in the house topology tree, and set the compensation weight according to the node contribution.

7. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 6, characterized in that, Incremental cross-validation methods include: Incremental cross-validation includes data compliance verification, logical branch matching verification, and correlation verification. For nodes that fail the data compliance verification, the corresponding data is directly cut from the house topology tree. For nodes that fail the logical branch matching verification and correlation verification, the position of the corresponding new branch node in the house topology tree is retained, and it is marked as inactive. When newly entered data is related to the new branch node, incremental cross-validation is re-executed.

8. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 1, characterized in that, Step S4 involves determining the implementation method of the priority sequence under each round of compensation, including: S41, transform the time dimension of the retrieval into the time window of the expropriation compensation business cycle, and obtain the update interval corresponding to each round of retrieval; S42, based on the update interval corresponding to each round of retrieval, repeatedly obtain the change path of the branch node in continuous time, combine the relative flow direction of the change path, lock the feature dimension corresponding to the branch node, and determine the update probability of the feature dimension where the compensation weight is located. S43. According to the update probability of the feature dimension where the compensation weight is located, the feature dimension is decomposed into multiple compensation sub-items, and the compensation sub-items are sorted in descending order according to the value of the update probability to obtain the output priority sequence.

9. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The implementation methods for configuring the compensation strategy under the global index in step S5 include: S51, based on the input priority sequence, determine the logical branch and conditional branch corresponding to the priority sequence; S52, perform regular expression matching between the logical branch and the conditional branch and the historical compensation data in sequence to obtain multiple regular expression matching results, and regard the union of the regular expression matching results as the output compensation strategy.

10. The intelligent calculation method for expropriation compensation based on multi-source heterogeneous data fusion according to claim 9, characterized in that, In step S5, when the logical branch and conditional branch are sequentially matched with the historical compensation data using regular expressions, the implementation methods include: According to the rule entries of the global index, priority sequences of the same scenario and the same batch are divided into the same comparison group; For the same comparison group, a global benchmark strategy is set based on the features of logical branch mapping; For the same comparison group, based on the characteristics of conditional branch mapping, a household adjustment strategy is set, and the global benchmark strategy and the household adjustment strategy are regarded as the output compensation strategy.

Images