Financial data identification and accounting methods based on artificial intelligence

Abstract

The present invention relates to the technical field of combining financial accounting with artificial intelligence, and in particular to a method for identifying and calculating financial data based on artificial intelligence. The steps of the present invention include: first calculating fractal dimensions using a multi-level grid splitting technique to construct a fractal financial semantic map covering transactions and interoperable entities; then utilizing quantum hybrid aggregator interference operations to fuse fractal map attributes with subject candidate results, obtain interference scores, and generate calibrated subject information through symbolic rule screening or manual review and confirmation. This subject information is then used to automatically generate accounting entries and archive ledgers, and the entire decision-making process is recorded using quantum hashing technology. By continuously analyzing and archiving decision data, the system can adaptively adjust fractal dimension calculations and interference parameters to achieve accurate identification and efficient accounting of complex financial data.

Description

Technical Field

[0001] The present invention relates to the technical field of combining financial accounting with artificial intelligence, and in particular to a financial data recognition and accounting method based on artificial intelligence. Background Art

[0002] In current financial management, artificial intelligence (AI) is being widely applied in data collection, intelligent journal entries, and anomaly detection to reduce the burden of manual operations and improve accounting accuracy. Existing technologies typically employ multi-level digital employees coupled with numerous rule-based scripts to handle diverse business scenarios (existing literature, Publication No. CN118691418A, Title: A Digital Employee Intelligent Financial Accounting Method and System). While this method achieves a certain degree of automation, it still faces bottlenecks in cross-system integration, complex rule maintenance, and unexpected situation handling. In particular, as business volume and rules continue to expand, traditional script-based accounting processes are difficult to dynamically optimize, resulting in deficiencies in compliance and timeliness. Summary of the Invention

[0003] To address the numerous issues with the aforementioned existing technologies, the present invention provides an AI-based method for identifying and calculating financial data. This method first constructs a financial semantic graph using multi-level grids and fractal dimensions, presenting the multi-scale attributes of transaction nodes. A quantum hybrid aggregator then performs interference operations on the fractal graph vectors and candidate account vectors to screen for trusted accounts. Finally, a combination of symbolic rule verification and manual review outputs the calibrated account information, and quantum hashing technology is used to archive and trace decision records. This method balances the adaptability of AI in complex financial scenarios with the robustness of traditional compliance accounting, significantly improving automation and accuracy.

[0004] A financial data identification and accounting method based on artificial intelligence, comprising the following steps:

[0005] Obtain financial source data and perform preliminary formatting and verification to obtain usable financial data;

[0006] Associating nodes and entities in the available financial data to establish an initial semantic graph, and calculating fractal dimensions based on node attribute vectors to generate fractal similarity edges or nested relationship edges, so that transactions and entity nodes form a fractal financial semantic graph;

[0007] The fractal financial semantic graph and the account candidate results are input into the quantum hybrid aggregator, and the constructive or destructive effects are evaluated using the quantum state superposition method to determine the interference score. After verification using symbolic rules, conflicting or abnormal transaction nodes are reviewed and corrected, and the updated account information is output to obtain the calibrated account data.

[0008] Accounting entries are generated based on the calibrated account data and ledger archiving is performed. Quantum hashing technology is used to record decision information of the transaction processing and correction processes. After analyzing the decision information, the fractal dimension calculation parameters and the weight factors of the quantum hybrid aggregator are adjusted.

[0009] Preferably, the process of obtaining financial source data and performing preliminary formatting and verification includes:

[0010] Use character detection algorithms to perform optical recognition on paper documents or scanned images to obtain text records;

[0011] Align the fields of the recognized text records with those in the structured business system, unify the currency unit and set the date format, and filter out records that do not meet the field integrity requirements;

[0012] Entries whose numerical ranges do not meet the preset thresholds are marked as abnormal data to be reviewed, and available financial data that meets the field completeness requirements is output.

[0013] Preferably, the step of associating nodes and entities in the available financial data includes:

[0014] Use transaction identifiers or unique indexes to map each record to the corresponding entity number to generate the basic nodes of the initial semantic graph;

[0015] When multiple transactions occur under the same entity number at different times, the amount, category, or current account information is recorded in the node attributes in chronological order;

[0016] Records that fail to match any entity number are marked as isolated nodes.

[0017] Preferably, the step of calculating the fractal dimension based on the node attribute vector includes:

[0018] Multi-level grid splitting is used to discretize the amount range, transaction frequency and number of related entities;

[0019] The fractal dimension of a node is calculated for discretized data. When the difference between the fractal dimension and the fractal dimension of another node is lower than a preset threshold, a fractal similarity edge is generated.

[0020] When node attributes are hierarchically recursively associated on the main feature components, nested relationship edges are established and the association is updated to the fractal financial semantic graph.

[0021] Preferably, after generating fractal similarity edges or nested relationship edges, the abnormal transaction nodes are preliminarily identified by comparing the transaction frequency and cumulative amount between the nodes, and the preliminary identification results are saved in the fractal financial semantic graph.

[0022] Preferably, the step of inputting the fractal financial semantic graph and the subject candidate results into the quantum hybrid aggregator comprises:

[0023] The node attribute vectors of the fractal financial semantic graph and the subject candidate result vectors are used to construct quantum state sequences respectively;

[0024] Based on the principle of quantum state superposition, constructive merging of similar vectors and destructive processing of conflicting vectors are performed to obtain the interference score.

[0025] Candidate subjects with interference scores higher than the threshold are taken as priority attribution suggestions to provide input data for subsequent symbol rule verification.

[0026] Preferably, when using symbol rule verification, candidate accounts with high interference scores are judged one by one according to transaction amount limits, compliance accounting account regulations and internal approval terms, and candidate accounts that do not meet the rules are eliminated and archived.

[0027] Preferably, when auditing and correcting transaction nodes with conflicts or anomalies, the transaction number, accounts before and after correction, operator and modification time are recorded, and this correction information is associated with the interference score and written into the calibrated account data to form updated account information.

[0028] Preferably, when generating accounting entries based on the calibrated account data and performing ledger archiving, a transaction identification tag, specific account information and related party information are configured for each entry.

[0029] Preferably, quantum hashing technology is used to write the decision information of the transaction processing process and the correction process, and after analyzing the decision information, the grid splitting scale used in the fractal dimension calculation and the interference parameters of the quantum hybrid aggregator are adjusted.

[0030] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0031] This invention uses fractal financial semantic graph technology to achieve adaptive aggregation of transaction nodes in multiple dimensions, effectively avoiding the rule explosion problem caused by over-reliance on complex scripts in the comparison solution.

[0032] This invention uses quantum hybrid aggregator interference computing technology to achieve deep fusion of subject candidate information and fractal attribute vectors, solving the problems of low efficiency and easy misjudgment of comparative solutions in multi-layer digital employee sorting.

[0033] The present invention uses quantum hashing to achieve tamper-proof technology to achieve secure archiving and self-learning analysis of the decision-making process, further solving the defect of the comparison scheme in lacking closed-loop optimization in rule iteration and anomaly tracing. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Schematic diagram of the process of the present invention;

[0035] Figure 2 Schematic diagram of the quantum hybrid aggregator and fractal financial semantic graph in the present invention. DETAILED DESCRIPTION

[0036] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0037] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0038] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0039] like Figure 1 As shown, a financial data identification and accounting method based on artificial intelligence includes the following steps:

[0040] Obtain financial source data and perform preliminary formatting and verification to obtain usable financial data;

[0041] Preferably, the process of obtaining financial source data and performing preliminary formatting and verification includes:

[0042] Use character detection algorithms to perform optical recognition on paper documents or scanned images to obtain text records;

[0043] Align the fields of the recognized text records with those in the structured business system, unify the currency unit and set the date format, and filter out records that do not meet the field integrity requirements;

[0044] Entries whose numerical ranges do not meet the preset thresholds are marked as abnormal data to be reviewed, and available financial data that meets the field completeness requirements is output.

[0045] This invention combines the character detection algorithm in artificial intelligence with the data normalization strategy to achieve preliminary integration and verification of financial data from multiple sources:

[0046] Text information is extracted from paper documents or scanned images through an optical recognition process and converted into a machine-readable structured record. The AI ​​model primarily utilizes principles of character location and classification to generate a text map for each image block, reducing errors caused by manual data entry. In one embodiment, the character detection algorithm processes the scanned image in blocks: first, the image is divided into rows and columns, and the characters to be classified are output block by block. The characters are then stored bundled with their confidence values, improving traceability for subsequent verification of financial data fields.

[0047] Since financial data may contain multiple currencies, for subsequent processing and accounting, all original amount fields need to be converted to a unified currency unit. In one embodiment, if the US dollar is used as the unified unit, the formula is used:

[0048]

[0049] in Indicates the converted amount. Indicates the original amount, Indicates the exchange rate from the original currency to the unified currency.

[0050] The processing of date formats is similar. Dates in different formats (such as "YYYY / MM / DD" or "DD-MM-YYYY") are re-split and re-labeled, and then stored in the same type of date data structure.

[0051] By setting field completeness rules, each record is checked to see if it meets the minimum required fields (for example, it must include the transaction date, amount, and transaction object), and records that do not meet the requirements are marked as "abnormal pending review"; in one embodiment, if the value in the amount field is lower than zero or higher than a certain upper limit, the entry is marked as an abnormal entry, providing an entry for subsequent manual verification or further screening by the algorithm.

[0052] Through the above process, the system can initially integrate financial data from multiple sources and formats into a standardized data structure, reducing type conflicts and data omissions in subsequent analysis stages.

[0053] In practical applications, the specific process of text record extraction includes:

[0054] Scanned image access: The input image is first segmented by line, and the possible locations of characters in each line are identified. A character detection algorithm is used to detect characters line by line, outputting detection box coordinates and predicted characters. The predicted characters are then sequentially merged into a text record. If a character detection fails to meet the confidence threshold, a "failed detection" label is added to the output text record to facilitate subsequent manual review. This process significantly reduces the repetitive workload and error rate of manual data entry.

[0055] In actual application, the specific process of currency conversion includes:

[0056] The system loads the current or current effective exchange rate from the exchange rate interface. , converting all original amounts to the same benchmark. If the currency code in a record doesn't match (e.g., an unknown currency), the record will be placed on the "Exception Pending Review" list. This unified currency approach facilitates subsequent comparison and association of amount attributes across different transaction nodes in the semantic graph, eliminating the need to separately process multiple currency types.

[0057] Set the date storage specification, such as "YYYY-MM-DD", and split and combine dates from different sources:

[0058] If the original date is in the format of "DD / MM / YYYY," the string position is sliced ​​and reassembled. If a date field contains non-numeric characters or does not fit within the legal date range, it is marked as an anomaly. This unified date storage method facilitates the understanding of the timing properties of transaction nodes during subsequent fractal dimension calculations, such as calculating the number of days between two transactions.

[0059] In actual applications, abnormal entry marking and output include:

[0060] By presetting thresholds, you can filter out entries with amounts outside the normal range or missing key information.

[0061] Flagged entries are temporarily removed from the available financial dataset or placed in a "pending review dataset" for manual confirmation or secondary review using other rule-based algorithms. Ultimately, only datasets with complete fields and meeting threshold requirements are retained, forming the usable financial data for subsequent "semantic graph construction and accounting processes."

[0062] The combination of an AI-based character detection algorithm and a unified rule-based verification strategy can automatically identify paper or unstructured data, eliminate abnormal entries or mark defective fields, and lay a unified data foundation for the subsequent analysis of fractal financial semantic maps and quantum hybrid aggregators. By means of exchange rate conversion and date format standardization, the confusion caused by "messy data source formats, missing fields, and inconsistent amount units" is reduced. The system as a whole reduces the workload of manual collection, proofreading, and cleaning of multi-channel financial data, and explicitly records the uncertain items in the processing link in the abnormal mark. Under the process of the present invention, there is no need to repeatedly deal with currency, date and other issues when constructing the fractal semantic map; the quantum hybrid aggregator can also directly read the standardized amount field for further interference score calculation. In actual deployment scenarios, this step can make the data integration process more orderly and transparent, and once a problem occurs, it can be traced back to the specific scan record or field conflict source, greatly improving the efficiency of auditing and error checking.

[0063] Associating nodes and entities in the available financial data to establish an initial semantic graph, and calculating fractal dimensions based on node attribute vectors to generate fractal similarity edges or nested relationship edges, so that transactions and entity nodes form a fractal financial semantic graph;

[0064] In available financial data, each transaction can be considered a "transaction node," while the parties, items, or categories of transactions can be considered "entity nodes." This paper connects transaction nodes and entity nodes in a graph structure to form an "initial semantic graph." Based on this, an attribute vector is defined for each node to describe its numerical characteristics (e.g., amount, transaction frequency, and temporal distribution).

[0065] When multiple nodes are mapped to a graph structure, in order to measure their distribution characteristics under multi-dimensional attributes, this paper introduces the concept of "fractal dimension". Suppose a node attribute vector is:

[0066]

[0067] in Indicates that the node is The fractal dimension is used to evaluate the discrete coverage or distribution level of a node at multiple scales. For example, in a multi-level grid partitioning scheme, if the attribute characteristics of a node show increasing coverage at each scale, it will show a high fractal dimension.

[0068] If two nodes (denoted as and ) and the distance between their main attribute vectors are both less than a threshold, a "fractal similarity edge" is added to the graph structure; this indicates that they have similar distribution characteristics at multiple scales. If a node has multiple hierarchical nested relationships with another node on the fractal dimension, a "nested relationship edge" is created, indicating recursive dependency or hierarchical subordination between the nodes, which can be seen in multi-level contracting or cross-layer project finance scenarios.

[0069] In actual applications, each transaction record is marked as , mark each transaction party or item as , and insert the corresponding node in the graph structure. If a transaction record contains a party identifier, then the connection and , forming a preliminary "transaction-entity" mapping. In one embodiment, if a project has multiple related transactions scattered across different months, they are merged into the project node to fully consider the transaction diversity of the project in the subsequent fractal dimension evaluation.

[0070] For transaction nodes , you can set The specific attribute types depend on the financial scenario. For example, in procurement transactions, attributes may include unit price, batch quantity, supplier rating, etc.; in scientific research project funding management, attributes may include supporting fund ratio, funding arrival batch, etc. You can also define corresponding attributes, such as the total amount of historical cooperation, the number of transactions involved, the level of related projects, etc.

[0071] Under the conventional grid subdivision idea, first the attribute vector The space is divided into multiple scales, and each scale is recorded as . In scale The number of grids covered by the statistical nodes If a box-counting fractal dimension estimation method is defined, the following core expression can be used:

[0072]

[0073] in is the node approximate fractal dimension, is the number of grids in which nodes are distributed at the corresponding scale, and This calculation only requires the necessary formulas and parameter definitions to be provided in the user manual.

[0074] Set node Has fractal dimension , attribute vector ,node Has fractal dimension , attribute vector .like: and In the present invention, the node and Add a "fractal similarity edge". and Representing the pre-set dimension difference threshold and attribute vector distance threshold, respectively. This operation can explicitly indicate in the graph that two nodes are converging on multi-scale and multi-dimensional features, which is very helpful in determining transaction ownership or verifying similar records in the subsequent quantum hybrid analysis stage.

[0075] If the node Presenting nodes under multi-scale division Repeated nested coverage of nodes At each level of the scale The sub-distribution interval of and Fractal hierarchical dependencies exist. This is common in complex financial networks, such as multi-level internal transfers or sub-fund disbursements. Building nested edges can help financial personnel identify cross-level and cross-project transaction chains.

[0076] Preferably, the step of associating nodes and entities in the available financial data includes:

[0077] Use transaction identifiers or unique indexes to map each record to the corresponding entity number to generate the basic nodes of the initial semantic graph;

[0078] When multiple transactions occur under the same entity number at different times, the amount, category, or current account information is recorded in the node attributes in chronological order;

[0079] Records that fail to match any entity number are marked as isolated nodes.

[0080] Each financial record usually contains a unique transaction identifier, which is recorded as , and the entity number corresponding to the transaction party or project party In order to represent the association between transactions and entities in a graph structure, it is necessary to map transaction identifiers to their matching entity numbers, thereby forming two types of nodes (transaction nodes and entity nodes) and their connecting edges in the initial semantic graph.

[0081] If the same entity number accumulates multiple transactions in different time periods, the transactions can be sorted in chronological order, and information such as the amount, subject or current account can be appended to the attribute record of the entity node, so that subsequent artificial intelligence algorithms can use the time dimension features for identification and accounting.

[0082] If an entity number in a transaction record cannot be found in the system, it is considered an "orphan node" and exists alone in the graph structure. This orphan node can alert the system or financial personnel to possible anomalies such as new suppliers, unregistered items, or missing records, and should be addressed in subsequent processes (such as quantum state interferometry analysis or manual review).

[0083] In actual application, the system first reads the (e.g. "TX20230912-001") and (For example, "SUP-A001" represents a supplier). In the initial map, for each different Create a transaction node, if it has never appeared in the graph , a new entity node is created; if it has already appeared, the existing entity node is used and a transaction-entity edge is added.

[0084] If an entity number Multiple transactions occur consecutively If the difference is only in the date, specific amount, or subject, then sort by date from earliest to latest:

[0085]

[0086] At the entity node In the attribute set, the amounts of these transactions are recorded in chronological order. , subject information , current account identifier , providing a serialized data source for subsequent fractal dimension calculation or anomaly detection.

[0087] When the entity number of the transaction record appears If a transaction fails to match the system's registered database or mapping rules, the system automatically marks the transaction node as an "isolated node." In the initial semantic graph, isolated nodes have no physical edges, which helps quickly identify "unknown transactions" or "requires manual re-entry" scenarios during subsequent quantum hybrid analysis or exception handling.

[0088] In one embodiment, a company has three main suppliers, numbered , and more than ten new transaction records Recorded on the same day.

[0089] The system reads the transaction ID and supplier number one by one, and ”) directly establishes the connection between transaction node and entity node, and appends the amount and items in the time sequence to In the properties of If the corresponding entity number cannot be found in the system, the transaction node will be marked as an isolated node.

[0090] Finally, on the initial semantic graph, some transaction nodes converged to three main supplier nodes to form a star or chain structure, and another part of the nodes (with only one unmatched entity) were classified into the "isolated" node queue for subsequent manual verification.

[0091] Centralized transaction mapping is achieved during the initial graphing phase, enabling the management of transactions across multiple time periods for the same entity through a single node view. Later, when conducting fractal dimension analysis or quantum state interference, the system can perform deep aggregation of connected transactions, focusing on the investigation and supplementation of isolated nodes, laying the foundation for visualization and traceability of financial data integrity.

[0092] Preferably, the step of calculating the fractal dimension based on the node attribute vector includes:

[0093] Multi-level grid splitting is used to discretize the amount range, transaction frequency and number of related entities;

[0094] The fractal dimension of a node is calculated for discretized data. When the difference between the fractal dimension and the fractal dimension of another node is lower than a preset threshold, a fractal similarity edge is generated.

[0095] When node attributes are hierarchically recursively associated on the main feature components, nested relationship edges are established and the association is updated to the fractal financial semantic graph.

[0096] In the present invention, each transaction node or entity node has several attributes, such as amount, transaction frequency, and number of associated entities. These attributes can be integrated into a vector ,in Indicates the attribute values ​​(such as the total transaction amount, the cumulative number of times within a specified period, the number of entities with which transactions occur, etc.).

[0097] Multi-level grid splitting means setting several levels of scale for each attribute dimension, dividing the node attributes into corresponding intervals at each level, and performing discretization statistics on the number or frequency of grids that the nodes fall into. For example, taking the amount field as an example, The value space of is divided into several intervals, and the number of transactions in each interval is recorded; if the transaction frequency If the range is large, the frequency value can be discretized by using logarithmic scale or other methods.

[0098] The fractal dimension is used in the present invention to quantify the complexity or self-similarity of node attributes at different scales. By counting the node coverage on a multi-level grid, the fractal eigenvalue of the node can be obtained. If the distribution of a node at different scales maintains a highly recursive or delicate structure, the fractal dimension tends to be higher; if the distribution is single or coarse-grained, the fractal dimension is lower. In the common "box counting" or "grid counting" methods in this field, if a node is The number of grid covers under , then the fractal dimension You can refer to similar

[0099]

[0100] in and The specific formulas and parameters are listed in detail above and do not need to be detailed here.

[0101] When the fractal dimension values ​​of nodes A and B differ slightly, and the distance between their main attribute vectors is also lower than a certain threshold, a fractal similarity edge is added to them in the fractal financial semantic graph, indicating that the two nodes are highly similar in multiple scales and dimensions.

[0102] If a node C recursively contains or hierarchically nests another node A at multiple grid scales, nested relationship edges are created so that their master-slave relationship or multi-level dependency relationship is explicitly reflected in the graph.

[0103] In financial scenarios, fractal similarity edges can help identify similar transaction groups (such as those with similar amount ranges and transaction frequencies), while nested relationships can indicate the deep structure of multi-level transfers or cross-project fund management.

[0104] In actual application, for the amount range, you can set A partition interval, such as , count the frequency or cumulative amount of transactions for each node within the interval. The processing of transaction frequency and number of associated entities is similar, and their value ranges can be segmented or split on a logarithmic scale. In one embodiment, if there is a node Amount attribute Large, but frequent If it is not high, the system will find the grid corresponding to "high amount and low number of times" under more scales. Exhibits special patterns in fractal dimension assessment.

[0105] After calculating the fractal dimension, the fractal dimension differences between nodes are compared. With node The dimension difference satisfies:

[0106]

[0107] in is a pre-set threshold, which can be considered and It has similarity under multi-level grid splitting. If its main attribute vector distance also meets another threshold , then a fractal similarity edge is established. The specific vector distance metric can be in the form of Euclidean distance or Manhattan distance.

[0108] In the specific implementation, the fractal dimension can be calculated for all nodes first. , and then perform pairwise matching when judging the distance, for those that meet:

[0109]

[0110] and

[0111]

[0112] Node pairs Adding a "fractal similarity edge" indicates that the two are consistent or closely related in multiple dimensions, which is conducive to quickly locking similar group transactions during subsequent quantum interference or symbol verification.

[0113] If the node For Node In the grid division, inclusive coverage is shown, that is, at each level of scale Down, The grid and The grid where it is located has recursive inclusion, which can be determined and There is a fractal level dependency.

[0114] So insert "nested relationship edges" into the graph structure ” and updates it to the fractal financial semantic graph to help identify potential hierarchical associations in cross-layer project finance or multi-stage funding chains.

[0115] In one embodiment, in a system, there are ≥10 nodes, some of which represent aggregated transaction clusters and some represent single items, and the attributes include three dimensions: "amount range", "transaction frequency" and "transaction entity count".

[0116] By multi-level grid splitting, we first count the frequency of coverage of each dimension of each node in different scale intervals, and then calculate the fractal dimension. With node If the fractal dimensions of the two are very close and their properties are similar, then add fractal similarity edges. If the node At most scales, the nodes is included in the higher-level distribution, then and Creates nested relationship edges.

[0117] Finally, a fractal financial semantic graph is generated, in which some nodes form similar groups, while other nodes present a multi-level structure through nested relationships, thereby showing more refined aggregation in subsequent quantum interference or automatic classification of subjects.

[0118] Through fractal dimension calculation and a multi-level grid splitting strategy, the system can distinguish the degree of aggregation and hierarchical relationships of nodes under multi-dimensional attributes, which is more flexible than single threshold filtering. Fractal similarity edges reveal strong connections between short-distance nodes, and nested relationship edges represent hierarchical dependencies, which is conducive to discovering potential connections across departments, multi-level projects, or recursive funding allocations in financial scenarios. In subsequent AI accounting and anomaly detection processes, the graph structure of this fractal financial semantic graph provides a visual and traceable foundational network for quantum state interference calculations or symbolic rule matching.

[0119] Preferably, after generating fractal similarity edges or nested relationship edges, the abnormal transaction nodes are preliminarily identified by comparing the transaction frequency and cumulative amount between the nodes, and the preliminary identification results are saved in the fractal financial semantic graph.

[0120] In the fractal financial semantic graph, if two nodes are connected by fractal similarity edges or nested relationship edges, it means that they are highly similar or hierarchically dependent on the multi-scale fractal dimension. The present invention further compares the transaction frequency and cumulative amount between nodes to identify possible abnormal transactions. With node The fractal similarity edge has been established, and the system can access its transaction count in the attribute vector and cumulative transaction amount If these values ​​conflict with the overall distribution or the node's own threshold, it can be determined that there are signs of anomalies.

[0121] When the system detects that a node's transaction frequency or cumulative amount is too high or too low compared to its associated nodes, it will mark the node with an "abnormal" or "needs review" label within the graph. This preliminary identification is not a final conclusion, but it is sufficient to draw attention to subsequent quantum state interference or symbol rule verification, thereby reducing missed inspections during financial audits.

[0122] To ensure the integrity and traceability of subsequent data flows, this invention associates anomaly identifiers with node or edge attributes and stores them in a fractal financial semantic graph. This allows subsequent analysis to quickly locate suspected anomaly nodes simply by reading the graph data, further improving accounting efficiency and accuracy.

[0123] In practical applications, the nodes With node The transaction records are summarized as follows:

[0124] The frequency of travel can be expressed as , where 1 is the indicator function, counting all the arrive The number of transaction entries.

[0125] The cumulative amount can be expressed as ,in Representative The amount of all transactions of transactions are totaled in this value.

[0126] If the node If some nodes on the fractal similar edge have extreme values ​​in frequency or amount distribution, it can be determined that There is a possibility of abnormal transactions in itself or in adjacent nodes.

[0127] The system can pre-set a set of thresholds for each edge or each node and , respectively used for alert detection of transaction frequency and cumulative amount. or In one embodiment, if the node Connected to multiple similar nodes and the cumulative amount of each connecting edge is higher than ,but The abnormal level of the event will be raised first, and a "high-risk" status will be marked in the figure to guide subsequent algorithm or human review.

[0128] If the node By nesting the relationship between edges and nodes To form a hierarchical relationship, we should compare and The frequency and cumulative amount of cross-level funds or transaction records. When the flow direction or number of funds under the nested relationship exceeds the expected range, or Mark it as abnormal, or judge that there is inexplicable large transaction between the two.

[0129] To maintain process integrity, this invention writes flags (e.g., "abnormal_flag = true") into node or edge attributes within the fractal financial semantic graph and records threshold trigger information. This writeback operation ensures that subsequent steps (quantum hybrid aggregator or symbolic rule verification) can directly read this information from the graph, eliminating the need to recalculate transaction frequency or cumulative amounts.

[0130] In one embodiment, within a group company, the node is ≥ 5 departments and each department has different fractal associations with supplier or project nodes. In the fractal semantic graph, it has similar edges with multiple supplier nodes.

[0131] System Statistics Frequency of transactions with relevant suppliers and cumulative amount If one of the edges is found Far above the average level, the system marks or This preliminary identification is recorded in the fractal financial semantic graph, and the potential risk can be immediately noted during the subsequent review stage, human-machine collaboration, or quantum interference assessment.

[0132] Comparing the actual transaction frequencies and dollar amounts derived from fractal similarity and nested relationship edges automatically identifies possible anomalies, focusing analysis of large amounts of distributed financial data on a small number of high-risk areas. By visualizing these initial indicators within the graph, auditors and accountants can quickly identify suspicious entities or transaction links, enabling more efficient follow-up investigations.

[0133] like Figure 2 As shown, the fractal financial semantic graph and the subject candidate results are input into the quantum hybrid aggregator, and the constructive or destructive effect is evaluated by the quantum state superposition method to determine the interference score. After the symbol rule verification, the transaction nodes with conflicts or anomalies are reviewed and corrected, and the updated subject information is output to obtain the calibrated subject data.

[0134] The present invention first internally matches the "fractal financial semantic graph" with the "account candidate results." In the fractal financial semantic graph, each node already possesses fractal attributes and multi-dimensional transaction records. The account candidate results are often derived from prior artificial intelligence or rule-based inference (e.g., large language models classifying text descriptions), providing a list of possible accounting categories for each transaction node. These two types of information are then fed into a "quantum hybrid aggregator," where the system evaluates their consistency or conflict using the principle of quantum superposition.

[0135] In conventional numerical fusion, most methods are based solely on weighting or simple voting. This paper introduces the concept of quantum state superposition: the node attribute vectors in the fractal map and the subject candidate result vectors are treated as two sets of "quantum states," which are superimposed in the same Hilbert space. When the two configurations are similar or facing each other, a "constructive" effect can occur, improving the applicability of the subject. If there is a clear contradiction, a "destructive" effect occurs, reducing the priority of the subject. Through this interference operation, the system calculates an interference score, with a higher score indicating a stronger consistency between the fractal map and the subject prediction.

[0136] The interference score is merely a measure of quantum state compatibility and must be re-verified against the symbolic rules built into this invention (including accounting standards and account usage specifications). If an account does not comply with the corresponding rules (e.g., if the amount exceeds a certain limit or the account type and project type conflict), it will be removed from the candidate list. For transaction nodes that still contain conflicts or anomalies after symbolic rule verification, the system will automatically prompt for manual review and correction. Manual reviewers can review the interference score calculation process and rule conflicts and ultimately confirm the account or make revisions.

[0137] All transaction nodes that have passed the quantum interference plus symbol rule screening and manual review will eventually select the target account and update it into the data set; after this step, the node officially has a certain account information, called "calibrated account information", and is returned to the fractal financial semantic map, which can be directly used in the subsequent accounting and auditing links.

[0138] In practical applications, if the nodes are represented by vectors in the fractal graph Indicates that the candidate subject list is a vector Represented, the quantum hybrid aggregator will map the combination of the two into a “quantum state” and evaluate their superposition effects.

[0139] When the node When the "subject state" and "subject state" converge, they will be mutually constructive (interference score increases), and vice versa, they will be mutually destructive (score decreases). The specific quantum mathematical details are generalizable knowledge in this field. The present invention only needs to list the main interference calculation logic in the specification. The system will set an interference score threshold. If the node Candidates for a subject Calculated interference score Exceed , it is determined to be high confidence; those below the threshold are downgraded. In one embodiment, if a transaction node is superimposed with multiple candidate items, only one item has an interference score significantly higher than , then intuitively mark this subject as the preferred attribution.

[0140] Symbol rules typically include amounts, account types, and project regulations from financial regulations, and may even encompass internal corporate processes (such as travel expense limits and departmental account availability). Candidate accounts with high interference scores are individually matched against these rules. Any non-compliance (exceeding limits, conflicting types) is immediately removed and the reason for the conflict is recorded. For the remaining accounts, if only one meets the criteria, it is approved. Otherwise, if multiple accounts still pass or all conflict, the transaction node is placed in the "manual review required" queue.

[0141] In the user interface, the system centrally displays the interference score, symbol rule matching results, and conflict point descriptions. If the auditor finds the item reasonable and feasible, it will be approved. If the system identifies an error, it can be modified to another item and a note can be provided. The audit action is written back to the Fractal Financial Semantic Graph, becoming an important basis for subsequent algorithm self-learning or system traceability.

[0142] After completing this process, each transaction node has a "Verified Account Information" field. Nodes without a valid account or in dispute remain in a pending state for further manual intervention. Ultimately, these verified nodes can be automatically referenced in subsequent accounting processing to generate vouchers or financial entries.

[0143] Preferably, the step of inputting the fractal financial semantic graph and the subject candidate results into the quantum hybrid aggregator comprises:

[0144] The node attribute vectors of the fractal financial semantic graph and the subject candidate result vectors are used to construct quantum state sequences respectively;

[0145] Based on the principle of quantum state superposition, constructive merging of similar vectors and destructive processing of conflicting vectors are performed to obtain the interference score.

[0146] Candidate subjects with interference scores higher than the threshold are taken as priority attribution suggestions to provide input data for subsequent symbol rule verification.

[0147] In a fractal financial semantic graph, each node has several numerical attributes, such as amount, fractal dimension, and number of transactions. The subject candidate results are information about the subjects that the artificial intelligence or rules in the previous stage have determined to be the possible subjects of a transaction node, and can be represented as several vectors. This invention regards the fractal graph node vectors as "fractal forms" and the subject candidate vectors as "subject states." Each of these forms a sequence of quantum states, which are then superimposed in a quantum hybrid aggregator.

[0148] In conventional weighted or simple voting fusion, the system often only performs linear addition or maximum selection. This invention introduces the concept of quantum superposition: when two vectors (the fractal state and the subject state) are similar in their primary characteristics, the system interprets this as a constructive effect, resulting in an increase in the score after interference. When the two vectors conflict in their primary characteristics, they are considered destructive, resulting in a decrease in the score after interference. The resulting interference score reflects the degree of match between the fractal financial information and the subject candidate; a larger value indicates a more likely correct subject candidate.

[0149] If the interference score exceeds the preset threshold, the candidate subject is considered a priority and provides input data for subsequent symbol rule verification. If the score is below the threshold, it means that the candidate subject does not match the fractal map information and needs to be abandoned or carefully checked.

[0150] In practical applications, transaction nodes For example, its fractal graph vector can contain several elements, such as ,in represents the fractal dimension, Indicates the frequency of travel, Represents the total amount or other key indicators. The subject candidate result vector can be recorded as ,in Represents a candidate subject, which is a node Possible matching accounting account information, dimensions may include account category vector coding, historical attribution success rate, etc.

[0151] The present invention integrates the and Mapped into two quantum states respectively and .when and If the similarity is high, the superposition will be constructive, and if the similarity is low, the superposition will be destructive. The result after interference is recorded as The interference score is calculated by the system The specific superposition formula can refer to the interference principle of quantum mechanics in theory, and there is no need to fully expand it here.

[0152] Interference score Measuring Node and subjects The degree of fit between fractal and subject level. ,in If it is the threshold value, it is considered as the priority attribution subject; if it is lower than In one embodiment, the node The superimposed interference score of the sub-form and the subject "office expenses" If it is above the threshold, the system will treat "office expenses" as the most promising subject label for subsequent symbol rule verification.

[0153] Candidate subjects that reach the threshold are recorded in the "Priority Attribution Suggestions" list. If there is no subsequent conflict in the symbol rules, they can directly become the final selected subject information. If there is a conflict, they will be weighed during manual review or secondary processing.

[0154] In one embodiment, it is assumed that the node For a research project expense, the system provides three candidate items: "material expense", "consulting fee" and "equipment expense" through preliminary text analysis, and maps each to 、 、 .

[0155] Quantum hybrid aggregators will Fractal vector Perform interference operations on the above three groups of subject states respectively to obtain interference scores .like If the amount is the highest and exceeds the threshold, "material cost" will be listed as the priority attribution recommendation.

[0156] Through quantum state superposition assessment, rather than relying solely on simple scoring or voting, the system integrates the degree of compatibility between fractal attributes (such as transaction frequency and amount range) and account characteristics, reducing the conflicts or omissions that can occur with traditional linear fusion. The combination of fractal dimensions and quantum interference significantly enhances the flexibility of multidimensional financial information fusion, enabling more accurate identification of appropriate accounts and providing a fine-grained screening for subsequent symbolic rule verification.

[0157] Preferably, when using symbol rule verification, candidate accounts with high interference scores are judged one by one according to transaction amount limits, compliance accounting account regulations and internal approval terms, and candidate accounts that do not meet the rules are eliminated and archived.

[0158] In this invention, the quantum hybrid aggregator has selected several candidate accounts with "high interference scores" for each transaction node. These accounts have a strong theoretical match. However, the financial system must be implemented under the constraints of real-world compliance and internal audits. This step ensures compliance through symbolic rule verification to prevent incorrect accounts or illegal accounting behavior. Symbolic rules may include:

[0159] Transaction amount limit: upper limit of account amount, lower limit of funds for special account, etc.

[0160] Legality regulations for accounting accounts: For example, fixed asset accounts only apply to single transactions ≥ a certain amount;

[0161] Internal approval terms: if the signature of the project leader is required or if the amount exceeds a certain financial amount, additional approval is required, etc.

[0162] If a category candidate violates the above rules—for example, if the transaction amount exceeds the applicable scope of the category or the transaction attributes do not meet internal approval requirements—the category will be removed from the candidate list and archived. Archiving can also record the "reason for removal" to provide a basis for subsequent review or secondary verification. Meanwhile, the remaining categories that meet the rules will be retained for the next stage. If only one category remains, it will be directly designated as the final category information.

[0163] In practical applications, after the quantum hybrid aggregator generates candidate subjects, the system usually lists the subjects in descending order of interference scores:

[0164]

[0165] in The highest interference score, Minimum. The symbol rule engine checks each Make judgments, check factors such as amount and approval terms, and immediately eliminate any unqualified items, and write the elimination records into the log.

[0166] If there is a minimum amount requirement for the "Fixed Assets" account , then the system determines if the transaction amount , then this subject candidate will be invalid immediately; if the subject "office expenses" has a maximum amount limit Or it is only applicable to a specified department, then transactions exceeding the limit or transactions that do not meet the department requirements are also eliminated here; in one embodiment, if the node attribute vector shows that the amount is too high and the candidate account is "office expenses", the system searches the symbol rule table and finds that it exceeds the limit. If the scope is not met, remove the "Office Expenses" account from the candidate list and record the "Reason for Excess".

[0167] If the rule base stipulates that the single transaction amount is greater than a certain threshold Items that require the signature of a leader can be recorded in "Conference Fees". The system will then check whether the node attributes contain a leader signature field or related approval fields. If the necessary approval mark is missing, the candidate item will be invalid. The system will indicate "Approval conditions not met" when archiving for subsequent manual review or re-entry.

[0168] When the subject Removed, the system will remove the entry An audit log or database table is written to facilitate subsequent tracing of why the quantum interference result, despite being high, still does not meet actual compliance rules. If only one or zero candidates remain, the system automatically executes the subsequent actions: if only one remains, it is considered the final candidate; if there are zero candidates, it indicates that the node fails the rules and enters the "manual review" process.

[0169] In one embodiment, a transaction node The amount is 2,500 currency units. After quantum interference, three categories are given: "office expenses", "conference expenses", and "material expenses", with scores decreasing in descending order. Symbol rule verification:

[0170] "Office expenses" received the highest score, but the company stipulated that the upper limit for "office expenses" was 2,000; since 2,500 exceeded the limit, it was eliminated and the reason for "excess" was recorded; "Conference expenses" required the signature of the leader and supporting documents for venue fees, but this node lacked a signature and did not meet the approval terms, so it was eliminated again; although "Material expenses" had the lowest score, it did not trigger any violation conditions and was ultimately retained.

[0171] The system retains a valid account "Material Cost", completes automatic judgment and outputs "Corrected Account".

[0172] This link ensures the compliance and depth of review of financial processing, and builds a bridge between AI automated fusion scores and traditional financial standards; detailed records of elimination operations are kept to facilitate audits or manual reviews; if the rules are upgraded, previous elimination records can also be traced back for replenishment; the credibility of the subject information output in the final link is greatly improved, which facilitates downstream accounting and management decisions.

[0173] Preferably, when auditing and correcting transaction nodes with conflicts or anomalies, the transaction number, accounts before and after correction, operator and modification time are recorded, and this correction information is associated with the interference score and written into the calibrated account data to form updated account information.

[0174] After verification by the quantum hybrid aggregator and symbolic rules, if a transaction node still exhibits interference score conflicts, multiple account competition, or non-compliance with audit rules, it will be classified as a "conflicting or abnormal transaction node." At this stage, further review and correction by humans or auxiliary algorithms are required to ensure the final account information is correct and reliable.

[0175] Once finance personnel or system administrators modify the account at a transaction node, it means that the recommendations previously provided by AI or preliminary rules are no longer fully suitable for the actual business. Within the AI ​​learning loop, retaining information about the account before and after the correction, the reason for the correction, and the interference score will provide key information support for subsequent model optimization and audit tracking.

[0176] The final confirmed or revised account information is written back into the system's "corrected account data" collection, creating a unique version that can be used for downstream accounting, reporting, or auditing. AI can also leverage these real-world corrections to improve the accuracy of subsequent matching or interference calculations, achieving adaptive evolution.

[0177] In actual applications, the interface displayed by the system to the reviewer usually includes:

[0178] Transaction number, such as "TX20230910-001"; interference score, symbol rule verification results; original candidate subjects (several possible options) and reasons. The reviewer clicks or edits in the interface to specify the final subject or split multiple items, and the operator ID (such as "UserID=FIN001") and the timestamp are combined. Record them together.

[0179] If the subject is changed from "office expenses" to "material expenses", then record: "Subject before correction = office expenses", "Subject after correction = material expenses"; Interference score and They can be saved together, and later it can be evaluated why the items with high interference scores were not finally adopted (possibly due to rule conflicts or lack of supporting invoices, etc.).

[0180] In this invention, the interference score is a key metric derived from quantum state superposition. Associating it with modification information allows the system to identify scenarios where, during subsequent self-learning, even high scores may be subject to manual correction, thereby optimizing interference calculations or weighting strategies. In practice, the system can log data in a dedicated "correction log" table or write "final_subject=XX" and "interference_score=YY" in the node attributes of the fractal financial semantic graph.

[0181] After the audit is completed, the system updates the final account field of the node in the "calibrated account data" library to the revised result, and marks the previous account as "invalid" or "overwritten". In one embodiment, if the audit determines that the account should be "conference fee" instead of "travel fee", the transaction node in the "calibrated account data" will be updated. Store "finalSubject=Conference Fee" and delete or mark the previously stored "Travel Expenses" as deprecated.

[0182] In one embodiment, the node After quantum interference analysis, the "Material Fees" item showed a high interference score, but symbol verification revealed the presence of a "Conference"-related field in the transaction description. Manual review determined that the item should be classified as "Conference Fees."

[0183] System record: Transaction number: "TXID=T_{7}"; Subject before correction: "Material expenses"; Subject after correction: "Conference expenses"; Interference score: ;Operator: "FIN007", modification time: "2023-09-10T15:20";When writing back the corrected account data, finally set T_{7} to "Conference Fees" and retain the correction log for self-study or audit review.

[0184] Artificial intelligence or preliminary automated solutions struggle to capture all financial details, so manual intervention and record corrections constitute a crucial safety valve. This correction process can be used as a high-value "adversarial example" during the next model training, helping the system recognize scenarios where high interference scores still violate business reality or symbolic rules, thereby improving the next round of quantum interference or account predictions. Auditors can also review the logs for transaction "T_{7}" to understand why the system initially identified it as "material expense" and why it was changed to "conference expense" during the audit, enhancing the traceability and transparency of the entire financial process.

[0185] Accounting entries are generated based on the calibrated account data and ledger archiving is performed. Quantum hashing technology is used to record decision information of the transaction processing and correction processes. After analyzing the decision information, the fractal dimension calculation parameters and the weight factors of the quantum hybrid aggregator are adjusted.

[0186] In the previous process, each transaction node has finalized account information (possibly through a quantum hybrid aggregator, symbolic rule verification, and manual review). This is called "calibrated account data." Based on this data, the system automatically generates accounting entries for each node. For example, in a debit-entry accounting system, nodes correspond to debit or credit accounts, forming paired accounting entries to meet financial system requirements.

[0187] This invention incorporates the concept of "quantum hashing" when archiving ledgers: Decision information about transaction processing and corrections is written into a blockchain or distributed ledger, and the hash value is quantum-securely encrypted to ensure it cannot be subsequently tampered with or forged. In this process, "decision information" encompasses a series of operations, from manual review to automated interference assessment, and quantum hash signatures ensure the integrity of the entire processing chain.

[0188] After accounting is complete, the system centrally analyzes the decision information stored in the quantum hash ledger. By aggregating large amounts of "calibrated subject data" and process records, it is possible to test the actual performance of the quantum hybrid aggregator in subject assessment and the accuracy of fractal dimension calculations in graph modeling. If high system errors are detected in certain scenarios or node types, model elements such as fractal dimension calculation parameters and quantum interference weighting factors can be automatically or semi-automatically adjusted to achieve adaptive correction and optimization of subsequent processing.

[0189] In actual applications, for each transaction node , reads the established account information (such as "office expenses", "material expenses", etc.), and retrieves the corresponding accounting account code in the internal account mapping table; according to the debit and credit accounting principle, it merges key fields such as amount and transaction time to form the following entry example: debit account: account code, amount; credit account: account code, amount. In one embodiment, if This represents a purchase transaction with the final account being "Material Cost." The system then generates a debit entry for Material Cost and a credit entry for Bank Deposit or Cash. This information is then written to the company's financial system, completing the accounting process.

[0190] The system collects decision-making information generated during the transaction processing and correction process (including interference scores, reasons for symbolic rule conflicts, and manual review and modification records), forms a sequence of events, and packages them. Using quantum hashing technology, a hash value is generated for this packaged information and digitally signed using a quantum-resistant algorithm (such as a lattice-based or multivariate polynomial), providing strong tamper-resistance. This digital signature can be written to a blockchain or local distributed ledger, ensuring seamless traceability for subsequent audits or model improvements.

[0191] Periodically or when the number of nodes reaches a certain scale, the system automatically reads all processing and correction records from the quantum hash ledger. By analyzing metrics such as the final correction rate of each node, common conflict causes, and interference score distribution, the system assesses whether existing models or rule bases need to be updated. If the fractal dimension calculation is found to exhibit deviations in high-value transactions, the multi-level grid splitting strategy may be adjusted. If the quantum hybrid aggregator experiences high conflict rates in certain items, the corresponding weighting factors may be adjusted downward or upward.

[0192] Taking the fractal dimension as an example, if the system finds through statistics that the number of nodes covered by large-value transactions is abnormally distributed in the high-scale interval, it can be fine-tuned. Threshold values ​​or adjustments to multi-scale segmentation methods; Example of weight factor adjustment for quantum hybrid aggregators: If a subject is often manually rejected under high interference scores, it means that the interference weight assigned to the subject is too high, and its superposition gain can be lowered; these parameter changes will automatically take effect during subsequent transaction processing, enabling the system to achieve adaptive correction and improvement, reducing the recurrence of similar errors.

[0193] Preferably, when generating accounting entries based on the calibrated account data and performing ledger archiving, a transaction identification tag, specific account information and related party information are configured for each entry.

[0194] After quantum interference, symbol rule verification, and manual review, each transaction node has finalized the accounting entries, known as "calibrated account data." In traditional accounting systems, this final information must be converted into specific entries to ensure debit and credit symmetry and meet accounting standards.

[0195] When the system automatically generates accounting entries, it adds the following key information to each entry to facilitate subsequent auditing, retrieval, and statistics:

[0196] Transaction identification tag: Indicates the specific transaction or transaction node from which the entry originates (for example, "TXID=TX20231015-001").

[0197] Account information: the accounting account code used for the entry, as well as the classification or sub-account of the account in the internal financial standards.

[0198] Transaction information: Record the name and number of the supplier, customer, or department associated with this transaction to facilitate subsequent report analysis and fund flow tracking.

[0199] In the artificial intelligence financial identification and accounting system, this archiving process is not just about storing records, but also provides a data source for subsequent algorithm audits or self-learning: when auditing or reviewing model performance in the future, the mapping relationship between the transaction in the graph, quantum interference and the final entry output can be quickly matched.

[0200] In practice, transaction identification tags can take the form of a unique system-wide number (e.g., "TXID=TX20231015-001"), already bound to the transaction node in previously calibrated account data. When an entry is generated, this tag is written into the entry field (a common practice is to add a "Transaction ID" field to the database's "Voucher Line Table"), maintaining a one-to-one or one-to-many traceability relationship between the entry and the graph node.

[0201] The calibrated account data indicates the final account, such as "Office Expenses-6601" or "Material Expenses-5101". The system obtains the complete code based on the internal "Account Dictionary" query and fills in the debit or credit position in the entry. In one embodiment, if a transaction node The final account is "Material Cost", and the system automatically fills in "Material Cost (Account Code 5101), Amount X, Debit / Credit Mark D / C" etc. in the "Entry Line" of the accounting software.

[0202] Trading party information can include supplier numbers, customer numbers, or internal department codes, such as "S001" and "Dept02." When generating journal entries, these are added to auxiliary accounting items (for example, in a debit / credit accounting system, auxiliary accounting dimensions are often used for projects, suppliers, and departments). This facilitates the subsequent issuance of trading details, statistical reports, and exception analysis.

[0203] Once each entry is complete, the system performs an "archive" operation, which typically involves writing the entry to the master ledger or ERP financial database; recording the time the entry was created, and, if a blockchain or quantum hashing solution is available, appending an unalterable hash record for subsequent auditing. In this scenario, AI-powered recognition and accounting results are formally implemented as traceable accounting vouchers.

[0204] In one embodiment, the node This transaction, calibrated as "Conference Fees (Account 6503)," has an amount of 1,000 currency units and is traded to "Dept05." When generating the journal entry, the system automatically creates a debit and a credit line: "Debit: Conference Fees (6503), Amount 1,000; Credit: Bank Deposit (1002), Amount 1,000." It also adds "TXID=T_{8}" and "Dept=Dept05" to the auxiliary fields. This entry line is successfully archived in the accounting software library, providing clarity for subsequent financial reports, account inquiries, and manual review.

[0205] Implementing the fractal financial semantic graph and AI analysis results into accounting entries creates a complete end-to-end process: from data cleaning → graph analysis → quantum interference → account verification → final entry, all in one go. Because the entries contain transaction identification tags and related transaction information, the system can retroactively trace and verify the cause and effect of any specific transaction node, greatly improving audit efficiency and transparency. It also provides accurate "ground truth" or annotated data for subsequent iterations of the AI ​​model. If similar transactions are encountered again, previously archived results can be used to make more accurate account judgments or risk assessments.

[0206] Preferably, quantum hashing technology is used to write the decision information of the transaction processing process and the correction process, and after analyzing the decision information, the grid splitting scale used in the fractal dimension calculation and the interference parameters of the quantum hybrid aggregator are adjusted.

[0207] In this invention, decision-making information for every transaction and subsequent corrections (such as manual review, account revisions, and anomaly annotations) is recorded in the ledger. When quantum hashing is used in the ledger, this information is quantum-securely hashed and signed, rendering it tamper-proof during transmission or storage and ensuring its integrity and reliability during subsequent audits and reconciliations. The core of quantum hashing technology is the use of quantum-resistant mathematical algorithms (such as those based on lattices or multivariate polynomials), ensuring that hash values ​​remain highly resistant to cracking even with future quantum computing capabilities.

[0208] The decision-making information collected mainly includes:

[0209] Transaction process: node attributes, fractal calculation results, quantum interference scores, symbol rule conflicts, etc.; correction process: manual or automatic correction of account information before and after, timestamps, operators, correction reasons, etc.

[0210] After the system analyzes this information regularly or in a triggered manner, if it finds that the existing fractal dimension calculation or quantum hybrid aggregator cannot accurately reflect the actual business situation, it will adjust the grid splitting scale used in the fractal dimension calculation and the interference parameters in the quantum hybrid aggregator to make the next round of processing closer to the real scenario.

[0211] In practical applications, the fields written into the quantum hash ledger are:

[0212] Transaction node identification: for example, "TXID=TX20231101-002"

[0213] Quantum interference stage information: interference fraction, main vector comparison

[0214] Decision-making process: revised subjects, conflict reasons, manual review notes

[0215] Timestamp: record writing time

[0216] The above information is quantum hashed and a tamper-proof signature is generated, which is then written into the distributed ledger for safekeeping.

[0217] Regularly summarize all recently processed transaction nodes, compare their initial fractal dimension assessments, quantum interference scores, and final results to see which nodes have a higher correction rate or which items have more conflicts. If multiple nodes show obvious inconsistencies in the same amount range or transaction frequency range, it is speculated that the grid division in the fractal dimension calculation does not handle the interval well, and it may be necessary to add a denser split scale to the amount axis or time frequency axis. If the quantum interference scores of multiple nodes are apparently high but are often rejected by the sign rule, it means that the interference parameter setting of the quantum hybrid aggregator may be too high, and it is necessary to lower the gain or refine the interference judgment threshold.

[0218] In the fractal dimension calculation stage, attribute dimensions such as amount, transaction frequency, and number of inter-entity transactions are usually split into multiple levels of grids:

[0219]

[0220] From the perspective of decision-making information, arrive If there are often a lot of misjudged transactions between the intervals, the grid range can be increased, decreased, or redistributed to allow the fractal dimension to reflect the node attribute distribution in a more fine-grained manner. , the system may locally refine and subdivide the original uniform grid, so that fractal statistics can capture node details more accurately.

[0221] Quantum Hybrid Aggregator for Fractal Vectors With subject candidate vector When performing interference, a gain factor or interference threshold is often used to determine constructive or destructive scores. If analysis shows that a certain type of transaction node has a generally high interference score but is being corrected extensively, this indicates that the gain setting is excessive. Alternatively, if a large number of nodes fail to achieve a high enough interference score, resulting in candidate rejection, this indicates that the interference parameters can be appropriately relaxed. By automatically comparing the correction rate with the conflict rate, the system can semi-automatically generate an "interference gain correction" plan and apply it during the next processing.

[0222] In one example, within a month, 20 or more transaction nodes with amounts ranging from 3,000 to 5,000 occurred. While the quantum interference scores were generally high, the actual manual review and correction rate reached 50%. Analysis of the correction records in the quantum hash ledger revealed that these transactions exhibited subtle over-coarsening of the fractal dimension.

[0223] The amount range of 3000-5000 is split into a finer grid scale for recalculating the fractal dimension; the quantum interference gain parameters are fine-tuned to reduce excessive preferential treatment for the medium amount range;

[0224] After the next batch of similar transactions were processed, the manual correction rate dropped significantly, indicating that the system's adaptability had been improved.

[0225] Through the above solution, the quantum hash ledger of the present invention provides secure audit traceability; after reading these "real decisions", the fractal dimension and quantum interference make adaptive fine-tuning to reduce misjudgments; the overall guarantee is that financial automation processing has the ability to continue to evolve in the middle and late stages, constantly adapting to real business scenarios and reducing manual rework.

[0226] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware.

[0227] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A financial data identification and accounting method based on artificial intelligence, characterized in that: The following steps are involved: Obtain financial source data and perform preliminary formatting and verification to obtain usable financial data; The transactions and entities in the available financial data are associated to establish an initial semantic graph, wherein the entities include counterparties and projects, and the fractal dimension is calculated based on the node attribute vector, and fractal similarity edges or nested relationship edges are generated, so that the transactions and entity nodes form a fractal financial semantic graph. Specifically, the fractal dimension is calculated based on the node attribute vector, and fractal similarity edges or nested relationship edges are generated, so that the transactions and entity nodes form a fractal financial semantic graph. Specifically, the following steps are performed: a multi-level grid splitting is used to perform discretization statistics on the amount range, transaction frequency and the number of associated entities; the fractal dimension of the node is calculated for the discretized data, and when the difference between the fractal dimension and the fractal dimension of another node is lower than a preset threshold, a fractal similarity edge is generated; when the node attributes are hierarchically recursively associated on the main feature components, a nested relationship edge is established and the association is updated to the fractal financial semantic graph; The fractal financial semantic graph and the subject candidate results are input into the quantum hybrid aggregator together, where the subject candidate results are the list of accounting subjects given for each transaction node, and the quantum state superposition method is used to evaluate the constructive or destructive effect to determine the interference score. Specifically, the fractal financial semantic graph and the subject candidate results are input into the quantum hybrid aggregator together, and the quantum state superposition method is used to evaluate the constructive or destructive effect to determine the interference score. Specifically, the node attribute vector of the fractal financial semantic graph and the subject candidate result vector are respectively constructed into quantum state sequences; based on the principle of quantum state superposition, similar vectors are constructively merged and conflicting vectors are destructively processed to obtain the interference score; the candidate subjects with interference scores higher than the threshold value are used as priority attribution suggestions to provide input data for subsequent symbol rule verification; after the symbol rule verification, the transaction nodes with conflicts or anomalies are reviewed and corrected, and the updated subject information is output to obtain the calibrated subject data; Accounting entries are generated based on the calibrated account data and ledger archiving is performed. Quantum hashing technology is used to record decision information of the transaction processing and correction processes. After analyzing the decision information, the fractal dimension calculation parameters and the weight factors of the quantum hybrid aggregator are adjusted.

2. The method according to claim 1, characterized in that The process of obtaining financial source data and performing preliminary formatting and verification includes: Use character detection algorithms to perform optical recognition on paper documents or scanned images to obtain text records; Align the fields of the recognized text records with those in the structured business system, unify the currency unit and set the date format, and filter out records that do not meet the field integrity requirements; Entries whose numerical ranges do not meet the preset thresholds are marked as abnormal data to be reviewed, and available financial data that meets the field completeness requirements is output.

3. The method according to claim 1, characterized in that The steps of associating transactions and entities in the available financial data to establish an initial semantic graph include: Use transaction identifiers or unique indexes to map each record to the corresponding entity number to generate the basic nodes of the initial semantic graph; When multiple transactions occur under the same entity number at different times, the amount, category, or current account information is recorded in the node attributes in chronological order; Records that fail to match any entity number are marked as isolated nodes.

4. The method according to claim 1, wherein After generating fractal similarity edges or nested relationship edges, the abnormal transaction nodes are preliminarily identified by comparing the transaction frequency and cumulative amount between nodes, and the preliminary identification results are saved in the fractal financial semantic graph.

5. The method according to claim 1, wherein When using symbol rule verification, candidate accounts with high interference scores are judged one by one based on transaction amount limits, compliance accounting account regulations, and internal approval terms, and candidate accounts that do not meet the rules are eliminated and archived.

6. The method according to claim 1, characterized in that When reviewing and correcting transaction nodes with conflicts or anomalies, record the transaction number, accounts before and after the correction, operator, and modification time. Associate this correction information with the interference score and write it into the calibrated account data to form updated account information.

7. The method according to claim 1, characterized in that When generating accounting entries based on calibrated account data and performing ledger archiving, each entry is configured with a transaction identification tag, specific account information, and related party information.

8. The method according to claim 7, characterized in that Quantum hashing technology is used to write decision information of the transaction processing process and the correction process. After analyzing the decision information, the grid splitting scale used in the fractal dimension calculation and the interference parameters of the quantum hybrid aggregator are adjusted.

Images