A multi-dimensional reconciliation method and system based on a group meal platform

Abstract

The present application relates to the technical field of data processing, and especially to a multi-dimensional reconciliation method and system based on a group meal platform, which comprises the following steps: determining the state of a sub-node according to a steady-state metric value and a correlation index, and determining a reference data selection method according to the state of the sub-node; in the single-node direct selection, the sub-node corresponding to the maximum comprehensive evaluation value is selected as the reference data; in the multi-node compensation selection, a compensation method is determined according to a steady-state mutual correlation degree and a steady-state balance coefficient to select the reference data; and a reconciliation sub-node selection method is determined according to a reference comparison coefficient and a cross-correlation coefficient of the sub-node corresponding to the to-be-reconciled data and the reference data; and the present application can improve the reconciliation processing efficiency.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a multi-dimensional reconciliation method and system based on a group meal platform. Background Technology

[0002] As the scale of group meal platform business expands, the volume of transaction data surges. Traditional manual reconciliation methods can no longer efficiently process massive amounts of data, and abnormal data is easily buried in normal transactions, resulting in poor reconciliation efficiency. Therefore, how to select key data for reconciliation to improve reconciliation efficiency is a technical problem that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN117149797B discloses a reconciliation method and system based on multi-dimensional data monitoring. The method includes: determining the account update status value of different servers based on account update processing failure data and account update interruption data from different servers; determining the probability of reconciliation problems based on the account update status value; determining the account update data of different servers based on server operation log data; determining the account update busyness based on the account update data of different servers; monitoring the operation data of different servers to obtain monitoring operation data; and determining the overall problem probability by combining the reconciliation problem probability and the account update busyness. When the overall problem probability does not meet the requirements or reaches the preset reconciliation cycle, reconciliation processing is performed on the account data of different servers. However, the above technical solution has the following problems: it mainly relies on the overall problem probability and the preset reconciliation cycle for reconciliation processing, lacks in-depth mining of the inherent correlation between different account data, and does not select reference data for dynamic optimization of the comparison benchmark, resulting in poor reconciliation processing efficiency and accuracy. Summary of the Invention

[0004] To address this, the present invention provides a multi-dimensional reconciliation method and system based on a group meal platform, which overcomes the problems of poor reconciliation efficiency and accuracy caused by the lack of in-depth mining of the inherent correlation between different account data and the failure to select reference data for dynamic optimization of the comparison benchmark in the existing technology.

[0005] To achieve the above objectives, the present invention provides a multi-dimensional reconciliation method based on a group meal platform, comprising:

[0006] The status of the sub-nodes is determined based on the stable attitude value and the correlation index. The method of selecting reference data is determined based on the status of the sub-nodes. The method of selecting reference data is either direct selection of a single node or compensation selection of multiple nodes.

[0007] In the direct selection of a single node, the reconciliation data corresponding to the sub-node with the largest comprehensive evaluation value is used as the reference data.

[0008] In the selection of multi-node compensation, the compensation method is determined based on the steady-state cross-correlation degree and steady-state equilibrium coefficient to select reference data. The compensation method is partial compensation of multiple steady-state sub-nodes or correlation compensation of reference steady-state sub-nodes.

[0009] The selection method for reconciliation sub-nodes is determined based on the reference comparison coefficient and cross-relation coefficient between the data to be reconciled and the reference data corresponding to the sub-nodes. The selection method for reconciliation sub-nodes is selected based on the interaction influence coefficient or comparison threshold.

[0010] The feature data is determined based on the combined bias influence degree and the balance coefficient. The combined bias influence degree is determined based on the bias reference value of the cluster corresponding to the data to be reconciled.

[0011] Furthermore, if the child node's status is a stable attitude value greater than or equal to a preset stable attitude value and the correlation index is greater than or equal to a preset correlation index, then the reference data selection method is direct selection of a single node.

[0012] Furthermore, if the child node status is a stable attitude value less than the preset stable attitude value or a correlation index less than the preset correlation index, then the reference data selection method is multi-node compensation selection.

[0013] Furthermore, if the steady-state cross-correlation degree is greater than or equal to the preset steady-state cross-correlation degree and the steady-state equilibrium coefficient is greater than or equal to the preset steady-state equilibrium coefficient, then the compensation method is partial compensation of multiple steady-state sub-nodes.

[0014] In the multi-stable sub-node partial compensation, the selected sub-nodes are determined based on the correlation deviation degree and multidimensional covariance weights, and the feature data in each selected sub-node are selected as reference data based on the difference coefficient and class balance degree.

[0015] Furthermore, if the steady-state cross-correlation degree is less than the preset steady-state cross-correlation degree or the steady-state equilibrium coefficient is less than the preset steady-state equilibrium coefficient, the compensation method is reference steady-state sub-node correlation compensation.

[0016] The reference steady-state child node association compensation includes:

[0017] Reference sub-nodes are determined based on the evaluation threshold, and compensation sub-nodes are selected based on the sub-influence coefficients and combinations of associated nodes.

[0018] The data compensation range for each compensation sub-node is determined based on the cross-influence threshold and the iterative correlation degree.

[0019] The supplementary selection method is determined based on the deviation reference coefficient of the data compensation range corresponding to each compensation sub-node;

[0020] All feature data in the reference child node and the feature data in the data compensation range corresponding to each compensation child node are used as reference data.

[0021] Furthermore, the supplementary selection method is determined based on the deviation reference coefficient of the data compensation range corresponding to each compensation sub-node, including:

[0022] If the deviation reference value is greater than or equal to the preset deviation reference value, the supplementary selection method is to increase the number of compensation sub-nodes.

[0023] If the deviation reference value is less than the preset deviation reference value, the supplementary selection method is to increase the adjustment of the data compensation range.

[0024] Furthermore, the clustering method is determined based on the data variability and increment coefficient of each child node, including:

[0025] For a single child node,

[0026] If the data variability is greater than or equal to the preset data variability or the increment coefficient is greater than or equal to the preset increment coefficient, the clustering method is to split the cluster according to the variability coefficient.

[0027] If the data variation is less than the preset data variation and the increment coefficient is less than the preset increment coefficient, then the clustering method is to perform overall clustering based on the correlation coefficient.

[0028] Furthermore, the feature data is the reconciliation data where the combined bias influence is less than the preset combined bias influence and the balance coefficient is greater than or equal to the preset balance coefficient.

[0029] Furthermore, the selection method for reconciliation sub-nodes is determined based on the reference comparison coefficients and cross-relation coefficients between the data to be reconciled and the reference data corresponding to the sub-nodes, including:

[0030] If the reference comparison coefficient is greater than or equal to the preset reference comparison coefficient and the cross-correlation number is greater than or equal to the preset cross-correlation number, then the sub-node selection method is based on the interaction influence coefficient.

[0031] If the reference comparison coefficient is less than the preset reference comparison coefficient or the cross-correlation number is less than the preset cross-correlation number, the reconciliation sub-node selection method is to select based on the comparison threshold.

[0032] This invention also provides a multi-dimensional reconciliation system based on a group meal platform, comprising:

[0033] The data acquisition module includes several sub-nodes for data acquisition.

[0034] The status analysis module, which is connected to the data acquisition module, is used to determine the status of the sub-nodes based on the stable state values ​​and correlation indices, and to determine the reference data selection method based on the sub-node status. The reference data selection method is either direct selection of a single node or compensation selection of multiple nodes.

[0035] The first selection module, which is connected to the status analysis module, is used to select the reconciliation data corresponding to the sub-node with the largest comprehensive evaluation value as reference data in the direct selection of a single node.

[0036] The second selection module, which is connected to the state analysis module, is used to determine the compensation method and select reference data in the multi-node compensation selection based on the steady-state cross-correlation degree and steady-state equilibrium coefficient. The compensation method is partial compensation of multiple steady-state sub-nodes or correlation compensation of reference steady-state sub-nodes.

[0037] The reconciliation selection module is connected to the first selection module and the second selection module respectively. It is used to determine the reconciliation sub-node selection method based on the reference comparison coefficient and cross-relation coefficient of the data to be reconciled and the reference data corresponding to the sub-node. The reconciliation sub-node selection method is selected based on the interaction influence coefficient or the comparison threshold.

[0038] The feature data is determined based on the combined bias influence degree and the balance coefficient. The combined bias influence degree is determined based on the bias reference value of the cluster corresponding to the data to be reconciled.

[0039] Compared with the prior art, the beneficial effects of the present invention are that, in the technical solution of the present invention, the status of the sub-node is determined according to the stability value and the correlation index. The stability value and the correlation index effectively reflect the stability and correlation of the data to be reconciled in each sub-node. Then, different reference data selection methods are adaptively selected according to the status of the sub-node, so that the selection of reference data selection methods is more in line with the actual application scenario and avoids the problem of poor reconciliation efficiency caused by inaccurate selection of reference data.

[0040] Furthermore, this invention effectively reflects the degree of interaction between stable and unstable child nodes and the proportion of stable child nodes through steady-state cross-correlation degree and steady-state equilibrium coefficient. Then, based on steady-state cross-correlation degree and steady-state equilibrium coefficient, different compensation methods are adaptively selected, which can deeply explore the inherent correlation of data in different child nodes, and then dynamically adjust the combination of data in child nodes, achieving a balance between reducing computational overhead and improving fault tolerance, thereby improving reconciliation processing efficiency.

[0041] Furthermore, this invention effectively reflects the degree of data anomaly and change status of sub-nodes through data anomaly degree and increment coefficient. Then, based on the data anomaly degree and increment coefficient corresponding to each sub-node, different clustering methods are adaptively selected, so that the selected clustering method can automatically split the reconciliation data in high-risk sub-nodes, balance the requirements of computing efficiency and accuracy, and thus improve the reconciliation processing efficiency.

[0042] Furthermore, this invention effectively reflects the abnormal state of sub-data and the potential deviation impact of specific data clusters on the overall result by combining the bias influence degree and the balance coefficient. Then, the feature data is determined according to the combined bias influence degree and the balance coefficient, which can achieve a dynamic balance between accuracy and efficiency, improve the monitoring efficiency of hidden abnormal data, and make the selected reference data more representative. Then, the reconciliation sub-node selection method is determined according to the reference comparison coefficient and cross-correlation coefficient of the data to be reconciled and the reference data corresponding to the sub-node, making the selection of reconciliation sub-nodes more accurate, thereby improving the efficiency and accuracy of reconciliation processing. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the multi-dimensional reconciliation method based on a group meal platform according to the present invention;

[0044] Figure 2 This is a flowchart illustrating how the reference data selection method is determined based on the child node status according to the present invention.

[0045] Figure 3 This is a flowchart illustrating how the compensation method is determined based on steady-state cross-correlation degree and steady-state equilibrium coefficient according to the present invention.

[0046] Figure 4 This is a module connection diagram of the multi-dimensional reconciliation system based on the group meal platform of the present invention. Detailed Implementation

[0047] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0048] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0049] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0050] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0051] Please see Figures 1 to 3 As shown, this invention provides a multi-dimensional reconciliation method based on a group meal platform, including:

[0052] The status of the sub-nodes is determined based on the stable attitude value and the correlation index. The method of selecting reference data is determined based on the status of the sub-nodes. The method of selecting reference data is either direct selection of a single node or compensation selection of multiple nodes.

[0053] In the direct selection of a single node, the reconciliation data corresponding to the sub-node with the largest comprehensive evaluation value is used as the reference data.

[0054] In the selection of multi-node compensation, the compensation method is determined based on the steady-state cross-correlation degree and steady-state equilibrium coefficient to select reference data. The compensation method is partial compensation of multiple steady-state sub-nodes or correlation compensation of reference steady-state sub-nodes.

[0055] The selection method for reconciliation sub-nodes is determined based on the reference comparison coefficient and cross-relation coefficient between the data to be reconciled and the reference data corresponding to the sub-nodes. The selection method for reconciliation sub-nodes is selected based on the interaction influence coefficient or comparison threshold.

[0056] The feature data is determined based on the combined bias influence degree and the balance coefficient. The combined bias influence degree is determined based on the bias reference value of the cluster corresponding to the data to be reconciled.

[0057] The application scenario of this invention is the selection of data to be reconciled when a group meal platform performs reconciliation processing. This invention includes several sub-nodes, each of which contains several data to be reconciled. Each data to be reconciled contains several data names and corresponding values ​​for each data name. The data names include, but are not limited to, order amount, delivery fee, settlement cycle, settlement amount, handling fee, and tax. This is content that is easy for those skilled in the art to understand, and will not be elaborated further.

[0058] This invention includes several historical records. Each historical record contains at least one stable state value, correlation index, comprehensive evaluation value, steady-state cross-correlation degree, steady-state equilibrium coefficient, difference coefficient, and category balance degree during the historical process of selecting reconciliation data. Each historical record also has a corresponding qualified mark, which records whether the selection of reconciliation data meets the user's needs. The qualified mark can be recorded manually. It is understood that the user can determine whether the selection process of reconciliation data meets the needs based on self-defined indicators. Self-defined indicators can be, but are not limited to, the misjudgment rate, which will not be elaborated here. The misjudgment rate is the number of times the reconciliation sub-node is incorrectly selected.

[0059] The present invention sets a continuous cyclic monitoring cycle. At the end of each monitoring cycle, the status of the sub-node is determined. The duration of the monitoring cycle can be set according to the user's needs. The greater the user's need for monitoring accuracy, the shorter the duration of the monitoring cycle. One value for the monitoring cycle is provided, which is 1 hour.

[0060] The comprehensive evaluation value corresponding to a single sub-node = the steady-state reference value corresponding to the sub-node + the associated reference value corresponding to the sub-node. The associated reference value corresponding to a single sub-node = 1 - (a1 / a). The average value of the name reference values ​​corresponding to each reconciliation data collected by the sub-node during the target monitoring period will be recorded as a1.

[0061] Specifically, if the child node's status is a stable attitude value greater than or equal to the preset stable attitude value and the correlation index is greater than or equal to the preset correlation index, then the reference data selection method is to directly select a single node.

[0062] The sub-node status includes the first sub-node status and the second sub-node status. The first sub-node status is when the stable attitude value is greater than or equal to the preset stable attitude value and the correlation index is greater than or equal to the preset correlation index. The second sub-node status is when the stable attitude value is less than the preset stable attitude value or the correlation index is less than the preset correlation index.

[0063] The steady-state attitude value is the average of the steady-state reference values ​​corresponding to each sub-node. The steady-state reference value is determined as follows: for a single sub-node, the sub-node is recorded as the target sub-node. The steady-state reference value corresponding to the target sub-node is 1 / (comparison reference value + fluctuation deviation value). The fluctuation deviation value is the maximum value among the fluctuation thresholds corresponding to each data name in the target sub-node. The comparison reference value is |comparison coefficient - preset comparison coefficient|. The comparison coefficient is the standard deviation of the fluctuation thresholds corresponding to each data name in the target sub-node. The preset comparison coefficient is the average of the comparison coefficients corresponding to the sub-nodes whose steady-state attitude values ​​are greater than or equal to the preset steady-state attitude value in each historical record that can meet the user's needs. The fluctuation threshold corresponding to a single data name is the standard deviation of the value corresponding to that data name in each reconciliation data received by the target sub-node in the target monitoring period. The previous monitoring period adjacent to the current monitoring period is recorded as the target monitoring period.

[0064] The correlation index = a2 / a, where a is the average of the name reference values ​​corresponding to each data to be reconciled collected by each sub-node in the target monitoring period, a2 is the number of data names that exist in each data to be reconciled collected by each sub-node in the target monitoring period, and the name reference value corresponding to a single data to be reconciled is the number of data names that exist in that data to be reconciled.

[0065] Users can determine the preset stable attitude value and preset correlation index based on the actual application scenario. The smaller the preset stable attitude value and preset correlation index, the greater the user's demand for direct selection of a single node. The system provides a preset stable attitude value and preset correlation index, detects the historical records of users' direct selection of a single node, and records the average stable attitude value corresponding to the historical records that meet the user's needs as the preset stable attitude value, and the average correlation index corresponding to the historical records that meet the user's needs as the preset correlation index.

[0066] Specifically, if the child node status is a stable attitude value less than the preset stable attitude value or a correlation index less than the preset correlation index, then the reference data selection method is multi-node compensation selection.

[0067] Specifically, if the steady-state cross-correlation degree is greater than or equal to the preset steady-state cross-correlation degree and the steady-state equilibrium coefficient is greater than or equal to the preset steady-state equilibrium coefficient, then the compensation method is partial compensation of multiple steady-state sub-nodes.

[0068] In the multi-stable sub-node partial compensation, the selected sub-nodes are determined based on the correlation deviation degree and multidimensional covariance weights, and the feature data in each selected sub-node are selected as reference data based on the difference coefficient and class balance degree.

[0069] Among them, the sub-nodes whose comprehensive evaluation value is greater than or equal to the preset comprehensive evaluation value are recorded as steady-state sub-nodes, and the sub-nodes whose comprehensive evaluation value is less than the preset comprehensive evaluation value are recorded as unstable sub-nodes;

[0070] Steady-state cross-correlation degree = number of data names that appear in both the target data set and the reference data set / number of different data names that appear in the reference data set. The set of all accounts to be reconciled collected by all steady-state sub-nodes during the target monitoring period is denoted as the target data set, and the set of all accounts to be reconciled collected by all unstable sub-nodes during the target monitoring period is denoted as the reference data set.

[0071] Steady-state equilibrium coefficient = Number of steady-state child nodes / Total number of child nodes;

[0072] The preset values ​​of the comprehensive evaluation value, the preset steady-state cross-correlation degree, and the preset steady-state equilibrium coefficient can be determined by the user according to the actual application scenario. The greater the user's demand for improving the accuracy of reconciliation processing, the larger the values ​​of the preset comprehensive evaluation value, the preset steady-state cross-correlation degree, and the preset steady-state equilibrium coefficient. A set of preset values ​​of the comprehensive evaluation value, the preset steady-state cross-correlation degree, and the preset steady-state equilibrium coefficient is provided. The minimum value of the comprehensive evaluation value corresponding to each steady-state sub-node in the historical records that can meet the user's needs is recorded as the preset comprehensive evaluation value. The historical records that are partially compensated for multiple steady-state sub-nodes are detected. The average value of the steady-state cross-correlation degree corresponding to the historical records that can meet the user's needs is recorded as the preset steady-state cross-correlation degree. The average value of the steady-state equilibrium coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset steady-state equilibrium coefficient.

[0073] The selection of child nodes is determined based on the correlation deviation degree and the multidimensional covariance weight. The number of child nodes to be selected is determined based on the correlation deviation degree, and steady-state child nodes are selected in descending order of multidimensional covariance weight until the number of child nodes to be selected is reached. The number of child nodes to be selected is positively correlated with the correlation deviation degree.

[0074] Feature data in each selected sub-node is selected as reference data based on the difference coefficient and the class balance. Specifically, for a single selected sub-node, feature data in that selected sub-node with a difference coefficient greater than a preset difference coefficient or a class balance greater than a preset class balance is used as reference data.

[0075] Correlation deviation = Standard deviation of similarity coefficients corresponding to each steady-state sub-node + Standard deviation of fluctuation coefficients corresponding to each steady-state sub-node;

[0076] The similarity coefficient and volatility coefficient are determined as follows: For a single steady-state sub-node, the steady-state sub-node is designated as the target steady-state sub-node, and other steady-state sub-nodes are designated as reference steady-state sub-nodes. The similarity coefficient corresponding to the target steady-state sub-node is 1 - (the number of identical data names in the steady-state feature data corresponding to the target steady-state sub-node and the steady-state feature data corresponding to each reference sub-node / the number of different data names appearing in each feature data of the target steady-state sub-node). The set of feature data in the steady-state sub-node is designated as the steady-state feature data. Each steady-state sub-node corresponds to steady-state feature data. The volatility coefficient corresponding to the target steady-state sub-node is the average value of the sub-volatility coefficients corresponding to each target data name. The data names appearing in the steady-state feature data corresponding to the target steady-state sub-node are designated as target data names. The sub-volatility coefficient corresponding to a single target data name is |standard deviation of each value corresponding to the data name in the steady-state feature data corresponding to the target steady-state sub-node -standard deviation of each value corresponding to the data name in the steady-state feature data corresponding to each reference steady-state sub-node|.

[0077] The multidimensional covariant weight corresponding to a single steady-state sub-node = the fluctuation coefficient corresponding to the steady-state sub-node / the similarity coefficient corresponding to the steady-state sub-node;

[0078] Difference coefficient = number of data names appearing in the feature data / number of different data names appearing in each feature data in the target steady-state sub-node; Category balance = balance coefficient / combined bias influence.

[0079] Users can determine the preset difference coefficient and preset category balance value according to the actual application scenario. The greater the user's demand for improving the accuracy of reconciliation processing, the larger the preset difference coefficient and preset category balance value should be. A preset difference coefficient and preset category balance value is provided. The system detects the user's historical records of partial compensation for multi-stable sub-nodes, and records the average difference coefficient corresponding to the historical records that meet the user's needs as the preset difference coefficient, and records the average category balance value corresponding to the historical records that meet the user's needs as the preset category balance value.

[0080] Specifically, if the steady-state cross-correlation degree is less than the preset steady-state cross-correlation degree or the steady-state equilibrium coefficient is less than the preset steady-state equilibrium coefficient, the compensation method is to refer to the steady-state sub-node correlation compensation.

[0081] The reference steady-state child node association compensation includes:

[0082] Reference sub-nodes are determined based on the evaluation threshold, and compensation sub-nodes are selected based on the sub-influence coefficients and combinations of associated nodes.

[0083] The data compensation range for each compensation sub-node is determined based on the cross-influence threshold and the iterative correlation degree.

[0084] The supplementary selection method is determined based on the deviation reference coefficient of the data compensation range corresponding to each compensation sub-node;

[0085] All feature data in the reference child node and the feature data in the data compensation range corresponding to each compensation child node are used as reference data.

[0086] Among them, reference child nodes are determined based on evaluation thresholds, and the steady-state child node with the largest evaluation threshold is selected as the reference child node;

[0087] Compensation sub-nodes are selected based on sub-influence coefficients and associated node combinations. The associated node combinations are determined based on associated reference values. Associated node combinations with sub-influence coefficients greater than preset sub-influence coefficients are recorded as reference combinations. Any sub-node in the reference combination is used as a compensation sub-node. Each reference combination has a corresponding compensation sub-node.

[0088] Determining the associated node combination based on the associated reference value includes: recording other child nodes besides the reference child nodes as child nodes to be selected; performing association analysis on each child node to be selected; when performing association analysis on a single child node to be selected, recording the child node to be selected as the target child node to be selected; recording other child nodes to be selected besides the target child node to be selected as reference child nodes to be selected; recording the set of reference child nodes to be selected whose association reference value with the target child node is greater than the preset association reference value and the target child node to be selected as an associated node combination; and continuing to perform association analysis on child nodes to be selected that are not recorded in the associated node combination until all child nodes to be selected are recorded in the associated node combination.

[0089] The method for confirming the correlation reference value is as follows: for any two child nodes to be selected, the correlation reference value = 1 - (the absolute value of the difference between the comprehensive evaluation values ​​corresponding to the two child nodes / the larger value among the comprehensive evaluation values ​​corresponding to the two child nodes); the preset correlation reference value can be determined by the user according to the actual application scenario. The greater the user's need to improve the accuracy of reconciliation processing, the smaller the preset correlation reference value will be. One preset correlation reference value is provided, which is 70%.

[0090] The evaluation threshold for a single steady-state sub-node = the comprehensive evaluation value of the steady-state sub-node + the multidimensional covariance weight of the steady-state sub-node;

[0091] The method for confirming the sub-influence coefficient is as follows: For a single associated node combination, each sub-node in the associated node combination is recorded as an influencing sub-node. The sub-influence coefficient corresponding to the associated node combination is the average of the influence reference values ​​corresponding to each influencing sub-node and the reference sub-node. The influence reference value corresponding to a single influencing sub-node and the reference sub-node = cross-influence threshold / deviation uniformity. The cross-influence threshold = 1 - (the number of data names that are the same in the feature data corresponding to the influencing sub-node and the feature data corresponding to the reference sub-node / the number of data names that are different in the feature data corresponding to the influencing sub-node). The data names that are different in the feature data corresponding to the influencing sub-node and the feature data corresponding to the reference sub-node are recorded as difference data names. The average of the sub-difference degree corresponding to each difference data name is recorded as the deviation uniformity. The sub-difference degree corresponding to a single difference data name is the number of candidate sub-nodes corresponding to the feature data with that difference data name.

[0092] The value of the preset sub-influence coefficient can be determined by the user according to the actual application scenario. The greater the user's demand for improving the accuracy of reconciliation processing, the smaller the value of the preset sub-influence coefficient. A preset sub-influence coefficient value is provided. The historical records of reference steady-state sub-node association compensation are detected, and the average value of the sub-influence coefficients corresponding to the reference combinations that can meet the user's needs is recorded as the preset sub-influence coefficient.

[0093] The data compensation range for each compensation sub-node is determined based on the cross-influence threshold and the iterative correlation degree. The data compensation range for a single compensation sub-node is the data to be reconciled that has a data evaluation index greater than the preset data evaluation index of the compensation sub-node and is located in the compensation sub-node. The preset data evaluation index of a single compensation sub-node is positively correlated with the compensation coefficient of the compensation sub-node. The compensation coefficient of a single compensation sub-node = the cross-influence threshold of the compensation sub-node + the iterative correlation degree of the compensation sub-node.

[0094] The method for confirming the iterative correlation is as follows: for a single compensation sub-node, the compensation sub-node is recorded as the target compensation sub-node, and the other compensation sub-nodes are recorded as reference compensation sub-nodes. The iterative correlation is the average of the iterative reference values ​​corresponding to the target compensation sub-node and each reference compensation sub-node. The iterative reference value corresponding to the target compensation sub-node and a single reference compensation sub-node is 1 / the absolute value of the difference between the cross-influence threshold corresponding to the target compensation sub-node and the cross-influence threshold corresponding to the reference compensation sub-node.

[0095] The data evaluation index corresponding to a single piece of data to be reconciled = the cross-influence threshold of the data to be reconciled + the category balance of the data to be reconciled.

[0096] Specifically, the supplementary selection method is determined based on the deviation reference coefficient of the data compensation range corresponding to each compensation sub-node, including:

[0097] If the deviation reference value is greater than or equal to the preset deviation reference value, the supplementary selection method is to increase the number of compensation sub-nodes.

[0098] If the deviation reference value is less than the preset deviation reference value, the supplementary selection method is to increase the adjustment of the data compensation range.

[0099] Wherein, the deviation reference coefficient = the number of feature data in the data compensation range corresponding to each compensation sub-node / deviation similarity coefficient, the deviation similarity coefficient = 1 / |standard deviation of the data evaluation index corresponding to each feature data - preset standard deviation|, the preset standard deviation is the standard deviation of the data evaluation index corresponding to each feature data in the data compensation range corresponding to each compensation sub-node in the historical record that performs reference steady-state sub-node association compensation and can meet user needs.

[0100] The user can determine the value of the preset deviation reference value according to the actual application scenario. The larger the value of the preset deviation reference value, the greater the user's need to increase the adjustment of the data compensation range. The system provides a preset deviation reference value, detects the historical records of the user's adjustment of the data compensation range, and records the average value of the deviation reference values ​​corresponding to the historical records that meet the user's needs as the preset deviation reference value.

[0101] When increasing the number of compensation sub-nodes, the increase in the number of compensation sub-nodes is positively correlated with the deviation reference value. Adjustment sub-nodes are selected in descending order of influence on the reference value until the number of compensation sub-nodes after the increase is reached. The adjustment sub-nodes are other sub-nodes to be selected besides the compensation sub-nodes.

[0102] When increasing the data compensation range, the adjustment should be made for the data compensation range corresponding to each compensation sub-node. When increasing the data compensation range for a single compensation sub-node, the adjustment should be made for the preset data evaluation index corresponding to that compensation sub-node. The reduction value of the preset data evaluation index corresponding to a single compensation sub-node is positively correlated with the deviation reference value.

[0103] Specifically, the clustering method is determined based on the data variability and increment coefficient of each child node, including:

[0104] For a single child node,

[0105] If the data variability is greater than or equal to the preset data variability or the increment coefficient is greater than or equal to the preset increment coefficient, the clustering method is to split the cluster according to the variability coefficient.

[0106] If the data variation is less than the preset data variation and the increment coefficient is less than the preset increment coefficient, then the clustering method is to perform overall clustering based on the correlation coefficient.

[0107] The data anomaly and incremental coefficient are determined as follows: For a single sub-node, the data anomaly is the maximum value among the fluctuation thresholds corresponding to each data name appearing in each reconciliation data collected by the sub-node during the target monitoring period; the incremental coefficient is equal to |the average value of the fluctuation thresholds corresponding to each data name appearing in each reconciliation data collected by the sub-node - the average value of the adjacent fluctuation thresholds corresponding to each data name appearing in each reconciliation data collected by the sub-node|; the adjacent fluctuation threshold corresponding to a single data name is the standard deviation of the value corresponding to that data name in each reconciliation data collected by the sub-node in the previous monitoring period adjacent to the target monitoring period.

[0108] Users can determine the preset values ​​of data anomaly and preset incremental coefficient based on their actual application scenarios. The greater the user's need to improve the accuracy of reconciliation processing, the smaller the preset values ​​of data anomaly and preset incremental coefficient will be. This provides a preset value of data anomaly and preset incremental coefficient. The system detects the historical records of the entire cluster based on the correlation coefficient, and records the average value of the data anomaly corresponding to the historical records that meet the user's needs as the preset data anomaly, and records the average value of the incremental coefficient corresponding to the historical records that meet the user's needs as the preset incremental coefficient.

[0109] Cluster splitting based on the coefficient of variation includes: denoteing the number of data names in a single name combination as n, where n is negatively correlated with the coefficient of variation (coefficient of variation = data anomaly + increment coefficient); sorting the data names in descending order of outlier reference values ​​and recording this as a reference sequence; starting from the leftmost end of the reference sequence, selecting the first n data names to form a name combination; repeating this selection process, each time selecting n more data names from the remaining sequence to form a name combination; when the number of remaining data names is less than n, recording all remaining data names as a separate name combination; each reconciliation data set corresponds to several split groups. In summary, a single splitting combination contains the data names and corresponding values ​​corresponding to a single name combination. Split cluster analysis is performed on the splitting combinations corresponding to each name combination. When performing split cluster analysis on a single splitting combination corresponding to a single name combination, the splitting combination is recorded as the target splitting combination, and other splitting combinations other than the target splitting combination corresponding to the name combination are recorded as reference splitting combinations. The set of reference splitting combinations with a combination matching degree greater than the preset combination matching degree and the target splitting combination is recorded as a splitting cluster. The splitting cluster analysis continues for the splitting combinations not recorded in the splitting cluster until all splitting combinations are recorded in the splitting cluster.

[0110] The abnormal reference value for a single data name = |the fluctuation threshold corresponding to the data name - the neighboring fluctuation threshold corresponding to the data name|;

[0111] The overall clustering is performed based on the correlation coefficient, including: performing overall cluster analysis on each piece of reconciliation data; when performing overall cluster analysis on a single piece of reconciliation data, the reconciliation data is recorded as the target reconciliation data, and other reconciliation data other than the target reconciliation data are recorded as reference reconciliation data; the set of reference reconciliation data with a correlation coefficient greater than the preset correlation coefficient and the target reconciliation data is recorded as an overall cluster, and the cluster analysis is continued on the reconciliation data not recorded in the overall cluster until all reconciliation data are recorded in the overall cluster;

[0112] The method for confirming the combination matching degree is as follows: for the two split combinations corresponding to a single name combination, the combination matching degree is the standard deviation of the difference coefficients corresponding to each data name in the name combination, and the difference coefficient corresponding to a single data name is the absolute value of the difference between the values ​​corresponding to the data name in the two split combinations.

[0113] For any two data points pending reconciliation, the correlation coefficient = 1 - (absolute value of the difference between the data evaluation indices corresponding to the two data points pending reconciliation / the larger value among the data evaluation indices corresponding to the two data points pending reconciliation).

[0114] Users can determine the values ​​of the preset combination matching degree and preset correlation coefficient according to the actual application scenario. The greater the user's demand for improving the accuracy of reconciliation processing, the higher the values ​​of the preset combination matching degree and preset correlation coefficient. The detection detects the historical records of the cluster split according to the coefficient of variation, and records the average combination matching degree corresponding to each historical record that meets the user's needs as the preset combination matching degree. The preset correlation coefficient is 80%.

[0115] Specifically, the feature data refers to the reconciliation data where the combined bias influence is less than the preset combined bias influence and the balance coefficient is greater than or equal to the preset balance coefficient.

[0116] The method for confirming the combined bias impact is as follows: For a single data to be reconciled, the data to be reconciled is recorded as the target data to be reconciled. If the target data to be reconciled corresponds to a split cluster, the combined bias impact is the maximum value among the bias reference values ​​of each split cluster corresponding to the target data to be reconciled. If the target data to be reconciled corresponds to an overall cluster, the combined bias impact is the bias reference value of the overall cluster corresponding to the target data to be reconciled. The method for confirming the bias reference value corresponding to a single cluster is as follows: the bias reference value corresponding to the cluster = |the sub-bias degree corresponding to the cluster - the preset sub-bias degree corresponding to the cluster|. The sub-bias degree corresponding to a single cluster is the maximum value among the abnormal reference values ​​corresponding to each reference name. Each data name appearing in the cluster is recorded as a reference name. The abnormal reference value corresponding to a single reference name = |the average value of the value corresponding to the reference name in the cluster - the average value of the value corresponding to the reference name in the associated cluster|. Clusters corresponding to feature data with the same data names as those contained in the cluster in the historical records are detected and recorded as associated clusters. The preset sub-bias degree corresponding to a single cluster is the average value of the sub-bias degrees corresponding to each associated cluster.

[0117] The balance coefficient corresponding to a single piece of data to be reconciled is the average of the sub-balance coefficients corresponding to each data name appearing in that piece of data. The sub-balance coefficient corresponding to a single data name = 1 / |the value corresponding to that data name in the data to be reconciled - the average value of the values ​​corresponding to that data name in the child node where the data to be reconciled is located|;

[0118] Users can determine the values ​​of the preset combined bias influence degree and the preset balance coefficient according to the actual application scenario. The greater the user's demand for improving the accuracy of reconciliation processing, the smaller the values ​​of the preset combined bias influence degree and the preset balance coefficient should be. One preset value for the preset combined bias influence degree and the preset balance coefficient is defined as the average value of the combined bias influence degree corresponding to each feature data in the historical records that can meet the user's needs.

[0119] Specifically, the selection method for reconciliation sub-nodes is determined based on the reference comparison coefficient and cross-relation coefficient between the data to be reconciled and the reference data corresponding to the sub-nodes, including:

[0120] If the reference comparison coefficient is greater than or equal to the preset reference comparison coefficient and the cross-correlation number is greater than or equal to the preset cross-correlation number, then the sub-node selection method is based on the interaction influence coefficient.

[0121] If the reference comparison coefficient is less than the preset reference comparison coefficient or the cross-correlation number is less than the preset cross-correlation number, the reconciliation sub-node selection method is to select based on the comparison threshold.

[0122] Wherein, the reference comparison coefficient = (the number of child nodes whose comparison threshold is greater than the preset comparison threshold) / the total number of child nodes;

[0123] The comparison threshold is determined as follows: for a single child node, the maximum value among the sub-comparison reference values ​​corresponding to each data name appearing in the child node is recorded as the comparison threshold; the sub-comparison reference value corresponding to a single data name = |the fluctuation threshold corresponding to the data name in the child node - the standard deviation of the values ​​corresponding to each reference data appearing with the data name|.

[0124] The cross-correlation coefficient is the standard deviation of the comparison threshold corresponding to each pre-reconciliation node. Sub-nodes whose comparison threshold is greater than the preset comparison threshold are recorded as pre-reconciliation nodes.

[0125] Users can determine the values ​​of the preset comparison threshold, preset reference comparison coefficient, and preset cross-correlation coefficient according to the actual application scenario. The smaller the values ​​of the preset comparison threshold, preset reference comparison coefficient, and preset cross-correlation coefficient, the greater the user's need for optimization detection selection. The average value of the comparison threshold corresponding to each pre-reconciliation node in the historical records that can meet the user's needs is recorded as the preset comparison threshold. The average value of the reference comparison coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset reference comparison coefficient, and the average value of the cross-correlation coefficient corresponding to the historical records that can meet the user's needs is recorded as the preset cross-correlation coefficient.

[0126] When selecting based on the interaction impact coefficient, the pre-reconciliation node with an interaction impact coefficient greater than the preset interaction impact coefficient is selected as the reconciliation sub-node.

[0127] When selecting a comparison threshold, child nodes whose comparison threshold is greater than the preset comparison threshold are used as reconciliation child nodes.

[0128] For a single pre-reconciliation node, the pre-reconciliation node is designated as the target pre-reconciliation node, and other pre-reconciliation nodes other than the target pre-reconciliation node are designated as reference pre-reconciliation nodes. The interaction impact coefficient corresponding to the target pre-reconciliation node is the average value of the sub-comparison reference values ​​corresponding to the target pre-reconciliation node for data names that appear in both the target pre-reconciliation node and the reference pre-reconciliation node / the number of data names that appear in both the target pre-reconciliation node and the reference pre-reconciliation node.

[0129] The value of the preset interaction impact coefficient can be determined by the user based on the actual application scenario. The greater the user's need to improve the accuracy of reconciliation processing, the smaller the value of the preset interaction impact coefficient. The system detects the historical records selected by the user based on the interaction impact coefficient, and records the average value of the interaction impact coefficients corresponding to each reconciliation sub-node selected from the historical records that meet the user's needs as the preset interaction impact coefficient.

[0130] Please see Figure 4 The diagram shown is a module connection diagram of the multi-dimensional reconciliation system based on a group meal platform according to the present invention. The present invention also provides a multi-dimensional reconciliation system based on a group meal platform, comprising:

[0131] The data acquisition module includes several sub-nodes for data acquisition.

[0132] The status analysis module, which is connected to the data acquisition module, is used to determine the status of the sub-nodes based on the stable state values ​​and correlation indices, and to determine the reference data selection method based on the sub-node status. The reference data selection method is either direct selection of a single node or compensation selection of multiple nodes.

[0133] The first selection module, which is connected to the status analysis module, is used to select the reconciliation data corresponding to the sub-node with the largest comprehensive evaluation value as reference data in the direct selection of a single node.

[0134] The second selection module, which is connected to the state analysis module, is used to determine the compensation method and select reference data in the multi-node compensation selection based on the steady-state cross-correlation degree and steady-state equilibrium coefficient. The compensation method is partial compensation of multiple steady-state sub-nodes or correlation compensation of reference steady-state sub-nodes.

[0135] The reconciliation selection module is connected to the first selection module and the second selection module respectively. It is used to determine the reconciliation sub-node selection method based on the reference comparison coefficient and cross-relation coefficient of the data to be reconciled and the reference data corresponding to the sub-node. The reconciliation sub-node selection method is selected based on the interaction influence coefficient or the comparison threshold.

[0136] The feature data is determined based on the combined bias influence degree and the balance coefficient. The combined bias influence degree is determined based on the bias reference value of the cluster corresponding to the data to be reconciled.

[0137] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0138] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-dimensional reconciliation method based on a group meal platform, characterized in that, include: The status of the sub-nodes is determined based on the stable attitude value and the correlation index. The method of selecting reference data is determined based on the status of the sub-nodes. The method of selecting reference data is either direct selection of a single node or compensation selection of multiple nodes. Among them, the steady-state reference value is the average of the steady-state reference values ​​corresponding to each sub-node. The steady-state reference value is determined as follows: for a single sub-node, the sub-node is recorded as the target sub-node. The steady-state reference value corresponding to the target sub-node is 1 / (comparison reference value + fluctuation deviation value). The fluctuation deviation value is the maximum value among the fluctuation thresholds corresponding to each data name in the target sub-node. The comparison reference value is |comparison coefficient - preset comparison coefficient|. The comparison coefficient is the standard deviation of the fluctuation thresholds corresponding to each data name in the target sub-node. The fluctuation threshold corresponding to a single data name is the standard deviation of the value corresponding to the data name in each reconciliation data received by the target sub-node in the target monitoring period. The correlation index is a2 / a. The average of the name reference values ​​corresponding to each reconciliation data collected by each sub-node in the target monitoring period is recorded as a. The number of data names that exist in each reconciliation data collected by each sub-node in the target monitoring period is recorded as a2. The name reference value corresponding to a single reconciliation data is the number of data names that exist in the reconciliation data. The sub-node states include steady-state values ​​less than preset steady-state values ​​or correlation indices less than preset correlation indices, and steady-state cross-correlation degrees greater than or equal to preset steady-state cross-correlation degrees and steady-state equilibrium coefficients greater than or equal to preset steady-state equilibrium coefficients. In the direct selection of a single node, the reconciliation data corresponding to the sub-node with the largest comprehensive evaluation value is used as the reference data. In the selection of multi-node compensation, the compensation method is determined based on the steady-state cross-correlation degree and steady-state equilibrium coefficient to select reference data. The compensation method is partial compensation of multiple steady-state sub-nodes or correlation compensation of reference steady-state sub-nodes. Wherein, steady-state cross-correlation degree = number of data names appearing in both the target data set and the reference data set / number of different data names appearing in the reference data set; the set of all reconciliation data collected by all steady-state sub-nodes during the target monitoring period is denoted as the target data set, and the set of all reconciliation data collected by all unstable sub-nodes during the target monitoring period is denoted as the reference data set; steady-state equilibrium coefficient = number of steady-state sub-nodes / total number of sub-nodes; In the multi-stable sub-node partial compensation, the selected sub-nodes are determined based on the correlation deviation degree and multidimensional covariance weights, and the feature data in each selected sub-node are selected as reference data based on the difference coefficient and class balance degree. Wherein, the correlation deviation = standard deviation of the similarity coefficients corresponding to each steady-state sub-node + standard deviation of the fluctuation coefficients corresponding to each steady-state sub-node; the similarity coefficient corresponding to the target steady-state sub-node = 1 - (the number of identical data names in the steady-state feature data corresponding to the target steady-state sub-node and the steady-state feature data corresponding to each reference sub-node / the number of different data names appearing in each feature data of the target steady-state sub-node), the set of feature data in the steady-state sub-node is denoted as steady-state feature data, and the fluctuation coefficient corresponding to the target steady-state sub-node is the average of the sub-fluctuation coefficients corresponding to each target data name, and the number of different data names appearing in each feature data of the target steady-state sub-node is denoted as steady-state feature data. The data name is denoted as the target data name. The sub-fluctuation coefficient corresponding to a single target data name is equal to |the standard deviation of each value corresponding to the target steady-state sub-node in the steady-state feature data - the standard deviation of each value corresponding to the reference steady-state sub-node in the steady-state feature data|. The multidimensional covariance weight corresponding to a single steady-state sub-node is equal to the fluctuation coefficient corresponding to the steady-state sub-node / the similarity coefficient corresponding to the steady-state sub-node. The difference coefficient is equal to the number of data names appearing in the feature data / the number of different data names appearing in each feature data of the target steady-state sub-node. The category balance is equal to the balance coefficient / the combined bias influence. In the reference steady-state sub-node correlation compensation, reference sub-nodes are determined based on the evaluation threshold, and compensation sub-nodes are selected based on the sub-influence coefficient and the combination of related nodes; the data compensation range corresponding to each compensation sub-node is determined based on the cross-influence threshold and the iterative correlation degree; and the supplementary selection method is determined based on the deviation reference coefficient of the data compensation range corresponding to each compensation sub-node. If the deviation reference value is greater than or equal to the preset deviation reference value, the supplementary selection method is to increase the number of compensation sub-nodes. If the deviation reference value is less than the preset deviation reference value, the supplementary selection method is to increase the data compensation range by using all feature data in the reference sub-node and the feature data in the data compensation range corresponding to each compensation sub-node as reference data. The supplementary selection method is determined based on the deviation reference coefficient of the data compensation range corresponding to each compensation sub-node; The evaluation threshold for a single steady-state sub-node is calculated as follows: the comprehensive evaluation value of the steady-state sub-node is equal to the multidimensional covariance weight of the steady-state sub-node. The sub-influence coefficient is determined by denoting each sub-node in a single associated node combination as an influencing sub-node. The sub-influence coefficient of this associated node combination is the average of the influence reference values ​​of each influencing sub-node and the reference sub-node. The influence reference value of a single influencing sub-node and the reference sub-node is calculated as: cross-influence threshold / deviation average. The cross-influence threshold is calculated as: 1 - (the number of identical data names in the feature data of the influencing sub-node and the feature data of the reference sub-node / the number of different data names in the feature data of the influencing sub-node). The different data names in the feature data of the influencing sub-node and the feature data of the reference sub-node are denoted as difference data names. The sub-differences corresponding to each difference data name are then denoted as... The average value of the degree is denoted as the deviation uniformity degree. The sub-difference degree corresponding to a single difference data name is the number of candidate sub-nodes corresponding to the feature data with that difference data name. The associated node combination is determined based on the association reference value, which is 1 - (the absolute value of the difference between the comprehensive evaluation values ​​corresponding to two sub-nodes / the larger of the comprehensive evaluation values ​​corresponding to two sub-nodes). The iterative association degree is the average value of the iterative reference values ​​corresponding to the target compensation sub-node and each reference compensation sub-node. The iterative reference value corresponding to the target compensation sub-node and a single reference compensation sub-node is 1 / the absolute value of the difference between the cross-influence threshold corresponding to the target compensation sub-node and the cross-influence threshold corresponding to the reference compensation sub-node. The deviation reference coefficient is the number of feature data in the data compensation range corresponding to each compensation sub-node / the deviation similarity coefficient. The deviation similarity coefficient is 1 / |the standard deviation of the data evaluation index corresponding to each feature data - the preset standard deviation|. The selection method for reconciliation sub-nodes is determined based on the reference comparison coefficient and cross-relation coefficient between the data to be reconciled and the reference data corresponding to the sub-nodes. The selection method for reconciliation sub-nodes is selected based on the interaction influence coefficient or comparison threshold. Wherein, the reference comparison coefficient = (number of child nodes whose comparison threshold is greater than the preset comparison threshold) / total number of child nodes; the comparison threshold is determined by, for a single child node, the maximum value among the sub-comparison reference values ​​corresponding to each data name appearing in the child node is recorded as the comparison threshold; the sub-comparison reference value corresponding to a single data name = |the fluctuation threshold corresponding to the data name in the child node - the standard deviation of the values ​​corresponding to each reference data appearing with the data name|; the cross-correlation coefficient is the standard deviation of the comparison threshold corresponding to each pre-reconciliation node, and child nodes whose comparison threshold is greater than the preset comparison threshold are recorded as pre-reconciliation nodes; The interaction impact coefficient corresponding to the target pre-reconciliation node = the average value of the sub-comparison reference value corresponding to the target pre-reconciliation node for data names that appear in both the target pre-reconciliation node and the reference pre-reconciliation node / the number of data names that appear in both the target pre-reconciliation node and the reference pre-reconciliation node; The feature data is determined based on the combined bias influence degree and the balance coefficient. The combined bias influence degree is determined based on the bias reference value of the cluster corresponding to the data to be reconciled. The method for confirming the combined bias impact is as follows: For a single piece of reconciliation data, this piece of data is designated as the target reconciliation data. If the target reconciliation data corresponds to a split cluster, the combined bias impact is the maximum value among the bias reference values ​​corresponding to each split cluster of the target reconciliation data. If the target reconciliation data corresponds to an overall cluster, the combined bias impact is the bias reference value of the overall cluster corresponding to the target reconciliation data. The bias reference value corresponding to a single cluster = |the sub-bias degree corresponding to the cluster - the preset sub-bias degree corresponding to the cluster|, and the sub-bias degree corresponding to a single cluster is the maximum value among the abnormal reference values ​​corresponding to each reference name. Each data name appearing in the data is denoted as a reference name. The abnormal reference value corresponding to a single reference name is equal to |the average value of the reference name in the cluster - the average value of the reference name in the associated clusters|. Clusters with the same feature data as the data names contained in the cluster are detected in the historical records and are denoted as associated clusters. The balance coefficient corresponding to a single data to be reconciled is the average of the sub-balance coefficients corresponding to each data name appearing in the data to be reconciled. The sub-balance coefficient corresponding to a single data name is 1 / |the value of the data name in the data to be reconciled - the average value of the data name in the child node where the data to be reconciled is located|.

2. The multi-dimensional reconciliation method based on a group meal platform according to claim 1, characterized in that, If the child node's status is a stable attitude value greater than or equal to the preset stable attitude value and the correlation index is greater than or equal to the preset correlation index, then the reference data selection method is to directly select a single node.

3. The multi-dimensional reconciliation method based on a group meal platform according to claim 2, characterized in that, If the child node status is a stable attitude value less than the preset stable attitude value or a correlation index less than the preset correlation index, then the reference data selection method is multi-node compensation selection.

4. The multi-dimensional reconciliation method based on a group meal platform according to claim 3, characterized in that, If the steady-state cross-correlation degree is greater than or equal to the preset steady-state cross-correlation degree and the steady-state equilibrium coefficient is greater than or equal to the preset steady-state equilibrium coefficient, then the compensation method is partial compensation of multiple steady-state sub-nodes.

5. The multi-dimensional reconciliation method based on a group meal platform according to claim 4, characterized in that, If the steady-state cross-correlation degree is less than the preset steady-state cross-correlation degree or the steady-state equilibrium coefficient is less than the preset steady-state equilibrium coefficient, the compensation method is reference steady-state sub-node correlation compensation.

6. The multi-dimensional reconciliation method based on a group meal platform according to claim 1, characterized in that, The clustering method is determined based on the data variability and increment coefficient of each child node, including: For a single child node, If the data variability is greater than or equal to the preset data variability or the increment coefficient is greater than or equal to the preset increment coefficient, the clustering method is to split the cluster according to the variability coefficient. If the data variation is less than the preset data variation and the increment coefficient is less than the preset increment coefficient, then the clustering method is to perform overall clustering based on the correlation coefficient.

7. The multi-dimensional reconciliation method based on a group meal platform according to claim 1, characterized in that, The characteristic data refers to the reconciliation data where the combined bias influence is less than the preset combined bias influence and the balance coefficient is greater than or equal to the preset balance coefficient.

8. The multi-dimensional reconciliation method based on a group meal platform according to claim 1, characterized in that, The selection method for reconciliation sub-nodes is determined based on the reference comparison coefficients and cross-relation coefficients between the data to be reconciled and the reference data corresponding to the sub-nodes, including: If the reference comparison coefficient is greater than or equal to the preset reference comparison coefficient and the cross-correlation number is greater than or equal to the preset cross-correlation number, then the sub-node selection method is based on the interaction influence coefficient. If the reference comparison coefficient is less than the preset reference comparison coefficient or the cross-correlation number is less than the preset cross-correlation number, the reconciliation sub-node selection method is to select based on the comparison threshold.

9. A reconciliation system applying the multi-dimensional reconciliation method based on a group meal platform as described in any one of claims 1 to 8, characterized in that, include: The data acquisition module includes several sub-nodes for data acquisition. The status analysis module, which is connected to the data acquisition module, is used to determine the status of the sub-nodes based on the stable state values ​​and correlation indices, and to determine the reference data selection method based on the sub-node status. The reference data selection method is either direct selection of a single node or compensation selection of multiple nodes. The first selection module, which is connected to the status analysis module, is used to select the reconciliation data corresponding to the sub-node with the largest comprehensive evaluation value as reference data in the direct selection of a single node. The second selection module, which is connected to the state analysis module, is used to determine the compensation method and select reference data in the multi-node compensation selection based on the steady-state cross-correlation degree and steady-state equilibrium coefficient. The compensation method is partial compensation of multiple steady-state sub-nodes or correlation compensation of reference steady-state sub-nodes. The reconciliation selection module is connected to the first selection module and the second selection module respectively. It is used to determine the reconciliation sub-node selection method based on the reference comparison coefficient and cross-relation coefficient of the data to be reconciled and the reference data corresponding to the sub-node. The reconciliation sub-node selection method is selected based on the interaction influence coefficient or the comparison threshold. The feature data is determined based on the combined bias influence degree and the balance coefficient. The combined bias influence degree is determined based on the bias reference value of the cluster corresponding to the data to be reconciled.

Images