Big data-based financial multi-dimensional dynamic analysis method and system

Abstract

The application discloses a financial multi-dimensional dynamic analysis method and system based on big data, and particularly relates to the field of financial data processing, which comprises the following steps: obtaining local business records collected to a center platform via edge data transmission, performing grouping according to business object identification and business number, and performing sorting according to occurrence time and source node sequence in each group to generate a candidate record sequence; the application performs transaction formation judgment, attribution branch checking, content migration identification and analysis segment merging on the local business records collected via edge data transmission, so as to solve the problem of how to avoid directly including local business records which have not reached the analysis formation condition and whose financial attributes are still in the changing process into the financial multi-dimensional dynamic analysis under the condition of edge data transmission.

Description

Technical Field

[0001] This invention relates to the field of financial data processing technology, and more specifically, to a method and system for multidimensional dynamic financial analysis based on big data. Background Technology

[0002] In current corporate financial analysis, the focus is usually on quickly generating dynamic analysis results that can be used for business management. A common approach is to continuously aggregate data from different business systems and terminal nodes, such as sales, procurement, inventory, reimbursement, invoicing, receipts and payments, and general ledger entries, to the central platform via edge data transmission. Then, the data is associated, collected, and dynamically updated according to organizational units, projects, customers, categories, cost centers, and accounting periods. In chain store scenarios, store POS terminals, warehouse terminals, mobile expense reimbursement terminals, invoicing terminals, and bank-enterprise interfaces need to continuously upload data under conditions of limited bandwidth, inconsistent upload order with the actual business occurrence order, and the same business transaction continuing to undergo supplementary entry, red-ink reversal, reclassification, and cross-period confirmation. Under these conditions, although the central platform can obtain analytical values ​​such as revenue, cost, expenses, and profit relatively quickly, in actual operation, phenomena such as repeated fluctuations in the profit contribution of the same store in a short period of time, the same expense falling into different cost centers at different times, customer payment contributions being included in statistics but contract fulfillment attribution not yet being completed, and abnormal fluctuation prompts being rewritten after subsequent invoice completion are frequently observed. The root cause is that the existing processing usually directly uses partial business records that have been transmitted to the central platform as input for financial analysis without first determining whether the partial business records have met the conditions required for the current analysis, or further distinguishing the changes in the financial attributes of the same business transaction at different confirmation stages. As a result, the object on which the subsequent multi-dimensional dynamic analysis is based remains in an unstable and undetermined state. Therefore, the technical problem to be solved by this application is: how to avoid directly including local business records that have not yet met the conditions for analysis and whose financial attributes are still changing under the condition of edge data transmission into the financial multidimensional dynamic analysis. Summary of the Invention

[0003] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method and system for multidimensional dynamic financial analysis based on big data. This method and system solves the problems mentioned in the background art by determining the execution of local business records collected through edge data transmission, verifying the attribution of branches, identifying content migration, and merging analysis fragments.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional dynamic financial analysis method based on big data, comprising: S1. Obtain local business records that are collected from edge data transmission to the central platform, group them according to business object identifier and business number, and sort them within each group according to the occurrence time and source node order to generate a candidate record sequence; S2. For local business records in the candidate record sequence, the business content, confirmation content and attribution content are merged according to the sorting order. Duplicate content with the same content is merged and retained. Conflicting content with different content is recorded in parallel to generate an event record. S3. Based on the required content items corresponding to the current analysis type, perform a completeness check and a unique attribution check on the item record, and write the business content, confirmed content, and attribution content that have passed the check, as well as the missing content that has failed the check, into the judgment record. S4. For the business content, confirmed content and belonging content that have been verified in the judgment record, perform adjacent comparison according to the sorting order in the candidate record sequence, and write the corresponding content item, the preceding content, the following content and the sorting position for the positions where the content is inconsistent between the two local business records, and generate migration record. S5. Based on the verified business content, confirmed content, and attribution content in the judgment record, merge the local business records in the migration record that are consecutive in the sorting position and have the same attribution content into the same analysis segment, and then perform multidimensional aggregation and time series statistics on each analysis segment to generate financial multidimensional dynamic analysis results.

[0005] In a preferred embodiment, S1 includes: S1-1. Extract the business object identifier and business number from the local business records, and group the local business records with the same business object identifier and business number to generate a record group. S1-2. Sort the local business records in each record group in ascending order according to the time of occurrence, and sort the local business records with the same time of occurrence in order according to the source node order to generate the initial sequence; S1-3. Read the previous local business record and the next local business record in the order of arrangement for each initial sequence. Perform consistency comparison on the related content of the two local business records. Keep the local business records with consistent related content in the same arrangement sequence. Break the local business records with inconsistent related content from the inconsistent position and form a new arrangement sequence to generate a candidate record sequence.

[0006] In a preferred embodiment, S2 includes: S2-1. Read the local business records in the candidate record sequence according to the sorting order, extract the business content corresponding to the business occurrence, the confirmation content corresponding to the financial confirmation, and the dimension from each local business record and classify them into the corresponding belonging content. Write the local business record corresponding to the first appearance of each content into the event record to generate the initial event record. S2-2. Continue reading the local business records after the initial event record, compare the business content, confirmation content and attribution content in the local business records with the corresponding content in the initial event record item by item, and add the corresponding content that has not yet been written in the initial event record to the event record to generate the supplemented event record.

[0007] In a preferred embodiment, S2 further includes: S2-3. Sequentially check the local business records after the supplementary item record, compare the business content, confirmation content and attribution content in the local business records with the corresponding content in the supplementary item record item by item, merge the local business records with the corresponding content, and generate a merged item record. S2-4. Continue to check the local business records after the merged item record. Compare the business content, confirmation content and attribution content in the local business records with the corresponding content in the merged item record item by item. Record the local business records with different content in parallel into the corresponding content to generate the item record.

[0008] In a preferred embodiment, S3 includes: S3-1. Read the required content items corresponding to the current analysis type, extract the business content, confirmation content, attribution content and corresponding sorting position from the event record, check whether the corresponding content exists in the event record item by item according to the required content items, if the corresponding content exists, write the corresponding content and corresponding sorting position into the checked item, if the corresponding content does not exist, write the corresponding required content item into the missing item, and generate the initial check record. S3-2. For item records where missing items are not written in the initial verification record, construct affiliation branches according to the affiliation content and corresponding sorting position, and perform acceptance verification, confirmation verification, and business verification on each affiliation branch. Among them, acceptance verification is used to check whether other affiliation content is written between adjacent sorting positions in the same affiliation branch. If no other affiliation content is written, an acceptance item is written in the corresponding position. If other affiliation content is written, a breakpoint item is written in the corresponding position. Confirmation verification is used to check whether each affiliation content corresponds to the confirmation content in the same item record at the corresponding sorting position. If it corresponds to the confirmation content, a confirmation item is written. If it does not correspond to the confirmation content, a mismatch item is written. Business verification is used to check whether each affiliation content corresponds to the business content in the same item record at the corresponding sorting position. If it corresponds to the business content, a consistency item is written. If it does not correspond to the business content, a conflict item is written. Generate a branch verification record.

[0009] In a preferred embodiment, S3 further includes: S3-3. Perform cross-determination on each attribution branch corresponding to the same item record. First, count the total number of breakpoint items, mismatch items, and conflict items in each attribution branch, and retain the attribution branch with the highest total number. If there is more than one remaining attribution branch, continue to count the number of receiving items in the remaining attribution branches, and retain the attribution branch with the highest number of receiving items. If there is still more than one remaining attribution branch, compare the starting sort positions of the remaining attribution branches and retain the attribution branch with the highest starting sort position. If there is only one remaining attribution branch, write that attribution branch into the passing branch. If there is more than one remaining attribution branch, write all remaining attribution branches into the return check branch and generate intermediate records. S3-4. Based on the intermediate records, write the business content, confirmation content, and attribution content corresponding to the item records written to the passing branch into the passing content, write the required content items corresponding to the item records with missing items into the missing content, and write the attribution content and corresponding sorting position corresponding to the item records written to the back check branch into the back check content, and generate a judgment record.

[0010] In a preferred embodiment, S4 includes: S4-1. Read the verified business content, confirmed content, and attribution content from the judgment record, and match each item to each local business record according to the sorting order in the candidate record sequence to generate a comparison sequence; S4-2. Read the previous local business record and the next local business record in the sorting order along the comparison sequence. Perform same-item comparison on the business content, confirmation content and belonging content in the previous local business record and the next local business record. Write the corresponding content items, the previous content, the next content and the sorting position of the next local business record that are inconsistent into the change item to generate the initial change sequence.

[0011] In a preferred embodiment, S4 further includes: S4-3. Read each change item in the initial change sequence in order of sorting position. Perform a follow-up check on the content after the previous change item and the content before the next change item. If the follow-up check is consistent, the previous change item and the next change item are assigned to the same migration segment. If the follow-up check is inconsistent, the next change item is assigned to another migration segment, and a migration segment sequence is generated. S4-4. Perform a continuous sorting position check on each migration segment in the migration segment sequence, and organize the migration segments with continuous sorting positions and the migration segments with non-contiguous sorting positions into migration content to generate migration records.

[0012] In a preferred embodiment, S5 includes: S5-1. Extract the verified business content, confirmed content, and attribution content from the judgment record, and combine them with the sorting position and migration content in the migration record. Group the local business records with consecutive sorting positions and consistent attribution content into the same analysis segment to generate an analysis segment sequence. S5-2. Sequentially process each analysis segment in the analysis segment sequence, split the content of each analysis segment into corresponding category combinations, merge according to the same category combination, and count the number of local business records, starting sort position, ending sort position, business content and confirmation content corresponding to each category combination, and generate a summary record sequence. S5-3. Then, arrange each summary record according to the starting and ending sorting positions in the summary record sequence. For two adjacent summary records, calculate the changes in the business content, confirmation content, and attribution item combinations, and generate financial multidimensional dynamic analysis results.

[0013] In a preferred embodiment, the big data-based multidimensional dynamic financial analysis system includes: The record grouping module is used to acquire local business records that are collected to the central platform via edge data transmission, group them according to business object identifier and business number, and sort them within each group according to the occurrence time and source node order to generate candidate record sequences. The event generation module is used to merge the business content, confirmation content and attribution content of local business records in the candidate record sequence according to the sorting order, merge and retain duplicate content with the same content, and record conflicting content with different content in parallel to generate event records. The condition judgment module performs a completeness check and a unique attribution check on the item records based on the required content items corresponding to the current analysis type. The business content, confirmed content, and attribution content that pass the check, as well as the missing content that fails the check, are all written into the judgment record. The migration identification module is used to perform adjacent comparisons on the business content, confirmed content and belonging content that have passed the verification in the judgment record according to the sorting order in the candidate record sequence. For the positions where the content is inconsistent between two local business records, the corresponding content item, the preceding content, the following content and the sorting position are written to generate a migration record. The analysis output module merges local business records with consecutive sort positions and consistent attribution in the migration records into the same analysis segment based on the verified business content, confirmation content, and attribution content in the judgment record. Then, it performs multidimensional aggregation and time-series statistics on each analysis segment to generate financial multidimensional dynamic analysis results.

[0014] The technical effects and advantages of this invention are as follows: By first grouping candidates by business object identifier and business number, and then organizing them by occurrence time, source node order and related content, we can avoid directly entering local business records that have not formed stable events into the analysis. By performing initial writing, gap filling, duplicate merging, and conflict parallel recording on the business content, confirmation content, and attribution content in the candidate record sequence, a traceable record of events is formed and the basis for subsequent verification is improved. By performing a complete content check based on the required content items of the current analysis type, and constructing attribution branches around the attribution content, multiple rounds of cross-judgment are carried out, which relatively reduces the probability of undetermined attribution items entering the analysis results layer. By mapping the approved content back to the candidate record sequence, and organizing it according to adjacent comparison, succession verification and migration segment to generate migration records, the identification and retention of the financial attribute change process is relatively improved. By grouping consecutive and consistent local business records into analysis segments, and then merging them according to their category and statistically analyzing changes in position and order, the phenomenon of repeated rewriting of analysis results can be alleviated. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method steps of the present invention.

[0016] Figure 2 This is a schematic diagram of the system modules of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Refer to the instruction manual appendix Figure 1-2 The present invention provides a big data-based multidimensional dynamic financial analysis method, comprising: S1. Obtain local business records that are collected from edge data transmission to the central platform, group them according to business object identifier and business number, and sort them within each group according to the occurrence time and source node order to generate a candidate record sequence; In this embodiment, S1 is used to separate the continuous record range corresponding to the same business item from the local business records received from the central platform, and organize the continuous record range into a candidate record sequence for subsequent item processing; the local business records are stored using a unified record structure, which at least includes a record identifier, a business object identifier, a business number, an occurrence time, a source node identifier, a source node order, and associated content; wherein, the business object identifier and the business number together serve as the grouping basis, the occurrence time and the source node order together serve as the arrangement basis, and the associated content serves as the basis for determining whether adjacent local business records maintain a continuous succession relationship; by sequentially performing grouping, arrangement, and disconnection and reorganization, the local business records that are mixed together under the original upload order can be organized into a candidate record sequence with clear boundaries; The implementation process includes the following steps: First, each local business record to be processed in the central platform is read one by one. The business object identifier and business number are extracted from each local business record. Local business records with the same business object identifier and business number are grouped into the same record group. When grouping, the business object identifier and business number are combined to form a grouping key. After reading a local business record, other local business records with the same grouping key are retrieved from the local business records to be processed, and the retrieved local business records are written into the same record group. If no other local business records with the same grouping key are retrieved, the current local business record is formed into a separate record group. After the record group is formed, the record identifier, occurrence time, source node identifier, source node order, and associated content corresponding to each local business record are retained in the record group as input for subsequent sorting and comparison. Next, the local business records in each record group are sorted within the group. During sorting, they are first sorted in ascending order by occurrence time, placing the local business records with earlier occurrence times first. For local business records with the same occurrence time, they are then sorted sequentially according to the source node order, placing the local business records with earlier source node order first. The source node order uses a fixed order value written into the local business records. This order value is registered when the source node connects to the central platform and remains unchanged during the same processing. After sorting, each local business record is written to its group position from front to back along the sorting result to form the initial sequence of the corresponding record group. After this processing, each local business record in the same record group has a clear and unique sorting position. After the initial sequence is formed, an adjacency consistency comparison is performed on the local business records in each initial sequence. Specifically, the preceding and following local business records are read sequentially along the arrangement order of the initial sequence. Related content is extracted from the two local business records, and a consistency comparison is performed on the related content. In this embodiment, the related content is used to indicate whether adjacent local business records still belong to the continuous undertaking content of the same business matter, and its composition remains fixed in the same processing process. When the related content of two local business records is consistent, the two local business records are kept in the same arrangement sequence, and the following local business record is read and its adjacent local business records are compared in the next round. When the related content of two local business records is inconsistent, the inconsistent position is used as the break point. The preceding local business record is kept in the current arrangement sequence, and the following local business record is taken as the first record of the new sequence. The same consistency comparison is then performed on the local business records after the following local business record. After processing in this way, each initial sequence is organized into one or more candidate record sequences, and each candidate record sequence corresponds to a segment of local business records continuously undertaken under the same business matter. Through the above processing, the original partial business records are first grouped according to the business object identifier and business number, then arranged according to the occurrence time and source node order, and finally disconnected and reassembled according to the consistency of related content, thus obtaining a candidate record sequence with clear order, clear boundaries, and clear succession relationship. This can reduce the impact of the cross-mixing of partial business records of different business matters on the generation of subsequent record items, and provide a continuous input basis for the content merging, verification and migration processing in subsequent steps. In practical applications: For a single sales return transaction at the same store, the sales record uploaded by the sales terminal, the return record uploaded by the warehouse terminal, the invoicing record uploaded by the invoicing terminal, and the refund record uploaded by the settlement terminal may not arrive at the central platform in the actual order of the transactions. When processing according to this implementation method, the aforementioned partial business records are first grouped into the same record group based on the business object identifier and business number corresponding to the transaction. Then, an initial sequence is formed according to the occurrence time and source node order. Subsequently, a consistency comparison is performed on the associated content of adjacent partial business records. If the associated content of the sales record, invoicing record, and refund record is consistent, they remain in the same candidate record sequence. If there is another partial business record in the group with the same business object identifier and business number, but whose associated content has been transferred to the next business transaction, it is disconnected at the point where the associated content is inconsistent, allowing that partial business record to enter a new candidate record sequence, thereby ensuring that each candidate record sequence corresponds to a single and continuous business transaction process.

[0019] S2. For local business records in the candidate record sequence, the business content, confirmation content and attribution content are merged according to the sorting order. Duplicate content with the same content is merged and retained. Conflicting content with different content is recorded in parallel to generate an event record. In this embodiment, the purpose of processing S2 is to organize the partial business records in the candidate record sequence into item records that can enter the subsequent verification process. Although the candidate record sequence is limited to a continuous segment of partial business records under the same business matter, the content in each partial business record may still correspond to different aspects of the business matter. Therefore, it is necessary to write the content in the candidate record sequence into the same item record according to a fixed standard. In this embodiment, the content extracted from the partial business records is divided into three categories: business content, confirmation content, and attribution content. Among them, business content is used to reflect the business matter itself, confirmation content is used to reflect the financial confirmation status corresponding to the business matter, and attribution content is used to reflect the dimension allocation destination corresponding to the business matter. The same content is only classified into one of the three categories and is not written repeatedly among the three categories. The item records are stored using a unified writing structure. The content value, source record identifier, and source sorting position are written simultaneously under each corresponding content item to facilitate subsequent full verification, attribution verification, and migration identification. The implementation process includes the following steps: First, read each local business record sequentially according to the sorting order of the candidate record sequence, and extract the business content, confirmation content, and attribution content from each local business record. During extraction, based on the content classification rules already written in the local business record, content reflecting the business transaction itself is written into the business content, content reflecting the financial confirmation status is written into the confirmation content, and content reflecting the dimension's destination is written into the attribution content. After reading the first local business record, write the first occurrence of the business content, confirmation content, and attribution content into the corresponding content item in the event record, and simultaneously write the record identifier and source sorting position of the first local business record under each corresponding content item. When continuing to read subsequent local business records, if a certain business content, confirmation content, or attribution content has not yet been written into the event record, write that content into the corresponding content item, and simultaneously write the record identifier and source sorting position of the current local business record. After this processing, the first round of writing results is formed in the event record, resulting in the initial event record. Here, the first occurrence is based on the sorting order in the candidate record sequence, that is, the local business record read first in the sorting order is given priority as the first source record of that content. Subsequently, the local business records following the initial item record are read in sorted order, and the business content, confirmation content, and attribution content in each local business record are compared item by item with the corresponding content of the same type in the initial item record. During item-by-item comparison, only within the same type of content are compared, that is, business content is only compared with the corresponding item of business content, confirmation content is only compared with the corresponding item of confirmation content, and attribution content is only compared with the corresponding item of attribution content; cross-category comparisons are not performed. If a corresponding content item in the initial item record is empty, the corresponding content of the same type in the current local business record is added to the corresponding content item, and the record identifier and source sorting position of the current local business record are written simultaneously. If the corresponding content item in the initial item record already has content, the original writing result is maintained, and the original content is not overwritten by the current local business record. After the addition is completed in this way, the previously missing corresponding content items in the item record are filled, resulting in the added item record. Here, the determination of emptiness is based on whether the corresponding content item in the item record is empty; corresponding content items with existing content are no longer considered as addition objects. After the supplementary item record is formed, the subsequent local business records are checked in sorting order, and the business content, confirmation content, and attribution content in each local business record are compared item by item with the corresponding content in the supplementary item record. For a certain item in the current local business record, if the item is the same as the content value already written in the corresponding content item in the supplementary item record, no new content value is written. Instead, the record identifier and source sorting position of the current local business record are appended under the corresponding content item to indicate that the item appears repeatedly in multiple local business records. If the same content value already corresponds to multiple source records, the source record identifier and source sorting position are appended in sorting order. After this process, local business records with the same content value but different source records are merged into the same corresponding content item to obtain a merged item record. After this process, the item record retains both the corresponding content value itself and the order and source of the content value in the candidate record sequence. Continue performing a peer-to-peer comparison on the local business records following the merged event record. For a certain business content, confirmation content, or attribution content in the current local business record, if the content is different from all the content values ​​already written in the corresponding content item of the same type in the merged event record, then the content is written as parallel content in the parallel position under the corresponding content item, and the record identifier and source sorting position of the current local business record are written simultaneously. If the same content value as the parallel content appears again later, no new parallel position is added, but the source record identifier and source sorting position are continued to be written under the parallel position. The parallel positions are arranged sequentially in the order of their first appearance within the same corresponding content item to maintain the reading basis when the subsequent attribution branch is formed. Through this process, different content values ​​are stored in parallel under the same corresponding content item, and finally an event record is formed. At this time, each corresponding content item in the event record is clearly distinguished as a single-value content or parallel content, and each content corresponds to a clear source record identifier and source sorting position. Through the above processing, the local business records in the candidate record sequence are organized into a unified item record. The item record not only retains the business content, confirmation content, and attribution content, but also retains the initial source, supplementary source, duplicate source, and parallel source of each item. This allows subsequent steps to directly perform content completeness verification, attribution uniqueness verification, and migration identification based on the item record. This processing can, on the one hand, prevent the content in different local business records from losing its source basis in subsequent processing, and on the other hand, it can separate the duplicate content of the same type from the conflicting content of the same type, so that the subsequent judgment process has a clear reading target. In practical applications: For a sequence of candidate records for the same procurement item, the procurement application record may first include the procurement category and the applying department, corresponding to the business content and attribution content; the procurement approval record then includes the approval status, corresponding to the confirmation content; if the warehousing record again includes the same procurement category as the procurement application record, the record identifier and source sorting position of the warehousing record are added under the corresponding business content item; if the subsequent settlement record simultaneously assigns the item to two cost destinations, these two different attribution contents are written in parallel positions under the same attribution content item according to their first appearance order, and the corresponding source record identifier and source sorting position are written respectively; the item record formed in this way can indicate what content the procurement item already has, as well as which local business record each content comes from and in which sorting position it appears, thus providing a direct basis for the generation of subsequent judgment records.

[0020] S3. Based on the required content items corresponding to the current analysis type, perform a completeness check and a unique attribution check on the item record, and write the business content, confirmed content, and attribution content that have passed the check, as well as the missing content that has failed the check, into the judgment record. In this embodiment, S3 is used to determine whether a record has met the conditions for subsequent migration identification and multidimensional dynamic analysis, based on the fact that the record has already been formed. Specifically, the necessary content items required for the current analysis are first determined according to the current analysis type. Then, the business content, confirmation content, and attribution content in the record are checked item by item to distinguish between existing and missing content. For record items without missing items, attribution branches are further formed around the attribution content, and acceptance checks, confirmation checks, and business checks are performed respectively to identify whether each attribution branch can stably represent the single destination of the same record. When multiple attribution branches are formed, cross-judgment is performed in a fixed order to output the passed branch or the re-checked branch. Finally, the passed result, missing result, and re-check result are uniformly written into the judgment record. The current analysis type is written by the analysis task, and the necessary content items are obtained from the content item mapping table corresponding to the current analysis type. Only one current analysis type corresponds to the same processing process, and the order of content items in the content item mapping table remains fixed to ensure that the item-by-item check process and the subsequent writing process have a consistent reading basis. The implementation process includes the following steps: First, the current analysis type written in the analysis task is read, and the corresponding required content items are extracted from the content item mapping table according to the current analysis type. The content item mapping table pre-stores the correspondence between the current analysis type and the required content items. When reading, the current analysis type is used as the search key, and all required content items are obtained according to the sorting order in the mapping table. Then, the business content, confirmation content, attribution content, and the sorting position of each content in the event record are read, and the corresponding content is checked item by item in the order of required content items. When the corresponding content is found, the corresponding content and the corresponding sorting position are written into the checked item. When the corresponding content is not found, the required content item is written into the missing item. The corresponding content here is based on the consistency between the content item category and the required content item category in the event record. Business content only corresponds to the required content items of the business category, confirmation content only corresponds to the required content items of the confirmation category, and attribution content only corresponds to the required content items of the attribution category. Cross-category correspondence is not performed. After all required content items are checked, an initial check record is generated. The initial check record should at least include the event record identifier, checked items, missing items, and the sorting position of the checked items, which will be used as the reading basis for subsequent branch construction and result writing. Subsequently, only for item records where missing items were not written in the initial review record, the construction and verification of attribution chains are performed. During construction, all attribution content and the corresponding sorting position of each attribution content in the item record are first read and sorted in ascending order of sorting position. For multiple parallel attribution contents at the same sorting position, each parallel attribution content is used as a candidate attribution content under that sorting position. Then, starting from the first sorting position, one candidate attribution content is selected for each sorting position and connected sequentially to form an attribution chain. Different selection methods at different sorting positions form different attribution chains. After the attribution chains are formed, acceptance verification, confirmation verification, and business verification are performed on each attribution chain. During acceptance verification, the interval enclosed by two adjacent sorting positions in the attribution chain is used as the inspection scope to check whether other attribution contents outside the attribution chain are written within this interval. If no other attribution contents are written, the interval is checked against the two adjacent sorting positions. Between sorting positions, write a successor item; for other subordinate content, write a breakpoint item between the two sorting positions. During verification, read each subordinate content in the subordinate chain one by one and check whether the corresponding sorting position of the subordinate content contains the confirmation content of the same item record. If confirmation content is written, write a confirmation item in the subordinate content position; if no confirmation content is written, write a mismatch item in the subordinate content position. During business verification, read each subordinate content in the subordinate chain one by one and check whether the corresponding sorting position of the subordinate content contains the business content of the same item record. If business content is written, write a consistency item in the subordinate content position; if no business content is written, write a conflict item in the subordinate content position. After all subordinate chains are verified, a branch verification record is generated. The branch verification record stores the successor item, breakpoint item, confirmation item, mismatch item, consistency item, and conflict item corresponding to each subordinate chain. Next, cross-validation is performed on all attribution chains corresponding to the same item record. First, the total number of breakpoint items, mismatched items, and conflicting items in each attribution chain is counted, and the attribution chain with the highest total number is retained. If only one attribution chain is retained, it is written into the passing chain, and the cross-validation for that item record ends. If more than one attribution chain is retained, the number of receiving items in each retained attribution chain is counted, and the attribution chain with the highest number of receiving items is retained. If only one attribution chain is retained, it is written into the passing chain, and the cross-validation for that item record ends. If more than one attribution chain is retained... If there is only one, then compare the starting positions of the remaining belonging chains and retain the belonging chain with the highest starting position. If only one belonging chain is retained, then write that belonging chain into the passing chain. If there are still more than one retained belonging chain, then write all the retained belonging chains into the return chain. Here, the condition for entering the next round of comparison is that there are more than one retained belonging chain after the previous round of comparison. If the previous round of comparison has already compressed it into one, it will not enter the next round of comparison. After all cross-judgments are completed, an intermediate record is generated. The intermediate record contains the item record identifier, the passing chain or the return chain, and the corresponding sorting position range. Finally, a judgment record is generated based on the intermediate records. For the item records with passed branches written in the intermediate records, the business content, confirmation content, and attribution content corresponding to the passed branches are extracted from the item records and written into the passed content in the original sorting order. For the item records with missing items written in the initial review records, the corresponding required content items are directly written into the missing content. For the item records with re-examination branches written in the intermediate records, the attribution content and corresponding sorting position corresponding to the re-examination branches are written into the re-examination content in the order within the branches. After writing is completed, a judgment record is generated. The judgment record must at least include the item record identifier, passed content, missing content, re-examination content, and corresponding sorting position. Among them, the passed content serves as the direct input for subsequent migration identification, while the missing content and re-examination content are retained in the judgment record for subsequent supplementation, review, or re-execution of the judgment processing corresponding to the current analysis type. Through the above processing, the event record first completes the necessary content item verification according to the current analysis type, then forms a belonging chain around the belonging content and performs three types of verification, and then completes multiple rounds of cross-judgment in a fixed order, finally forming a judgment record with a clear structure. After this processing, it is possible to distinguish between event records with incomplete content, event records with unstable belonging, and event records with stable belonging, and to make subsequent steps read only the content that passes, thereby reducing the interference caused by unclosed event records directly entering the migration recognition. In practical applications: For a transaction record generated from the same contract payment, if the current analysis type is payment attribution analysis, the required content items written in the content item mapping table include the contract-related content, payment confirmation content, and attribution content. During the initial review, if the transaction record has already included the contract-related content, payment confirmation content, and multiple parallel attribution contents, then the transaction record will not include missing items and will proceed to attribution branch construction. Subsequently, multiple attribution branches are formed based on the values ​​of each parallel attribution content in different sorting positions, and each branch is checked for the insertion of other attribution contents and the presence of other attribution contents within each branch. Whether the corresponding position of the content is written with the payment confirmation content, and whether the corresponding position of each attributed content is written with the contract business content; if the total number of breakpoint items, mismatch items, and conflict items of a certain attributed branch is first, and the number of acceptance items is first, then the attributed branch is written into the approved branch, and the business content, confirmation content, and attributed content corresponding to the attributed branch are written into the approved content; if multiple attributed branches are still retained after the above comparison, then these attributed branches are written into the back-check branch, and the corresponding attributed content and sorting position are retained in the judgment record, so that the judgment process can be performed again after the subsequent supplementary records arrive.

[0021] S4. For the business content, confirmed content and belonging content that have been verified in the judgment record, perform adjacent comparison according to the sorting order in the candidate record sequence, and write the corresponding content item, the preceding content, the following content and the sorting position for the positions where the content is inconsistent between the two local business records, and generate migration record. In this embodiment, S4 is used to identify the content change process of the same item record in the candidate record sequence based on the formation of the judgment record, and to organize the scattered change items into migration records that can be directly entered into subsequent analysis and processing. In specific processing, the business content, confirmation content and attribution content that have been verified are first extracted from the judgment record, and according to the source sorting position of these contents in the item record, each content is mapped back to each local business record in the candidate record sequence to form a comparison sequence that can be compared in order. Then, the same item comparison between adjacent local business records is performed sequentially along the comparison sequence to identify the changes in the business content, confirmation content and attribution content in the preceding and following positions. Then, a migration segment is formed according to the succession relationship between the preceding and following change items. Finally, according to whether the sorting position within the migration segment is continuous, the migration segment is organized into migration content and written into the migration record. The migration record shall at least include the corresponding content item, the starting sorting position, the ending sorting position, the preceding content, the following content and the segment type, as a direct basis for subsequent analysis and segment merging. The implementation process includes the following steps: First, the approved content in the judgment record is read, and the business content, confirmation content, and attribution content are extracted from it. Simultaneously, the source sorting position of each content item is read from the corresponding item record. Then, the candidate record sequence corresponding to the judgment record is read, and the business content, confirmation content, and attribution content are mapped back to each local business record in the candidate record sequence according to their source sorting positions. During mapping, the source sorting position must match the sorting position in the local business record as the mapping condition. Once a content item is mapped to a specific sorting position, it is written to the local business record corresponding to that sorting position. If multiple contents correspond to the same sorting position, they are written to the business content, confirmation content, and attribution content positions in that local business record respectively. After all mappings are completed, each local business record in the candidate record sequence is reassembled into a sequential record with approved content, generating a comparison sequence. After this processing, subsequent comparison actions no longer directly address the summary results in the judgment record but instead return to the sequential structure of the candidate record sequence for execution. Subsequently, the preceding and following local business records are read sequentially along the comparison sequence in sorting order. Same-item comparisons are performed between similar content; that is, business content is compared only with business content, confirmation content only with confirmation content, and belonging content only with belonging content. Cross-category comparisons are not performed. During comparison, if the content value of a corresponding content item is the same in the preceding and following local business records, that corresponding content item is not written into a change item. If the content values ​​of a corresponding content item are different, that corresponding content item, the content value in the preceding local business record, the content value in the following local business record, and the sorting position corresponding to the following local business record are written into a change item. The sorting position corresponding to the following local business record is used as the writing position because this sorting position corresponds to the changed content state. After all comparisons along the comparison sequence are completed, an initial change sequence is generated. Each change item in the initial change sequence corresponds to a specific content change, and each change item retains information about which corresponding content item the change occurred in, what content changed to what content, and which sorting position the change occurred in. Next, each variable in the initial change sequence is read sequentially according to its sorted position, and a succession check is performed within the same corresponding content item. During processing, the content following the preceding variable is read first, followed by the content preceding the following variable. If the content following the preceding variable matches the content preceding the following variable, the two variables are considered to be connected end-to-end in terms of content evolution, and are grouped into the same transition segment. If the content following the preceding variable does not match the content preceding the following variable, it is determined that there is no continuous succession relationship between the two variables, and the following variable is grouped into the same transition segment. One change item is assigned to another migration segment; for multiple changes items in the same migration segment, they are written into the migration segment in the order of their corresponding sorting positions; for another migration segment, the next change item whose previous inheritance is invalid is used as the first change item to start writing again; after all changes items have been processed, a migration segment sequence is generated; here, only changes items generated by the same corresponding content item are allowed to enter the same inheritance verification process, and changes items generated by different corresponding content items form their own migration segments, so as to avoid business content changes, confirmation content changes and ownership content changes being mixed in the same migration segment; Finally, the sorting positions of each migration segment in the migration segment sequence are continuously checked. During the check, the sorting positions corresponding to the first and last changes in the same migration segment are read, and then the sorting positions of adjacent changes in the same migration segment are checked one by one to see if they are connected end to end. If the sorting positions of adjacent changes in the same migration segment are connected end to end, the migration segment is organized into a continuous segment type of migration content. If the sorting positions of adjacent changes in the same migration segment are not connected end to end, the migration segment is organized into a discontinuous segment type of migration content. During the organization, the corresponding content item, starting sorting position, ending sorting position, the preceding content of the first change, the following content of the last change, and the segment type are uniformly written for each migration segment. After all migration segments are organized, a migration record is generated. Each migration content in the migration record corresponds to a complete content migration process and can be directly read by subsequent steps for analysis of segment merging and change statistics. Through the above processing, the content in the judgment record is remapped to the candidate record sequence, and based on the order of the candidate record sequence, content comparison, change identification, migration segment formation, and migration content organization are performed sequentially. This allows the content change process in the same record to be broken down into migration records with clear location, clear succession, and clear segment type. After this processing, subsequent steps no longer need to backtrack the correspondence between the judgment record and the candidate record sequence, but can directly read the starting sort position, ending sort position, and segment type in the migration record to perform analysis and fragment merging. In practical applications: For judgment records formed by the same contract settlement matter, if the business content written in the content corresponds to contract settlement, the confirmation content corresponds to pending confirmation and confirmed confirmation in sequence, and the attribution content corresponds to two different attribution destinations in sequence, then first, these content entries are mapped back to the local business records in the candidate record sequence according to the source sorting position; then, the business content, confirmation content, and attribution content in adjacent local business records are compared in turn. If the confirmation content at a certain sorting position changes from pending confirmation to confirmed confirmation, then a confirmation content change item is written at the next sorting position; if the subsequent attribution content changes from the previous attribution destination to the next attribution destination, then another attribution content change item is formed; if the subsequent content of the previous change item is consistent with the preceding content of the next change item, then the two are grouped into the same migration segment; if they are inconsistent, they are divided into different migration segments; finally, according to whether the sorting position within the migration segment is continuous, it is organized into continuous segment type or discontinuous segment type migration content and written into the migration record, thereby providing a direct migration basis for the generation of subsequent financial multidimensional dynamic analysis results.

[0022] S5. Based on the verified business content, confirmed content and attribution content in the judgment record, merge the local business records in the migration record that are consecutive in the sorting position and have the same attribution content into the same analysis segment, and then perform multidimensional aggregation and time series statistics on each analysis segment to generate financial multidimensional dynamic analysis results. In this embodiment, S5 is used to organize partial business records in the same event record into statistically usable analytical segments based on the established judgment record and migration record. Then, it performs attribution dimension merging and before-and-after change identification on the analytical segments, ultimately generating a multi-dimensional dynamic financial analysis result. Specifically, the process first extracts the verified business content, confirmation content, and attribution content from the judgment record. Then, combining the sorting position and migration content already written in the migration record, it determines which partial business records are consecutive in sorting and consistent in attribution, and groups these partial business records into the same analytical segment. Subsequently, the attribution content in each analytical segment is processed according to… The data is split into categories in a fixed order, and then analysis segments with the same category category are merged. The number of local business records, starting sort position, ending sort position, business content, and confirmation content are counted after merging. Finally, the changes in the business content, confirmation content, and category category combination are identified according to the order in the summary record sequence to form a multi-dimensional dynamic financial analysis result. In this embodiment, the category category combination is formed by splitting the category content according to a pre-fixed category category order. The category category order remains unchanged in the same processing process, and the entry point for writing the business content and confirmation content in the summary record is consistent with the merging result in the analysis segment. The implementation process includes the following steps: First, extract the verified business content, confirmation content, and attribution content from the judgment record, and read the sorting position and migration content from the migration record. Then, read each local business record one by one according to the sorting order of the candidate record sequence, checking whether the sorting position corresponding to the current local business record falls within the sorting position range defined by the migration record, and checking whether the attribution content corresponding to the current local business record is consistent with the attribution content of the previous local business record. If the sorting position of the current local business record is consecutive to the previous local business record and their attribution content is consistent, then the current local business record is continued to be written into the current analysis segment. If the sorting position of the current local business record is not consecutive to the previous local business record, or their attribution content is inconsistent, then the current local business record is used as the first record of the new analysis segment and writing starts again. When the analysis segment is formed, the sorting position, business content, confirmation content, and attribution content corresponding to each local business record in the analysis segment are written synchronously. After all local business records are processed, an analysis segment sequence is generated. After this processing, each analysis segment corresponds to a segment of local business records with consecutive sorting positions and consistent attribution content. Subsequently, each analysis segment in the analysis segment sequence is processed sequentially. For each analysis segment, its attribution content is first read, and then the attribution content is split into corresponding attribution item combinations according to a pre-fixed attribution item order. The attribution item order is pre-written into an attribution item definition table in this embodiment. During reading, the segments are split sequentially according to the arrangement order in the attribution item definition table, so that the same attribution content forms attribution item combinations with the same order in different analysis segments. After the attribution item combinations are formed, analysis segments with the same attribution item combinations are merged. During merging, the attribution item combination is used as the merge key, and analysis segments with the same merge key are written into the same summary record. After being written into the same summary record, the overall... Count the number of local business records corresponding to the summary record, and take the first sorting position among all the merged local business records as the starting sorting position and the last sorting position as the ending sorting position; at the same time, write the business content and confirmation content in the merged analysis segment into the summary record in sorted order; if the merged business content or confirmation content corresponds to only a single content value, write that content value directly; if the merged business content or confirmation content corresponds to multiple content values, write them in sorted order; after all analysis segments are merged, a summary record sequence is generated; after this processing, each summary record corresponds to a continuous statistical result under a fixed set of attribution items; Next, arrange the summary records according to their starting and ending sorting positions in the summary record sequence. During arrangement, first sort by starting position in ascending order; for records with the same starting position, then sort by ending position in ascending order. After arrangement, read adjacent summary records sequentially and count the changes and order of changes among similar content. Specifically, compare the business content in the previous summary record with the business content in the next summary record item by item, and write the inconsistent items to the business change position. Compare the confirmation content in the previous summary record with the confirmation content in the next summary record item by item, and write the inconsistent items to the confirmation change position. Then, compare the previous summary record with the confirmation content in the next summary record item by item. The attribution item combination in the previous summary record is compared item by item with the attribution item combination in the next summary record, and the inconsistent attribution item positions are written into the attribution change position. Then, the order of the previous and next summary records in the summary record sequence is written into the change order, so that the change order can indicate between which two adjacent summary records the change occurred. The business change position, confirmation change position, attribution change position, and corresponding change order are uniformly written into the analysis result. After all adjacent summary records are compared, the financial multidimensional dynamic analysis result is generated. After this processing, the financial multidimensional dynamic analysis result retains both the merging results under each attribution item combination and the position and order in which the changes occurred. Through the above processing, the content of the records and the migration content in the migration records are further organized into an analysis fragment sequence, a summary record sequence, and the final financial multidimensional dynamic analysis results. This allows the merging results of the same event records in terms of the attribution dimension and the change results in terms of time sequence to be preserved simultaneously. After this processing, on the one hand, multidimensional merging of local business records can be performed according to the combination of attribution items. On the other hand, the change position and order of business content, confirmation content, and attribution item combinations can be identified according to the chronological order, so that the output results have both a merging basis and a change basis. In practical applications: For judgment and migration records of the same project expense items, if several partial business records are consecutive in sorting position and their attribution content corresponds to the same project, the same cost center, and the same period, these partial business records are grouped into the same analysis segment. Then, the attribution content in this analysis segment is split into attribution item combinations according to a fixed order of project, cost center, and period. Analysis segments with the same attribution item combinations are merged into the same summary record. Simultaneously, the number of partial business records, starting sorting position, ending sorting position, business content, and confirmation content corresponding to this summary record are counted. If the confirmation content in a later summary record changes relative to the previous summary record, or the cost center position in the attribution item combination changes, the corresponding confirmation change position or attribution change position is written into the analysis results, along with the order in which the change occurred between the two summary records. This allows the financial multidimensional dynamic analysis results to directly indicate what changes occurred in the continuous processing of the same project expense item and when these changes occurred.

[0023] Furthermore, a big data-based multidimensional dynamic financial analysis system includes: The record grouping module is used to acquire local business records that are collected to the central platform via edge data transmission, group them according to business object identifier and business number, and sort them within each group according to the occurrence time and source node order to generate candidate record sequences. The event generation module is used to merge the business content, confirmation content and attribution content of local business records in the candidate record sequence according to the sorting order, merge and retain duplicate content with the same content, and record conflicting content with different content in parallel to generate event records. The condition judgment module performs a completeness check and a unique attribution check on the item records based on the required content items corresponding to the current analysis type. The business content, confirmed content, and attribution content that pass the check, as well as the missing content that fails the check, are all written into the judgment record. The migration identification module is used to perform adjacent comparisons on the business content, confirmed content and belonging content that have passed the verification in the judgment record according to the sorting order in the candidate record sequence. For the positions where the content is inconsistent between two local business records, the corresponding content item, the preceding content, the following content and the sorting position are written to generate a migration record. The analysis output module merges local business records with consecutive sort positions and consistent attribution in the migration records into the same analysis segment based on the verified business content, confirmation content, and attribution content in the judgment record. Then, it performs multidimensional aggregation and time-series statistics on each analysis segment to generate financial multidimensional dynamic analysis results.

[0024] Working Principle: This solution first aggregates local business records scattered across different business nodes to the central platform via edge data transmission. Then, records belonging to the same business matter are grouped together according to business object identifier and business number. Combining the occurrence time, source node order, and related content, a truly continuous segment of records is organized into a candidate record sequence. Based on this, business content, confirmation content, and attribution content are extracted from the candidate record sequence to generate event records. Then, the completeness of the event records and whether their attribution can stably fall into the same direction are checked according to the required content items corresponding to the current analysis type, forming a judgment record. For content that passes the judgment, the changes in content before and after are identified along the candidate record sequence to generate migration records. Finally, local business records that are consecutively ordered and have consistent attribution are merged into analysis segments. Merging and statistical analysis of changes before and after are performed according to the attribution items, outputting multidimensional dynamic financial analysis results. The entire process is that each step provides the input basis for the next: first, the mixed records are sorted out; then, the scattered content is combined into events; then, the usable events are screened out; then, the change process is extracted; finally, results that can be directly used for financial analysis are formed. For example, in a chain retail scenario, a single store return transaction may generate multiple local business records at the POS terminal, warehouse terminal, invoicing terminal, and settlement terminal, and be transmitted to the central platform in batches via edge data transmission. These records may not arrive in the same order as the actual transaction and may also be mixed with records of other transactions. This solution first groups records with the same business object identifier and business number, then arranges them according to the time of occurrence and source node, and uses related content to separate records that do not belong to the same continuous process, resulting in a candidate record sequence corresponding only to this one return transaction. Subsequently, the sales content, invoice status, refund status, and destination are written into the transaction record. It checks whether this return transaction already has the content required for the current analysis, then identifies and confirms whether the status has changed and whether the destination has shifted. Finally, continuous and consistent records are organized into an analysis fragment, statistically analyzing the return transaction's inclusion in the store, project, cost center, or period, as well as the changes before and after. In this way, the system outputs no longer scattered raw records, but rather a multi-dimensional dynamic financial analysis result that directly reflects the attribution and changes of this transaction in actual operations.

[0025] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A financial multidimensional dynamic analysis method based on big data, characterized in that: include: S1. Obtain local business records that are collected from edge data transmission to the central platform, group them according to business object identifier and business number, and sort them within each group according to the occurrence time and source node order to generate a candidate record sequence; S2. For local business records in the candidate record sequence, the business content, confirmation content and attribution content are merged according to the sorting order. Duplicate content with the same content is merged and retained. Conflicting content with different content is recorded in parallel to generate an event record. S3. Based on the required content items corresponding to the current analysis type, perform a completeness check and a unique attribution check on the item record, and write the business content, confirmed content, and attribution content that have passed the check, as well as the missing content that has failed the check, into the judgment record. S4. For the business content, confirmed content and belonging content that have been verified in the judgment record, perform adjacent comparison according to the sorting order in the candidate record sequence, and write the corresponding content item, the preceding content, the following content and the sorting position for the positions where the content is inconsistent between the two local business records, and generate migration record. S5. Based on the verified business content, confirmed content, and attribution content in the judgment record, merge the local business records in the migration record that are consecutive in the sorting position and have the same attribution content into the same analysis segment, and then perform multidimensional aggregation and time series statistics on each analysis segment to generate financial multidimensional dynamic analysis results.

2. The financial multidimensional dynamic analysis method based on big data according to claim 1, characterized in that: S1 includes: S1-1. Extract the business object identifier and business number from the local business records, and group the local business records with the same business object identifier and business number to generate a record group. S1-2. Sort the local business records in each record group in ascending order according to the time of occurrence, and sort the local business records with the same time of occurrence in order according to the source node order to generate the initial sequence; S1-3. Read the previous local business record and the next local business record in the order of arrangement for each initial sequence. Perform consistency comparison on the related content of the two local business records. Keep the local business records with consistent related content in the same arrangement sequence. Break the local business records with inconsistent related content from the inconsistent position and form a new arrangement sequence to generate a candidate record sequence.

3. The financial multidimensional dynamic analysis method based on big data according to claim 2, characterized in that: S2 includes: S2-1. Read the local business records in the candidate record sequence according to the sorting order, extract the business content corresponding to the business occurrence, the confirmation content corresponding to the financial confirmation, and the dimension from each local business record and classify them into the corresponding belonging content. Write the local business record corresponding to the first appearance of each content into the event record to generate the initial event record. S2-2. Continue reading the local business records after the initial event record, compare the business content, confirmation content and attribution content in the local business records with the corresponding content in the initial event record item by item, and add the corresponding content that has not yet been written in the initial event record to the event record to generate the supplemented event record.

4. The financial multidimensional dynamic analysis method based on big data according to claim 3, characterized in that: S2 further includes: S2-3. Sequentially check the local business records after the supplementary item record, compare the business content, confirmation content and attribution content in the local business records with the corresponding content in the supplementary item record item by item, merge the local business records with the corresponding content and generate a merged item record. S2-4. Continue to check the local business records after the merged item record. Compare the business content, confirmation content and attribution content in the local business records with the corresponding content in the merged item record item by item. Record the local business records with different content in parallel into the corresponding content to generate the item record.

5. The financial multidimensional dynamic analysis method based on big data according to claim 4, characterized in that: S3 includes: S3-1. Read the required content items corresponding to the current analysis type, extract the business content, confirmation content, attribution content and corresponding sorting position from the event record, check whether the corresponding content exists in the event record item by item according to the required content items, if the corresponding content exists, write the corresponding content and corresponding sorting position into the checked item, if the corresponding content does not exist, write the corresponding required content item into the missing item, and generate the initial check record. S3-2. For item records where missing items are not written in the initial verification record, construct affiliation branches according to the affiliation content and corresponding sorting position, and perform acceptance verification, confirmation verification, and business verification on each affiliation branch. Among them, acceptance verification is used to check whether other affiliation content is written between adjacent sorting positions in the same affiliation branch. If no other affiliation content is written, an acceptance item is written in the corresponding position. If other affiliation content is written, a breakpoint item is written in the corresponding position. Confirmation verification is used to check whether each affiliation content corresponds to the confirmation content in the same item record at the corresponding sorting position. If it corresponds to the confirmation content, a confirmation item is written. If it does not correspond to the confirmation content, a mismatch item is written. Business verification is used to check whether each affiliation content corresponds to the business content in the same item record at the corresponding sorting position. If it corresponds to the business content, a consistency item is written. If it does not correspond to the business content, a conflict item is written. Generate a branch verification record.

6. The financial multidimensional dynamic analysis method based on big data according to claim 5, characterized in that: S3 further includes: S3-3. Perform cross-determination on each attribution branch corresponding to the same item record. First, count the total number of breakpoint items, mismatch items, and conflict items in each attribution branch, and retain the attribution branch with the highest total number. If there is more than one remaining attribution branch, continue to count the number of receiving items in the remaining attribution branches, and retain the attribution branch with the highest number of receiving items. If there is still more than one remaining attribution branch, compare the starting sort positions of the remaining attribution branches and retain the attribution branch with the highest starting sort position. If there is only one remaining attribution branch, write that attribution branch into the passing branch. If there is more than one remaining attribution branch, write all remaining attribution branches into the return check branch and generate intermediate records. S3-4. Based on the intermediate records, write the business content, confirmation content, and attribution content corresponding to the item records written to the passing branch into the passing content, write the required content items corresponding to the item records with missing items into the missing content, and write the attribution content and corresponding sorting position corresponding to the item records written to the back check branch into the back check content, and generate a judgment record.

7. The financial multidimensional dynamic analysis method based on big data according to claim 6, characterized in that: S4 includes: S4-1. Read the verified business content, confirmed content, and attribution content from the judgment record, and match each item to each local business record according to the sorting order in the candidate record sequence to generate a comparison sequence; S4-2. Read the previous local business record and the next local business record in the sorting order along the comparison sequence. Perform same-item comparison on the business content, confirmation content and belonging content in the previous local business record and the next local business record. Write the corresponding content items, the previous content, the next content and the sorting position of the next local business record that are inconsistent into the change item to generate the initial change sequence.

8. The financial multidimensional dynamic analysis method based on big data according to claim 7, characterized in that: S4 further includes: S4-3. Read each change item in the initial change sequence in order of sorting position. Perform a follow-up check on the content after the previous change item and the content before the next change item. If the follow-up check is consistent, the previous change item and the next change item are assigned to the same migration segment. If the follow-up check is inconsistent, the next change item is assigned to another migration segment, and a migration segment sequence is generated. S4-4. Perform a continuous sorting position check on each migration segment in the migration segment sequence, and organize the migration segments with continuous sorting positions and the migration segments with non-contiguous sorting positions into migration content to generate migration records.

9. The financial multidimensional dynamic analysis method based on big data according to claim 8, characterized in that: S5 includes: S5-1. Extract the verified business content, confirmed content, and attribution content from the judgment record, and combine them with the sorting position and migration content in the migration record. Group the local business records with consecutive sorting positions and consistent attribution content into the same analysis segment to generate an analysis segment sequence. S5-2. Sequentially process each analysis segment in the analysis segment sequence, split the content of each analysis segment into corresponding category combinations, merge according to the same category combination, and count the number of local business records, starting sort position, ending sort position, business content and confirmation content corresponding to each category combination, and generate a summary record sequence. S5-3. Then, arrange each summary record according to the starting and ending sorting positions in the summary record sequence. For two adjacent summary records, calculate the changes in the business content, confirmation content, and attribution item combinations, and generate financial multidimensional dynamic analysis results.

10. A financial multidimensional dynamic analysis system based on big data, characterized in that: include: The record grouping module is used to acquire local business records that are collected to the central platform via edge data transmission, group them according to business object identifier and business number, and sort them within each group according to the occurrence time and source node order to generate candidate record sequences. The event generation module is used to merge the business content, confirmation content and attribution content of local business records in the candidate record sequence according to the sorting order, merge and retain duplicate content with the same content, and record conflicting content with different content in parallel to generate event records. The condition judgment module performs a completeness check and a unique attribution check on the item records based on the required content items corresponding to the current analysis type. The business content, confirmed content, and attribution content that pass the check, as well as the missing content that fails the check, are all written into the judgment record. The migration identification module is used to perform adjacent comparisons on the business content, confirmed content and belonging content that have passed the verification in the judgment record according to the sorting order in the candidate record sequence. For the positions where the content is inconsistent between two local business records, the corresponding content item, the preceding content, the following content and the sorting position are written to generate a migration record. The analysis output module merges local business records with consecutive sort positions and consistent attribution in the migration records into the same analysis segment based on the verified business content, confirmation content, and attribution content in the judgment record. Then, it performs multidimensional aggregation and time-series statistics on each analysis segment to generate financial multidimensional dynamic analysis results.

Images