Bill cooperative management method and system based on multi-source heterogeneous data fusion

Abstract

The embodiment of the application provides a kind of based on multi-source heterogeneous data fusion bill cooperative management method and system, belong to data processing technical field.The method comprises: according to standardization bill data carries out hash calculation, obtains unique identifier.Standardization bill data corresponding graph node is obtained, and target graph node is obtained.The path of graph root node to target graph node is target structure path.According to unique identifier and behavior log, the complete behavior sequence of standardization bill data is obtained.According to target structure path and complete behavior sequence, behavior structure joint sequence is constructed.Behavior structure joint sequence is input to structure perception embedding model, and sequence embedding vector is obtained.According to structure behavior consistency evaluation function, behavior structure joint sequence and sequence embedding vector, structure behavior inconsistency score and abnormal type label are obtained.According to structure behavior inconsistency score and abnormal type label, suggestion text is generated and is shown.Accurate identification exception bill is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a bill cooperative management method and system based on multi-source heterogeneous data fusion. BACKGROUND

[0002] In rail transit engineering, district development and related supporting industries, bills serve as a key medium between engineering execution and financial disbursement, and run through multiple links such as project planning, contract performance, progress verification, fund payment and tax archiving, and bear data interaction and process driving between multiple systems and roles. The same bill may exist in different systems in different codes, formats, times and amounts, resulting in serious fragmentation of bill data at the semantic and process levels, which not only has high reconciliation cost and poor process penetration, but also is difficult to support attribution judgment and behavior audit with engineering structure as the core. The common method in the industry at present mainly depends on field comparison or fixed rule matching, which cannot identify the legal behavior path of bills in the cross-system life cycle, and is difficult to find hidden irregular risks caused by process bypassing, step execution and permission mismatch. Therefore, how to accurately identify abnormal bills has become a technical problem to be solved. SUMMARY

[0003] The main purpose of the embodiments of the present application is to provide a bill cooperative management method and system based on multi-source heterogeneous data fusion, which aims to accurately identify abnormal bills.

[0004] To achieve the above-mentioned purpose, the first aspect of the embodiments of the present application provides a bill cooperative management method based on multi-source heterogeneous data fusion, which comprises:

[0005] obtaining standardized bill data;

[0006] performing hash calculation on each of the standardized bill data to obtain a unique identifier;

[0007] obtaining behavior logs and a preset structure graph; wherein the structure graph comprises a graph root node and a plurality of graph nodes;

[0008] obtaining the graph node corresponding to the standardized bill data to obtain a target graph node; wherein the path from the graph root node to the target graph node is a target structure path;

[0009] obtaining a complete behavior sequence of each of the standardized bill data according to the unique identifier and the behavior logs;

[0010] constructing a behavior structure joint sequence according to the target structure path and the complete behavior sequence;

[0011] inputting the behavior-structure joint sequence into a preset structure-aware embedding model to obtain a sequence embedding vector;

[0012] performing calculation according to a preset structure-behavior consistency evaluation function, the behavior-structure joint sequence and the sequence embedding vector to obtain a structure-behavior inconsistency score and an abnormal type label;

[0013] generating and displaying a suggestion text according to the structure-behavior inconsistency score and the abnormal type label.

[0014] In some embodiments, the generating and displaying a suggestion text according to the structure-behavior inconsistency score and the abnormal type label comprises:

[0015] performing scoring according to a preset collaborative suggestion scoring function, the sequence embedding vector, the structure-behavior inconsistency score, the abnormal type label and a preset role responsibility vector to obtain a collaborative suggestion score;

[0016] if the collaborative suggestion score is greater than a preset suggestion threshold, generating the suggestion text according to the structure-behavior inconsistency score and the abnormal type label;

[0017] displaying the suggestion text in a ticket view corresponding to the role.

[0018] In some embodiments, the performing scoring according to a preset collaborative suggestion scoring function, the sequence embedding vector, the structure-behavior inconsistency score, the abnormal type label and a preset role responsibility vector to obtain a collaborative suggestion score comprises:

[0019] the collaborative suggestion scoring function is:

[0020] ;

[0021] wherein, represents a normalized ticket data , the system determines whether it should be intervened by the role focus and intervention of the collaborative suggestion score, represents a sequence embedding vector, represents a role focus preference vector, represents transpose, represents a structure-behavior inconsistency score, represents an indicator function, represents an abnormal type label , represents a role focus on the abnormal set. 、 represents a weight parameter.

[0022] In some embodiments, the standardized bill data is obtained, including:

[0023] Obtaining original bill data from different systems;

[0024] According to the original bill data, field unification processing is performed to obtain initial bill data;

[0025] According to the initial bill data, retrieval is performed, and the initial bill data meeting the preset condition is merged to obtain the standardized bill data.

[0026] In some embodiments, the standardized bill data corresponding to the graph node is obtained to obtain a target graph node, including:

[0027] According to a preset anchor scoring function, a score of the standardized bill data mapped to each graph node is calculated to obtain an anchor score set;

[0028] The anchor score with the highest score in the anchor score set is obtained to obtain a target anchor score;

[0029] The graph node corresponding to the target anchor score is obtained to obtain the target graph node.

[0030] In some embodiments, the anchor scoring function is calculated according to the preset anchor scoring function, and the score of the standardized bill data mapped to each graph node is obtained to obtain an anchor score set, including:

[0031] The anchor scoring function is:

[0032] ;

[0033] ;

[0034] ;

[0035] Wherein, represents the anchor score of the standardized bill data mapped to the graph node , represents the proportion similarity of the amount in the standardized bill data and the historical amount of the graph node , represents the probability value of the bill issuing time of the standardized bill data under the time density function of the graph node , represents a structure level penalty regular term, denotes the length of the longest common sub-path missing between the bill contract path and the node path, denotes the structure penalty coefficient, denotes the amount field of the standardized bill data , denotes the graph node , denotes the average amount of historical bills of the graph node , denotes the contract number of the standardized bill data , denotes the engineering level path where the graph node , , respectively denote the weight coefficients of the amount similarity, the time similarity and the structure level penalty term in the overall anchor score.

[0036] In some embodiments, the structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector are calculated according to the preset structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector to obtain the structure behavior inconsistency score and the abnormal type label, comprising:

[0037] The structure behavior consistency evaluation function is:

[0038] ;

[0039] ;

[0040] wherein, denotes the structure behavior consistency score, denotes the sigmoid activation function, denotes the sequence embedding vector, denotes the normal behavior embedding mean of the structure node , denotes the position offset penalty term, denotes the structure behavior inconsistency score, denotes the weight coefficient of the position offset penalty term.

[0041] To achieve the above purpose, a second aspect of the embodiment of the present application proposes a bill collaborative management system based on multi-source heterogeneous data fusion, the system comprising:

[0042] A first acquisition module is configured to acquire standardized bill data.

[0043] A first calculation module is configured to perform hash calculation on each standardized bill data to obtain a unique identifier.

[0044] A second acquisition module is configured to acquire behavior logs and a preset structure graph, wherein the structure graph comprises a graph root node and a plurality of graph nodes.

[0045] a third obtaining module, configured to obtain the graph node corresponding to the standardized ticket data, to obtain a target graph node; wherein a path from the graph root node to the target graph node is a target structure path;

[0046] a fourth obtaining module, configured to obtain a complete behavior sequence of each standardized ticket data according to the unique identifier and the behavior log;

[0047] a constructing module, configured to construct a behavior structure joint sequence according to the target structure path and the complete behavior sequence;

[0048] an inputting module, configured to input the behavior structure joint sequence into a preset structure-aware embedding model, to obtain a sequence embedding vector;

[0049] a second calculating module, configured to calculate according to a preset structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector, to obtain a structure behavior inconsistency score and an abnormal type label;

[0050] a display module, configured to generate and display a suggestion text according to the structure behavior inconsistency score and the abnormal type label.

[0051] To achieve the above object, a third aspect of the embodiments of the present application provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the method of the first aspect when executing the computer program.

[0052] To achieve the above object, a fourth aspect of the embodiments of the present application provides a computer readable storage medium, the computer readable storage medium storing a computer program, and the computer program being executed by a processor to implement the method of the first aspect.

[0053] The bill collaborative management method and system based on multi-source heterogeneous data fusion provided in the application obtain standardized bill data, perform hash calculation on each standardized bill data to obtain a unique identifier. Behavior logs and a preset structure graph are obtained, wherein the structure graph includes a graph root node and a plurality of graph nodes. A graph node corresponding to the standardized bill data is obtained to obtain a target graph node. The path from the graph root node to the target graph node is a target structure path. The complete behavior sequence of each standardized bill data is obtained according to the unique identifier and the behavior logs. The behavior structure joint sequence is constructed according to the target structure path and the complete behavior sequence. The behavior structure joint sequence is input into a preset structure perception embedding model to obtain a sequence embedding vector. The structure behavior inconsistency score and the abnormal type label are obtained by performing calculation according to the preset structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector. The suggestion text is generated and displayed according to the structure behavior inconsistency score and the abnormal type label. The sensitivity score and the suggestion of the bill abnormality are generated, so that different functional departments can accurately identify the risk focus and obtain personalized processing suggestions in their respective views, and the abnormal bill is accurately identified. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 is a flowchart of the bill collaborative management method based on multi-source heterogeneous data fusion provided by the embodiment of the application;

[0055] Figure 2 is a flowchart of step S101 in Figure 1

[0056] Figure 3 is a flowchart of step S104 in Figure 1

[0057] Figure 4 is a flowchart of step S109 in Figure 1

[0058] Figure 5 is a structural schematic diagram of the bill collaborative management system based on multi-source heterogeneous data fusion provided by the embodiment of the application;

[0059] Figure 6 is a hardware structural schematic diagram of the electronic device provided by the embodiment of the application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical scheme and advantages of the application more clear, the application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application, and are not used to limit the application.

[0061] ​​​It should be noted that although the system diagrams illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the system or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0063] The same bill may exist in multiple systems with different codes, formats, times, and amounts, resulting in severe fragmentation of bill data at both the semantic and process levels. This not only increases reconciliation costs and reduces process penetration, but also makes it difficult to support attribution judgments and behavioral audits centered on engineering structures. Current industry-wide approaches rely on field comparisons or fixed rule matching, failing to identify the legitimate behavior paths of bills across their lifecycles. This makes it difficult to detect hidden violations caused by process bypasses, skipped execution steps, and permission mismatches.

[0064] Based on this, embodiments of the present application provide a collaborative bill management method and system based on multi-source heterogeneous data fusion. This system aims to achieve full-link modeling capabilities, from data fusion, path reconstruction, anomaly identification, to multi-role response, by integrating standardized bill objects with graph-based engineering structures. First, a unique cross-system bill primary key is constructed based on key attributes such as contract number, amount, and time. This key is then assigned to a precise engineering structure node through a graph anchoring mechanism, achieving a transition from traditional field matching to structural semantic alignment. Second, behavioral sequence optimization is guided by the structural path, effectively suppressing system drift and process misalignment. Furthermore, by fusing behavioral paths with structural anchoring information, a structure-behavior consistency assessment mechanism is constructed. This incorporates a structure-aware embedding model and a process deviation regularization term to identify multiple types of anomaly patterns, including behavioral misalignment, structural overreach, and path anomalies. Furthermore, a multi-role collaborative perception and recommendation output mechanism is designed. By combining the response preferences of each business role in historical processes, sensitivity scoring and recommendation generation are performed for bill anomalies. This enables different functional departments to accurately identify risk priorities and obtain personalized handling recommendations within their respective views, thus accurately identifying anomaly bills.

[0065] The collaborative bill management method and system based on multi-source heterogeneous data fusion provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the collaborative bill management method based on multi-source heterogeneous data fusion in the embodiments of the present application is described.

[0066] Embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (AI) is to use digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0067] The basic technology of artificial intelligence generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. The software technology of artificial intelligence mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0068] The bill collaborative management method based on multi-source heterogeneous data fusion provided by the embodiments of the present application can be applied to a terminal, can also be applied to a server end, and can also be software running in the terminal or the server end. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc.; the server end can be configured as an independent physical server, can also be configured as a server cluster or a distributed system composed of multiple physical servers, can also be configured as a cloud server providing basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, and big data and artificial intelligence platform; and the software can be an application that implements the bill collaborative management method based on multi-source heterogeneous data fusion, etc., but is not limited to the above forms.

[0069] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0070] Please refer to Figure 1 , Figure 1is a flowchart of a bill collaborative management method based on multi-source heterogeneous data fusion provided by the embodiment of the present application, Figure 1 The method in the method can include but is not limited to steps S101 to S109.

[0071] Step S101, obtaining standardized bill data;

[0072] Step S102, performing hash calculation according to each standardized bill data to obtain a unique identifier;

[0073] Step S103, obtaining behavior logs and a preset structure graph; wherein the structure graph includes a graph root node and a plurality of graph nodes;

[0074] Step S104, obtaining a graph node corresponding to the standardized bill data to obtain a target graph node; wherein the path from the graph root node to the target graph node is a target structure path;

[0075] Step S105, obtaining a complete behavior sequence of each standardized bill data according to the unique identifier and the behavior logs;

[0076] Step S106, constructing a behavior structure joint sequence according to the target structure path and the complete behavior sequence;

[0077] Step S107, inputting the behavior structure joint sequence into a preset structure perception embedding model to obtain a sequence embedding vector;

[0078] Step S108, calculating according to a preset structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector to obtain a structure behavior inconsistency score and an abnormal type label;

[0079] Step S109, generating and displaying a suggestion text according to the structure behavior inconsistency score and the abnormal type label.

[0080] Please refer to Figure 2 In some embodiments, step S101 further includes but is not limited to steps S201 to S203:

[0081] Step S201, obtaining original bill data from different systems;

[0082] Step S202, performing field uniform processing according to the original bill data to obtain initial bill data;

[0083] Step S203, performing retrieval according to the initial bill data, merging the initial bill data meeting the preset conditions to obtain standardized bill data.

[0084] In step S201 of some embodiments, raw invoice data from three core systems is obtained, namely invoice information from an invoice inspection platform , including 、 、 , payment information from a company financial system , including invoice number, amount, payment time, and data from a contract management system , using only the contract number field .

[0085] In steps S202 to S203 of some embodiments, the above raw invoice data faces problems such as inconsistent field naming, different formats, and chaotic time standards when imported into the system through the interface. Therefore, these data need to be processed for field unification: all time fields are converted to UTC timestamp format; the amount field is converted to a unified unit of "yuan", with two decimal places; the invoice number field is uniformly named ; and the contract number field is uniformly .

[0086] Subsequently, the system performs retrieval based on 、 、 three fields as the main matching key, combines records from the invoice inspection platform and the financial system, and considers them as the same invoice when the following preset conditions are met:

[0087] completely consistent;

[0088] , wherein represents the from the invoice inspection platform;

[0089] time difference days.

[0090] Records that meet the above conditions at the same time are combined into standardized invoice data, denoted as , representing the th standardized invoice data, with uniform fields and consistent structure. Among them, the standardized invoice set .

[0091] Through the above steps S201 to S203, without introducing any external manual intervention, through field unification and approximate merging strategies, the invoice data from the invoice inspection platform, the financial system, and the contract system is fused to generate standardized invoice data, realizing collaborative management of invoices under multi-source heterogeneous data.

[0092] In step S102 of some embodiments, in order to make the bill have a global unique identifier in the entire system, a hash unique identifier is constructed. , which is generated as shown in the following formula (1):

[0093] , (1)

[0094] in, Represents a unique identifier, Indicates the invoice number, Indicates the amount field, Indicates the contract number, Indicates the unified timestamp format. Indicates the use of a deterministic hash function based on algorithm SM3. Indicates a field concatenation operation, which is encoded in UTF-8 and then input into the hash function.

[0095] In one example, an invoice , Yuan, ", , after splicing, the input is , whose hash value is the bill .

[0096] To prevent data confusion caused by hash collisions, the system sets up an automatic verification mechanism to detect whether but If there is a conflict, mark it and submit it for manual review.

[0097] In step S103 of some embodiments, the behavior log comes from the contract system, invoice verification platform, approval platform, etc. , each is a triple , the record unique identifier is The bill at the time The behavior that occurred . Structural map Including the graph root node and multiple graph nodes, each node Represents a structured engineering unit, containing the following fields:

[0098] : The contract number associated with the node;

[0099] : The average amount of historically attributed bills, used to assist in judgment;

[0100] : The engineering level path where the node is located, such as "Line > Section > Section > Unit";

[0101] : Parcel number or stake number, used for structure adjacency calculation;

[0102] : Invoice time distribution density function of historical bills.

[0103] See Figure 3 In some embodiments, step S104 further includes, but is not limited to, steps S301-S303:

[0104] Step S301: Calculate the score of the mapping of the standardized bill data to each graph node according to the preset anchor scoring function, to obtain an anchor score set;

[0105] Step S302: Obtain the anchor score with the highest score in the anchor score set, to obtain a target anchor score;

[0106] Step S303: Obtain the graph node corresponding to the target anchor score, to obtain a target graph node.

[0107] In step S301 of some embodiments, to solve the ambiguity of the structure graph attribution, the anchor scoring function shown in the following formulas (2)-(4) is proposed:

[0108] , (2)

[0109] , (3)

[0110] , (4)

[0111] Wherein, denotes the anchor score of the mapping of the standardized bill data to the graph node , denotes the proportional similarity of the amount in the standardized bill data to the historical amount of the graph node . denotes the probability value of the invoice time of the standardized bill data under the time density function of the graph node , which is generated by kernel density estimation. denotes the structure level penalty regular term, which is used to punish the mapping of the standardized bill data to the graph node with large difference in logical path. denotes the longest common sub-path missing length between the bill contract path and the node path, for example, "line-section-contract" matches "line-section-contract", which is missing one layer, . denotes the structure penalty coefficient, denotes the standardized bill data The amount field, Represents a graph node The average amount of historical vested notes, Represents standardized bill data The contract number, Represents a graph node The project level path. 、 、 They respectively represent the weight coefficients of amount similarity, time similarity, and structural level penalty in the overall anchor score, and are used to regulate the influence of each indicator on the final anchor score.

[0112] In some embodiments, in step S302 to step S303, the highest anchor score in the anchor score set is directly obtained to obtain the target anchor score, and the graph node corresponding to the target anchor score is the target graph node. Among all the related candidate graph nodes, the candidate graph node with the highest score is selected as the target graph node, as shown in the following formula (5):

[0113] , (5)

[0114] in, represents the target graph node, represents a candidate graph node, Represents standardized bill data Mapping to graph nodes Anchor score.

[0115] Furthermore, from the root node of the graph to the target graph node The path is the target structure path, the target graph node set , each Identifies the structural entity. Represents a set of mapping relationships between unique identifiers and target graph nodes, which is used for subsequent behavioral link modeling.

[0116] Through the above steps S301 to S303, each standardized bill data is accurately matched to its corresponding target graph node , thereby establishing a semantic connection between standardized invoice data and engineering structures. This structural attribution is not only a key prerequisite for visualizing the entire invoice lifecycle, but also provides the basic coordinates of physical locations and business units for subsequent process behavior modeling and anomaly detection.

[0117] It should be noted that in rail transit projects, the contracts on which the bills are based often correspond to multiple bidding sections, and there are multiple contracts with the same name or partially duplicate numbers. For example, the "M101 Line Civil Engineering Bid 02" and the "M101 Line Civil Engineering Bid 03" differ by only one digit in the contract number, and may use different abbreviations in different systems. At the same time, there are obvious structural differences between different lines, sections and facility types, so simple contract number matching can easily lead to attribution errors. Traditional primary key connection or fuzzy comparison methods cannot fully characterize these complex structures. In this step, the anchor scoring function is not only based on contract number matching, but also integrates multiple features such as amount, time distribution, and structural hierarchy, and designs a structural penalty regularization term for "engineering structure heterogeneity" to improve the stability and accuracy of anchoring.

[0118] In some embodiments, in steps S105 to S106, the behavior log is sorted by unique identifier Log Aggregation is performed to obtain the complete behavior sequence of each standardized bill data, as shown in the following formula (5):

[0119] , (5)

[0120] in, Represents standardized bill data The complete sequence of actions performed in all business systems, Indicates sorting in ascending order of time. Indicates behavior, Represents a unique identifier, A unique identifier for the ticket corresponding to a behavior event in the behavior log record.

[0121] Next, a behavior-structure joint sequence is constructed based on the target structure path and the complete behavior sequence, as shown in the following formula (6):

[0122] , (6)

[0123] in, represents a joint sequence of behavioral structures, Indicates the target structure path (such as [line number, section number, lot number]). Represents a complete behavior sequence. The symbol ‖ represents the concatenation operation of structural information and behavioral information.

[0124] In step S107 of some embodiments, in order to further support subsequent anomaly detection, consistency assessment and other tasks, the behavior structure joint sequence is constructed. Then, the behavior structure is combined with the sequence Input into a structure-aware embedding model to generate a low-dimensional sequence embedding vector .

[0125] The structure-aware embedding model is based on a bidirectional gated recurrent unit (Bi-GRU) and a structure position encoding module, which can capture the temporal context dependence of the behavior sequence and the hierarchical constraint of the structure path. Specifically, each structure element and behavior node in the behavior-structure joint sequence will be encoded into a fixed-dimensional vector. The structure part uses path hierarchical embedding (e.g., [line, section, subsection] is mapped to 3-level position information), and the behavior part is converted into a time-sensitive input vector according to the dictionary mapping and position encoding method. The final sequence embedding vector is generated by the following formula (9):

[0126] , (9)

[0127] wherein, represents the sequence embedding vector, represents the bidirectional GRU network based on the behavior and structure joint input, which is used to extract the context sequence representation. represents the path level encoder output of the structure path , which encodes the structure depth information, represents the weight of the structure embedding in the overall representation, and the empirical value is set to 0.2-0.5, which is adjusted according to the project complexity.

[0128] The sequence embedding vector fuses structure information and behavior evolution patterns, which can provide stable, low-dimensional, and semantically sufficient input representation for the subsequent structure consistency evaluation module, while being compatible with distance calculation and classifier operation of subsequent modules, ensuring the end-to-end reproducibility and business adaptability of the overall scheme.

[0129] In step S108 of some embodiments, in actual engineering scenarios, "behavior-structure inconsistency" usually manifests as the following situations:

[0130] The subsection belongs to "civil engineering", but the bill behavior appears "material warehousing" or "equipment maintenance" operation;

[0131] The node ordering in the behavior sequence violates the regular process of the structure, such as "payment" before "audit";

[0132] Some behaviors never appear in the structure path (for example, "application for payment" appears in the supervision unit bill).

[0133] Therefore, it is necessary to identify abnormal bills with inconsistent behavior processes and belonging structures through a structure-behavior consistency evaluation function. The core of the structure-behavior consistency evaluation function is to evaluate whether the current structure node “matching probability”. For this purpose, the mean of each node corresponding historical normal embedding and the deviation degree is modeled based on Euclidean distance. The structural behavior consistency evaluation function is shown in the following formula (10) and formula (11):

[0134] , (10)

[0135] , (11)

[0136] wherein, represents the structural behavior consistency score, represents the sigmoid activation function, represents the sequence embedding vector, represents the normal behavior embedding mean of the structural node , which is constructed by a large number of artificial review of historical “normal bills”. represents the position offset penalty term, which represents whether the key node in the behavior sequence occurs position abnormality (for example, “payment” before “signature”). If there is a behavior order violation in the process template of the structural node where , otherwise 0, the process template comes from the business sub-process definition in the structural graph, such as the process under “bid B3” should be: [contract signature → acceptance → invoicing → audit → payment]. represents the structural behavior inconsistency score, the lower the score, the more consistent the behavior with the structure, the higher the score, the higher the risk of inconsistency. When is greater than the score threshold 0.6, it is marked as abnormal. represents the weight coefficient of the position offset penalty term.

[0137] Further, the abnormal type label is divided into three categories, type 1 is “behavior misplacement”, type 2 is “structural overreach”, and type 3 is “path abnormality”. It is generated by the following logic:

[0138] If , it is “behavior misplacement”;

[0139] If is closer to other structural nodes , that is, misattribution behavior, it is “structural overreach”;

[0140] If the behavior appears abnormal repetition, jumping, etc., it is classified as “path abnormality”.

[0141] Please refer to Figure 4In some embodiments, step S109 further includes, but is not limited to, steps S401-S403:

[0142] In step S401, a collaborative suggestion score is obtained according to a preset collaborative suggestion scoring function, a sequence embedding vector, a structural behavior inconsistency score, an anomaly type label, and a preset role responsibility vector.

[0143] In step S402, if the collaborative suggestion score is greater than a preset suggestion threshold, a suggestion text is generated according to the structural behavior inconsistency score and the anomaly type label.

[0144] In step S403, the suggestion text is displayed in a ticket view corresponding to the role.

[0145] In step S401 of some embodiments, a role-anomaly matching perception mechanism is introduced to enable different business positions (such as financial audit, project management, contract approval, etc.) to form their own perception of ticket anomalies. Each role has its attention preference vector , which can be learned from historical processing records and represents the sensitive feature focus of the role on different types of anomalies. The collaborative suggestion scoring function is shown in the following formula (11):

[0146] , (11)

[0147] wherein represents the standardized ticket data , and the system determines whether it should be focused on and intervened by the role collaborative suggestion score, represents the sequence embedding vector, represents the attention preference vector of the role , and the dimension is the same as . represents the transpose of , and represents the structural behavior inconsistency score. represents an indicator function, indicating whether the current anomaly type label belongs to the anomaly set focused on by the role . , represents a weight parameter for controlling the weight influence of anomaly intensity and role responsibility domain in scoring.

[0148] In steps S402-S403 of some embodiments, the suggestion threshold can be set to 0.6. When , it is considered that the standardized ticket data should attract the attention of the role focus. Therefore, the suggestion text is generated according to the structural behavior inconsistency score and the abnormal type label, and the suggestion text is displayed in the ticket view of the corresponding role. For example, the standardized ticket data is determined as a "structural overreach" abnormality, ; , the attention preference vector corresponds to the "structural overreach" class ticket has high sensitivity;

[0149] calculation , the suggestion threshold is exceeded, the system includes the standardized ticket data in the "project manager view", and generates the following suggestion text:

[0150] "The current ticket contains cross-structure operation behavior, and the process trigger system is not within the scope of the structure. It is recommended to check the contract ownership and personnel permission configuration."

[0151] Further, a role view dictionary is constructed, as shown in the following formula (12):

[0152] , (2)

[0153] wherein represents the ticket coordination view set of the role , the format is structured, and contains a unique identifier, an abnormal type label, a structural behavior inconsistency score, and a suggestion text. represents a unique identifier, represents an abnormal type label, represents a structural behavior inconsistency score, represents the coordination suggestion score of the role which the system determines whether it should be focused on and intervened by the role . represents a suggestion text, which is obtained based on a template filling and scoring generation mechanism, represents a suggestion threshold.

[0154] The system automatically generates front-end interface content according to the role view dictionary, displays abnormal ticket ID, risk type, score, suggestion text and other multi-dimensional information. The suggestion content is generated by a template combination method, the template is preset by the business team, and is filled with abnormal type and structural path context information to form a highly controllable and deployable template.

[0155] Through the above steps S401 to S403, the ticket that has completed the structural behavior consistency score and abnormality detection is outputted and encapsulated, a coordination view is generated for different business roles to perceive and process, and a structured and landable preliminary suggestion text is given.

[0156] The steps S101 to S109 shown in the embodiments of the present application are as follows: obtaining standardized bill data, performing hash calculation on each standardized bill data to obtain a unique identifier. Obtain the behavior log and the preset structure graph, wherein the structure graph includes a graph root node and a plurality of graph nodes. Obtain the graph node corresponding to the standardized bill data to obtain a target graph node. The path from the graph root node to the target graph node is a target structure path. Obtain the complete behavior sequence of each standardized bill data according to the unique identifier and the behavior log. Construct a behavior structure joint sequence according to the target structure path and the complete behavior sequence. Input the behavior structure joint sequence into a preset structure perception embedding model to obtain a sequence embedding vector. Calculate according to the preset structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector to obtain a structure behavior inconsistency score and an abnormal type label. Generate and display a suggestion text according to the structure behavior inconsistency score and the abnormal type label. The sensitivity score and the suggestion generation of the bill abnormality enable different functional departments to accurately identify risk highlights and obtain personalized processing suggestions in their respective views, thereby realizing accurate identification of abnormal bills.

[0157] Please refer to Figure 5 The embodiments of the present application also provide a bill collaborative management system based on multi-source heterogeneous data fusion, which can realize the bill collaborative management method based on multi-source heterogeneous data fusion. The system comprises:

[0158] The first obtaining module 501 is configured to obtain standardized bill data.

[0159] The first calculation module 502 is configured to perform hash calculation on each standardized bill data to obtain a unique identifier.

[0160] The second obtaining module 503 is configured to obtain a behavior log and a preset structure graph. The structure graph includes a graph root node and a plurality of graph nodes.

[0161] The third obtaining module 504 is configured to obtain the graph node corresponding to the standardized bill data to obtain a target graph node. The path from the graph root node to the target graph node is a target structure path.

[0162] The fourth obtaining module 505 is configured to obtain the complete behavior sequence of each standardized bill data according to the unique identifier and the behavior log.

[0163] The construction module 506 is configured to construct a behavior structure joint sequence according to the target structure path and the complete behavior sequence.

[0164] The input module 507 is configured to input the behavior structure joint sequence into a preset structure perception embedding model to obtain a sequence embedding vector.

[0165] The second calculation module 508 is configured to perform calculation according to the preset structure behavior consistency evaluation function, the behavior structure joint sequence and the sequence embedding vector to obtain a structure behavior inconsistency score and an abnormal type label.

[0166] The display module 509 is configured to generate and display a suggestion text according to the structure behavior inconsistency score and the abnormal type label.

[0167] The specific implementation of the bill collaborative management system based on multi-source heterogeneous data fusion is basically the same as the specific implementation of the bill collaborative management method based on multi-source heterogeneous data fusion, and will not be repeated here.

[0168] Embodiments of the present application also provide an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the bill collaborative management method based on multi-source heterogeneous data fusion. The electronic device can be any intelligent terminal, such as a tablet computer or a vehicle-mounted computer.

[0169] Please refer to Figure 6 , Figure 6 The hardware structure of the electronic device of another embodiment is illustrated, which includes:

[0170] The processor 601 can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, and is used to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0171] The memory 602 can be implemented in the form of a ROM (Read Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory). The memory 602 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 602 and called and executed by the processor 601 to implement the bill collaborative management method based on multi-source heterogeneous data fusion.

[0172] The input / output interface 603 is used to realize information input and output.

[0173] The communication interface 604 is configured to realize the communication interaction between the device and other devices, and can realize the communication through a wired manner (for example, a USB, a network cable, or the like) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, or the like).

[0174] The bus 605 is configured to transmit information between various components (for example, the processor 601, the memory 602, the input / output interface 603, and the communication interface 604) of the device.

[0175] The processor 601, the memory 602, the input / output interface 603, and the communication interface 604 are connected to each other through the bus 605 to realize the communication connection between the device.

[0176] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to realize the above-mentioned bill collaborative management method based on multi-source heterogeneous data fusion.

[0177] The memory is a non-transitory computer readable storage medium, and can be used to store a non-transitory software program and a non-transitory computer executable program. In addition, the memory can include a high-speed random access memory, and can further include a non-transitory memory, for example, at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and the remote memory can be connected to the processor through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0178] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0179] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than the figures, or combine certain steps, or different steps.

[0180] The system embodiments described above are only schematic, and the units described as separate components can be or can not be physically separated, that is, can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the modules can be selected to realize the purpose of the embodiments.

[0181] Those skilled in the art can understand that all or some of the steps in the method disclosed above, the function modules / units in the system and the device can be implemented as software, firmware, hardware or appropriate combination thereof.

[0182] The terms "first", "second", "third", "fourth" and the like in the description of the application and in the claims, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of these terms herein is to be construed to cover a general order and / or structure unless otherwise indicated. Furthermore, the terms "comprise", "comprising", "has", "having", "includes", "including", "contain", "containing" or any other similar forms are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, has, includes, contains items or components does not include items or components not explicitly recited. The terms "a" or "an", as used herein in the detailed description and in the claims, mean "one or more" or "at least one", unless otherwise indicated.

[0183] It should be understood that, in the present application, "at least one" means one or more, and "multiple" means two or more. "And / or" is used to describe the relationship between associated objects, which means that there can be three relationships, for example, "A and / or B" can mean that there are three cases: only A, only B, and A and B at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can mean a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be singular or plural.

[0184] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the above-described system embodiments are only illustrative, for example, the division of the above-mentioned units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed objects can be indirect coupling or communication connection between the systems or units through some interfaces, and can be electrical, mechanical or other forms.

[0185] The units described as separate components above can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0186] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0187] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0188] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, and are not limited to the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. A bill collaborative management method based on multi-source heterogeneous data fusion, characterized by: The method comprises: Obtain standardized bill data; Perform hash calculation based on each of the standardized bill data to obtain a unique identifier , , Indicates the invoice number, Indicates the amount field, Indicates the contract number, Indicates a unified timestamp format; Obtain behavior logs and preset structure maps; behavior logs come from the contract system, invoice verification platform, and approval platform. , each is a triple , the record unique identifier is The bill at the time The behavior that occurred , The unique identifier of the ticket corresponding to a certain behavior event in the behavior log record; the structure graph includes a graph root node and multiple graph nodes, each node represents a structured engineering unit, and contains the following fields: : The contract number associated with the node, : the average amount of historically attributed bills, : The engineering level path where the node is located, including line>section>section>unit, : plot number or stake number, used for structure adjacency calculation, : Invoice time distribution density function of historical bills; Obtain the graph node corresponding to the standardized bill data to obtain a target graph node; wherein the path from the graph root node to the target graph node is a target structure path, wherein the target structure path includes [line number, section number, bid section number]; According to the unique identifier and the behavior log, a complete behavior sequence of each standardized bill data is obtained, wherein the behavior log is divided into the unique identifier Log Aggregate to obtain the complete behavior sequence of each standardized bill data, representing the standardized bill data The complete sequence of actions performed in all business systems; Constructing a behavior structure joint sequence according to the target structure path and the complete behavior sequence; Inputting the behavior-structure joint sequence into a preset structure-aware embedding model to obtain a sequence embedding vector, wherein the structure-aware embedding model is based on a bidirectional gated recurrent network and a structure position encoding module. Each structural element and behavior node in the behavior-structure joint sequence is encoded as a vector of fixed dimension. The structure part adopts a path hierarchical embedding method, and the behavior part is converted into a time-sensitive input vector based on dictionary mapping and position encoding. Calculate based on a preset structure-behavior consistency evaluation function, the behavior-structure joint sequence, and the sequence embedding vector to obtain a structure-behavior inconsistency score and anomaly type label; Suggested text is generated and displayed according to the structure-behavior inconsistency score and the anomaly type label.

2. The method according to claim 1, characterized in that Generating and displaying suggested text according to the structure-behavior inconsistency score and the abnormality type label includes: Scoring is performed according to a preset collaborative suggestion scoring function, the sequence embedding vector, the structure-behavior inconsistency score, the anomaly type label, and a preset role responsibility vector to obtain a collaborative suggestion score; If the collaborative suggestion score is greater than a preset suggestion threshold, generating the suggestion text according to the structure-behavior inconsistency score and the anomaly type label; The suggested text is displayed in the bill view of the corresponding role.

3. The method according to claim 1, characterized in that Scoring according to a preset collaborative suggestion scoring function, the sequence embedding vector, the structure-behavior inconsistency score, the anomaly type label, and the preset role responsibility vector to obtain a collaborative suggestion score includes: The collaborative suggestion scoring function is: ； in, Indicates standardized bill data , the system determines whether it should be Collaborative recommendation scores for focused attention and interventional treatment, represents the sequence embedding vector, Representing a role The attention preference vector, express The transpose of Indicates the structural behavior inconsistency score, represents the indicator function, Indicates the exception type label , Representing a role The set of exceptions of interest, 、 Represents the weight parameter.

4. The method according to claim 1, wherein The obtaining of standardized bill data includes: Obtain original bill data from different systems; Performing unified field processing on the original bill data to obtain initial bill data; The initial bill data is searched and the initial bill data that meets the preset conditions is merged to obtain the standardized bill data.

5. The method according to claim 1, wherein The step of obtaining the graph node corresponding to the standardized bill data to obtain a target graph node includes: Calculating the score of mapping the standardized bill data to each of the graph nodes according to a preset anchor scoring function to obtain an anchor score set; Obtaining the highest anchor score in the anchor score set to obtain a target anchor score; Obtain the graph node corresponding to the target anchor score to obtain the target graph node.

6. The method according to claim 5, characterized in that The step of calculating the score of mapping the standardized bill data to each of the graph nodes according to a preset anchor scoring function to obtain an anchor score set includes: The anchor scoring function is: ； ； ； in, Represents standardized bill data Mapping to graph nodes The anchor score, Represents standardized bill data Amounts in the middle and graph nodes Proportional similarity of historical amounts, Represents standardized bill data The invoicing time is in the graph node The probability value under the time density function is generated by kernel density estimation, represents the structural level penalty regularization term, Indicates the missing length of the longest common subpath between the bill contract path and the node path, represents the structural penalty coefficient, Represents standardized bill data The amount field, Represents a graph node The average amount of historical vested notes, Represents standardized bill data The contract number, Represents a graph node The project level path, 、 、 They respectively represent the weight coefficients of amount similarity, time similarity and structural level penalty items in the overall anchor score.

7. The method according to claim 1, characterized in that The calculation is performed based on a preset structure-behavior consistency evaluation function, the behavior-structure joint sequence, and the sequence embedding vector to obtain a structure-behavior inconsistency score and an anomaly type label, including: The structural behavior consistency evaluation function is: ； ； in, represents the structural behavior consistency score, represents the sigmoid activation function, represents the sequence embedding vector, Represents a structure node The normal behavior embedding mean, Represents the position offset penalty term, indicating whether the key nodes in the behavior sequence have abnormal positions. Indicates the structural behavior inconsistency score, Represents the weight coefficient of the position offset penalty term.

8. A bill collaborative management system based on multi-source heterogeneous data fusion, characterized by: The system comprises: A first acquisition module is used to acquire standardized bill data; The first calculation module is used to perform hash calculation based on each of the standardized bill data to obtain a unique identifier , , Indicates the invoice number, Indicates the amount field, Indicates the contract number, Indicates a unified timestamp format; The second acquisition module is used to obtain behavior logs and preset structure maps; among them, behavior logs come from the contract system, invoice verification platform, and approval platform. , each is a triple , the record unique identifier is The bill at the time The behavior that occurred , The unique identifier of the ticket corresponding to a certain behavior event in the behavior log record; the structure graph includes a graph root node and multiple graph nodes, each node represents a structured engineering unit, and contains the following fields: : The contract number associated with the node, : the average amount of historically attributed bills, : The engineering level path where the node is located, including line>section>section>unit, : plot number or stake number, used for structure adjacency calculation, : Invoice time distribution density function of historical bills; A third acquisition module is configured to acquire the graph node corresponding to the standardized bill data to obtain a target graph node; wherein the path from the graph root node to the target graph node is a target structure path, wherein the target structure path includes [line number, section number, bid section number]; The fourth acquisition module is used to obtain the complete behavior sequence of each standardized bill data according to the unique identifier and the behavior log, wherein the behavior log is divided into the following categories: Log Aggregate to obtain the complete behavior sequence of each standardized bill data, representing the standardized bill data The complete sequence of actions performed in all business systems; A construction module, configured to construct a behavior-structure joint sequence according to the target structure path and the complete behavior sequence; An input module is configured to input the behavior-structure joint sequence into a preset structure-aware embedding model to obtain a sequence embedding vector. The structure-aware embedding model is based on a bidirectional gated recurrent network and a structure position encoding module. Each structural element and behavior node in the behavior-structure joint sequence is encoded as a vector of fixed dimension. The structure part adopts a path-layered embedding method, and the behavior part is converted into a time-sensitive input vector based on a dictionary mapping and position encoding method. A second calculation module is used to calculate according to a preset structure-behavior consistency evaluation function, the behavior-structure joint sequence and the sequence embedding vector to obtain a structure-behavior inconsistency score and an anomaly type label; A display module is used to generate and display suggested text based on the structural behavior inconsistency score and the abnormality type label.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the bill collaborative management method based on multi-source heterogeneous data fusion as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for collaborative bill management based on multi-source heterogeneous data fusion according to any one of claims 1 to 7 is implemented.

Images