A case assignment method and system based on data fusion

Abstract

The application discloses a case distribution method and system based on data fusion, and the method comprises the following steps: acquiring multi-source case data from multiple heterogeneous business systems, wherein the multi-source case data comprises structured data and unstructured data; performing fusion processing on the multi-source case data, and generating a structured case feature vector comprising multiple dimension quantification features for each case; generating an ability feature vector for a currently available handling subject based on a preset handling subject ability label library, wherein the ability feature vector comprises regional coverage information and efficiency indicators; and determining a distribution scheme for distributing each case to a corresponding handling subject according to a preset rule based on the structured case feature vector and the ability feature vector. It can be seen that the application improves the automatic task distribution performance of the computer system in a multi-source heterogeneous data environment from the three dimensions of data fusion efficiency, matching calculation efficiency and system load balancing capability.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a case assignment method and system based on data fusion. Background Technology

[0002] In large-scale business processing systems involving multi-stage and multi-role collaboration, such as integrated service platforms encompassing mediation, arbitration, and notarization, case source or task data is typically stored in heterogeneous business subsystems built at different times and developed by different vendors. These subsystems differ significantly in data format, storage medium, and interface protocols, resulting in inherent physical or logical data isolation.

[0003] Existing task assignment schemes have the following technical shortcomings when dealing with such multi-source heterogeneous data: First, there is insufficient data utilization capability. The system can typically only read structured fields (such as task number and processing status) from each subsystem. However, for massive amounts of unstructured data (such as scanned documents with signatures and audio / video files of communication records), the lack of effective automated parsing and fusion methods prevents the conversion of the key information contained within into data that can be directly used by subsequent decision-making processes. This results in a severe lack of input information for task profiling, with single and sparse feature dimensions.

[0004] Secondly, the matching computation is inefficient and inaccurate. Due to the lack of a unified data fusion view, existing assignment rules are mostly based on static, pre-set simple conditions (such as round-robin by region code or average allocation by quantity) for serial matching. When faced with a large number of tasks to be assigned and a large number of candidate processing units, the computational complexity of this serial matching method increases sharply, and it cannot quantify the correlation between task characteristics and processing unit capabilities, resulting in a mismatch between the matching results and the actual needs, directly reducing the processing efficiency and resource utilization of the entire system.

[0005] Third, the system lacks dynamic perception and adaptive capabilities. The existing solution cannot perceive the load status and processing efficiency changes of each processing unit in real time. Once the allocation decision is made, it cannot be dynamically adjusted, which can easily lead to some units being overloaded while others are idle, causing fluctuations in the overall processing capacity of the system and waste of resources.

[0006] Therefore, how to break down the barriers between multi-source heterogeneous data, realize the automated parsing and fusion of unstructured data, and on this basis build an adaptive assignment mechanism that can quantitatively match task characteristics and processing unit capabilities and has dynamic load awareness is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide a case assignment method and system based on data fusion, which can automatically parse and fuse multi-source structured and unstructured data from multiple heterogeneous business systems to form a structured case feature vector containing multiple dimensions of quantitative features. On this basis, the system can realize quantitative matching and adaptive assignment of task features and processing unit capabilities, thereby improving the automated task processing performance of computer systems in multi-source heterogeneous data environments.

[0008] To address the aforementioned technical problems, the first aspect of this invention discloses a case allocation method based on data fusion, the method comprising: Acquire multi-source case data from multiple heterogeneous business systems, including structured data and unstructured data; The multi-source case data is fused to generate a structured case feature vector for each case, which contains quantitative features of multiple dimensions. Based on a pre-set capability tag library for disposal entities, capability feature vectors are generated for currently available disposal entities. The capability feature vectors include geographic coverage information and efficiency indicators. Based on the structured case feature vector and the capability feature vector, an allocation scheme is determined according to preset rules to assign each case to the corresponding handling entity.

[0009] As an optional implementation, in the first aspect of the present invention, the step of fusing the multi-source case data to generate a structured case feature vector containing multiple dimensional quantitative features for each case includes: A text recognition module is used to extract first text information from the first unstructured data, which includes images of legal documents, and the first text information includes the case number, parties involved, or amount of the claim. The semantic analysis module is used to extract the second text information from the second unstructured data, which includes the transcribed text of the call recording. The second text information includes the parties involved, the amount in dispute, the cause of action, the legal status, or preset keyword information. Based on the extracted first and second text information, the subjective repayment dimension features, asset enforceability dimension features, and legal complexity dimension features of the case are constructed and calculated to generate the structured case feature vector.

[0010] As an optional implementation, in the first aspect of the present invention, generating capability feature vectors for currently available disposal subjects based on a preset disposal subject capability tag library includes: Based on the historical disposal records and publicly available data of the disposal entity, its capability tags are constructed and dynamically updated. These capability tags include: Attribute tags include the professional field category and geographical scope of practice or service of the entity handling the matter; Performance metrics include efficiency and quality indicators derived from historical data; and, Status tags are used to characterize the real-time availability status of the handling entity, including the current case load, the number of pending enforcement cases, or the available resource capacity; Based at least on the attribute tags, performance tags, and status tags, generate the capability feature vector of the currently available disposal subject.

[0011] As an optional implementation, in the first aspect of the present invention, the step of determining the assignment scheme for each case to the corresponding handling entity based on the structured case feature vector and the capability feature vector according to preset rules includes: Based on the structured case feature vector and the capability feature vector, a preliminary assignment scheme is calculated using a multi-objective optimization model; The preliminary allocation scheme is compared and verified with the mandatory business rules. If there is a conflict, it is corrected according to the mandatory business rules to output the final allocation scheme.

[0012] As an optional implementation, in the first aspect of the present invention, the multi-objective optimization model aims to maximize the overall disposal benefit, and the optimization objective function is defined as: ; Where C is the set of cases to be assigned, and P is the set of available entities for disposal; Assigning variables to represent cases Whether to assign to the main body , For the case By the main body The expected value of the proceeds from the disposal; This is the estimated disposal period; The resource load deviation of the main body after allocation is defined as the absolute difference between the current case volume and the ideal load; , , Weights can be configured for the corresponding business functions; The objective function is globally optimized in a single batch assignment to avoid local optima.

[0013] As an optional implementation, in the first aspect of the invention, the calculation of the preliminary allocation scheme through a multi-objective optimization model includes: Obtain jurisdictional information and legal complexity dimension characteristics of cases to be assigned; Obtain information on the business licenses or service coverage of each disposal entity, historical records of conflicts of interest, professional qualifications, historical winning records, and the preset upper limit of remaining processing capacity; If the jurisdiction of the case to be assigned is not within the scope of the entity's professional license or service coverage, the assignment plan will be deemed not to meet the geographical compliance constraints, and the subsequent calculation of the assignment plan will be terminated. If the disposing entity has a history of conflict of interest with the debtor in the case, the allocation plan will be deemed not to meet the constraints of the avoidance rule, and the subsequent calculation of the allocation plan will be terminated. If the legal complexity dimension of a case is determined to be high, and the handling entity does not possess the corresponding professional qualifications and its historical winning record is below a preset threshold, then the allocation plan is determined to not meet the qualification threshold constraints, and the subsequent calculation of the allocation plan is terminated. If the total number of pending cases handled by the entity exceeds the preset remaining processing capacity, the pairing is determined to not meet the capacity limit constraint, and the subsequent calculation of the allocation scheme is terminated.

[0014] As an optional implementation, in the first aspect of the invention, the calculation of the preliminary allocation scheme through a multi-objective optimization model includes: According to the formula: ; For each case to be assigned and the candidate handling entity, a comprehensive score is calculated, and the entity with the highest score is selected for assignment.

[0015] As an optional implementation, in the first aspect of the invention, the set of cases to be assigned and the set of subjects to be dealt with are considered as two parts of a bipartite graph, with edge weights set to... or The Hungarian algorithm is used to find the globally optimal one-to-one matching.

[0016] A second aspect of this invention discloses a case assignment system based on data fusion, the system comprising: The acquisition module is used to acquire multi-source case data from multiple heterogeneous business systems, including structured data and unstructured data. The fusion module is used to fuse the multi-source case data and generate a structured case feature vector containing multiple dimensions of quantitative features for each case. The generation module is used to generate capability feature vectors for currently available disposal entities based on a preset disposal entity capability tag library. The capability feature vectors include geographical coverage information and efficiency indicators. The determination module is used to determine, based on the structured case feature vector and the capability feature vector, an allocation scheme for assigning each case to the corresponding handling entity according to preset rules.

[0017] A third aspect of the present invention discloses a computer device, the computer device comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the data fusion-based case assignment method disclosed in the first aspect of the present invention.

[0018] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the case assignment method based on data fusion disclosed in the first aspect of the present invention.

[0019] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: This invention acquires and fuses multi-source structured and unstructured data from multiple heterogeneous business systems, transforming the original fragmented data into structured case feature vectors containing multi-dimensional quantitative features. Simultaneously, based on a pre-defined capability tag library for handling entities, it generates capability feature vectors for currently available handling entities, including geographical coverage information and efficiency indicators. Then, based on these two types of feature vectors and pre-defined rules, it determines the allocation scheme. This achieves automated parsing and unified representation of multi-source heterogeneous data, transforming previously scattered, inconsistently formatted, and machine-unusable unstructured data into calculable and comparable numerical coordinates, breaking down data barriers between heterogeneous systems. By quantifying both case features and entity capabilities into vector form, it lays the data foundation for subsequent efficient matrix operations for matching, avoiding the high computational complexity of traditional text-based conditional matching. Furthermore, by introducing capability feature vectors containing real-time status information, the allocation decision-making process gains the ability to dynamically perceive the load status of processing units, enabling adaptive balancing of system resources during batch allocation. Ultimately, this invention improves the overall performance of automated task assignment in a multi-source heterogeneous data environment by enhancing data fusion efficiency, matching calculation efficiency, and system load balancing capabilities. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1This is a flowchart illustrating a case assignment method based on data fusion disclosed in the first embodiment of the present invention; Figure 2 This is a detailed flowchart of S12 in the first embodiment of the present invention; Figure 3 This is a detailed flowchart of S13 in the first embodiment of the present invention; Figure 4 This is a detailed flowchart of S14 in the first embodiment of the present invention; Figure 5 This is a detailed flowchart of S142 in the first embodiment of the present invention; Figure 6 This is a structural block diagram of a case assignment system based on data fusion disclosed in an embodiment of the present invention.

[0022] Figure 7 This is a hardware structure diagram of a computer device for a case assignment system based on data fusion disclosed in an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] This invention discloses a case assignment method and system based on data fusion. The invention acquires and fuses multi-source structured and unstructured data from multiple heterogeneous business systems, transforming the original fragmented data into structured case feature vectors containing multi-dimensional quantified features. Simultaneously, based on a pre-defined capability tag library for handling entities, it generates capability feature vectors for currently available handling entities, including geographical coverage information and efficiency indicators. Then, based on these two types of feature vectors, an assignment scheme is determined according to pre-defined rules. This achieves automated parsing and unified representation of multi-source heterogeneous data, transforming previously scattered, inconsistently formatted, and machine-unusable unstructured data into calculable and comparable numerical coordinates, breaking down data barriers between heterogeneous systems. By quantifying both case features and entity capabilities into vector form, it lays the data foundation for subsequent efficient matrix operations for matching, avoiding the high computational complexity of traditional text-based conditional matching. Furthermore, by introducing capability feature vectors containing real-time status information, the assignment decision-making process gains the ability to dynamically perceive the load status of processing units, enabling adaptive resource balancing during batch assignment. Ultimately, this invention improves the overall performance of automated task assignment in a multi-source heterogeneous data environment by enhancing data fusion efficiency, matching calculation efficiency, and system load balancing capabilities.

[0027] Example 1 Please refer to Figures 1 to 5 As shown, this embodiment of the invention discloses a case assignment method based on data fusion, including S11-S14, wherein: S11, acquire multi-source case data from multiple heterogeneous business systems, the multi-source case data including structured data and unstructured data.

[0028] The executing entity in this embodiment can be an intelligent allocation engine deployed on an enterprise-level data platform. This step aims to solve the data input problem for case allocation decisions. It is exemplary and not limiting. Taking a post-loan management scenario as an example, raw data can be pulled from a first business system (such as a collection / debt collection management platform), a second business system (such as a court electronic litigation platform), and a third business system (such as a court enforcement case handling platform).

[0029] Specifically, taking post-loan management scenarios as an example, the first business system (collection system) typically refers to the post-loan collection / debt collection management platform, which stores debtors' repayment records, call activity, or commitment letters, etc.; the second business system (litigation platform) refers to the court's electronic litigation platform, which stores legal documents such as complaints, judgments, or mediation agreements; the third business system (enforcement system) refers to the court's enforcement case handling platform, which stores enforcement clue data such as property investigation and control receipts, freezing orders, or final judgments. Structured data refers to standardized records stored in the form of two-dimensional tables, such as case status tables, repayment records, or judge quota tables; unstructured data refers to text, image, or audio / video files without a predefined data model or not organized in rows and columns, such as call recordings, scanned PDFs of complaints, or photos of vehicle registration certificates, etc.

[0030] It should be noted that the above description using a post-loan management scenario as an example is only for the purpose of understanding the technical solution of this invention, and is not intended to limit the application field of this invention. The technical solution of this invention is based on the fusion processing and vectorized matching of multi-source heterogeneous data. Its core lies in solving the two general technical problems of "multi-source data fusion" and "task-resource quantitative matching," and therefore it can be widely applied to various business areas that require cross-system data integration and automated task allocation, including but not limited to: 1. Public affairs and government services In the "One-Stop Government Service" platform, applications submitted by enterprises or individuals may involve data from multiple independent government systems, such as market supervision, taxation, social security, and public security. The technical solution of this invention can be used to: automatically acquire and integrate application materials (including scanned documents, forms, and other structured and unstructured data) from various government systems to construct a unified "service task feature vector"; simultaneously, generate "processing unit capability vectors" for each government service window and approval department, including business scope, real-time load, and processing efficiency; thereby achieving intelligent assignment and dynamic load balancing of service tasks, improving the efficiency of government services and resource utilization.

[0031] 2. Medical and Health Services and Hierarchical Medical System In regional medical collaboration platforms, patient medical information is scattered across the HIS, imaging, and laboratory systems of different medical institutions. The technical solution of this invention can be used to: integrate electronic medical records, imaging reports, and laboratory data from multiple medical institutions to form a "patient medical feature vector"; simultaneously generate a "medical resource capability vector" for hospitals, departments, and doctors at all levels, including their professional fields, current bed capacity, and surgical schedules; thereby enabling the referral and allocation of complex cases and the optimized allocation of medical resources, promoting the implementation of hierarchical medical services.

[0032] 3. Emergency Command and Resource Dispatch In urban emergency management systems, emergency information may originate from multiple heterogeneous data sources, such as 110 / 119 emergency call platforms, IoT sensors, video surveillance, and social media. The technical solution of this invention can be used to: fused multi-source emergency data in real time to construct an "event feature vector" (such as event type, severity, and geographical location); simultaneously, to generate an "emergency resource capability vector" for each emergency team, rescue vehicle, and material reserve point, containing response range, current workload, and equipment status; thereby achieving optimal matching between emergencies and emergency resources, improving emergency response speed and resource allocation efficiency.

[0033] 4. Areas of scientific research collaboration and project allocation In scientific research project management platforms, researchers' achievements, project data, and experimental records are scattered across multiple heterogeneous systems. The technical solution of this invention can be used to: integrate scientific research data from different systems to construct a "scientific research task feature vector"; simultaneously generate a "scientific research resource capability vector" for each research team and laboratory, containing research directions, current project workload, and historical achievements; thereby achieving intelligent allocation of scientific research topics and cross-team collaboration, and optimizing the allocation of scientific research resources.

[0034] The above application scenarios are merely to further illustrate the versatility of the present invention, and not to limit the scope of protection. Regardless of the specific business field in which it is applied, as long as it involves the fusion processing of multi-source heterogeneous data and the quantitative matching of tasks and resources, it falls within the protection scope of the present invention.

[0035] It should be noted that, regarding the compliance of judicial data sources, this step mandates the deployment of a compliance filtering gateway at the data access layer: only case data with explicit authorization from the debtor (such as a "Post-Loan Management and Data Sharing Authorization Letter" signed in the loan agreement) will be accessed, including but not limited to call recordings and related debt data authorized by the debtor; for data from the court's electronic case file platform, it will only be obtained through official cooperation interfaces (such as the court data interface service platform) via an HTTPS encrypted channel, and will strictly comply with the requirements of the "People's Court Information System Security Technical Specifications" regarding minimizing data output, without actively crawling, batch caching, or retaining original case file images.

[0036] S12, perform fusion processing on the multi-source case data to generate a structured case feature vector containing multiple dimensions of quantitative features for each case.

[0037] The fusion process in this step is not a simple data splicing, but rather involves a series of operations such as data cleaning, entity alignment, and feature engineering. For example, fragmented raw data from three systems is mapped to a unified view centered on "case ID," and quantifiable and comparable numerical features are extracted from it. A structured case feature vector is a coordinate point in a multi-dimensional mathematical space, for example... , where each dimension represents the score or status of a case on a certain quantitative dimension (such as the willingness to cooperate dimension 0.75, the asset disposability index 0.60), serving as the direct / indirect input data for the subsequent matching algorithm.

[0038] Please refer to Figure 2 As shown, this step S12 includes S121 - S123, where: S121, use the text recognition module to extract the first text information from the first unstructured data. The first unstructured data includes image-based legal documents, and the first text information includes case number, parties, or litigation claim amount.

[0039] In this embodiment, the text recognition module can be an OCR (Optical Character Recognition) engine. OCR is a technology that converts text in images into machine-editable text. In this embodiment, the first unstructured data can be a scanned or photographed legal document image, such as a complaint without an electronic signature, a mediation agreement signed offline, a scanned copy of a judgment retrieved from the archive. The common formats of image-based legal documents are PDF, JPG, or PNG, and their text is stored in pixel form and cannot be directly retrieved by the database. This step preferably supports a highly robust OCR engine for complex layouts (such as forms, tables, stamped text) and blurred images (such as faxed documents, old dot matrix print traces), such as Qwen-VL or PaddleOCR. The first text information is the key-value pair of the keyword fields output by OCR, for example, {"case number": "(2025) Yue 01 Min Zhong 1234 Hao", "parties": "Zhang San", "litigation claim amount": "150,000.00 yuan"}.

[0040] S122, use the semantic analysis module to extract the second text information from the second unstructured data. The second unstructured data includes the transcribed text of the phone call recording, and the second text information includes parties, subject amount, cause of action, legal status, or preset keyword information.

[0041] The semantic analysis module can be an NLP (Natural Language Processing) model, which in this embodiment can refer to a pre-trained language model in the legal field, such as Legal-BERT or Lawformer. The second type of unstructured data primarily originates from call recordings in the debt collection system (these recordings require reminders and consent from the debtor before recording and use). These recordings are first transcribed into text by an ASR (Automatic Speech Recognition) engine. Pre-set keyword information refers to words or phrases pre-configured by business personnel that represent the debtor's specific intentions, such as "definitely paying next week," "selling the house," and "don't sue me." These keywords are then embedded and converted into factors for calculating the intention score.

[0042] S123, Based on the extracted first and second text information, construct and calculate the subjective repayment dimension features, asset enforceability dimension features, and legal complexity dimension features of the case, and generate the structured case feature vector.

[0043] The subjective willingness (W) dimension of repayment is not a direct observation but a latent variable synthesized from multiple factors. For example, if the NLP model identifies high-frequency positive commitment words (such as "repay" or "raise money") from the transcript of a phone call, and the debtor's past three commitments have a 100% fulfillment rate, then W converges towards 1; conversely, if it identifies adversarial words (such as "no money" or "sue at will"), and the debtor frequently loses contact, then W converges towards 0. The asset enforceability dimension (i.e., the enforceability score E) integrates asset clue density (number of vehicle and real estate registrations) and concealment behavior index (recent change of phone number, related parties holding properties on behalf of others). The legal complexity dimension (i.e., complexity L) is a binary variable, determined by the rule engine: if the case meets any of the following conditions—"number of defendants > 1," "previously raised an objection to jurisdiction," or "multiple layers of security"—then L = 1; otherwise, L = 0. Finally, these features are packaged into a triple vector (W, E, L) or a higher-dimensional extended vector.

[0044] It is important to note that in steps S121-S123 above, all text involving Personally Identifiable Information (PII) and financial account data must undergo mandatory de-identification before entering the OCR or NLP engine. This embodiment incorporates a dynamic de-identification middleware at the data access layer, employing Format Preservation Encryption (FPE) technology to irreversibly obfuscate ID card numbers, bank card numbers, and mobile phone numbers (e.g., retaining the first 6 and last 4 digits of the ID card number, with the middle 8 digits marked as "*"). For scenarios requiring auditing after de-identification, a "three-person separation" access control system (system administrator, security auditor, and business operator) is implemented, and each plaintext viewing requires electronic approval and record-keeping. The original plaintext data is never written to disk, stored in a database, or cached, thus avoiding the risk of "illegally processing personal information" as stipulated in Article 51 of the Personal Information Protection Law from an architectural perspective.

[0045] The aforementioned steps S121-S123, through an end-to-end automated heterogeneous data parsing pipeline, transform unstructured fragments such as images, audio, and text generated during the collection, litigation, and enforcement stages into quantifiable and computable case profile feature vectors. Compared to the traditional model that requires manual reading of case files and manual entry of key information, this step achieves a qualitative leap from human-reading to machine-reading, reducing feature extraction time from days to seconds, and completely avoiding operational risks such as typos and missing amounts caused by manual entry. Simultaneously, by integrating the entity recognition of legal documents with the emotional perception of collection calls into the same case view, it bridges the information gap between debtor communication records and business document records for the first time, providing deep behavioral and legal element features that traditional purely structured databases cannot generate for subsequent allocation decisions.

[0046] S13. Based on a preset capability tag library for disposal entities, generate capability feature vectors for currently available disposal entities. The capability feature vectors include geographical coverage information and efficiency indicators.

[0047] This step, along with S12, creates a case profile on one end and a profile of the handling entity on the other. The handling entity includes all legal or non-legal entities that may handle the case, such as mediation institutions, law firms, notary offices, or court enforcement teams, but excludes personnel within judicial organs. Preferably, in this embodiment, the capability tag library is not a static directory, but a real-time updated key-value storage system (such as Redis or HBase), dynamically refreshed through a combination of offline batch processing (daily) and real-time stream computing (hourly).

[0048] Please refer to Figure 3 As shown, this step S13 includes S131-S132, wherein: S131, Based on the historical disposal records and public data of the disposal entity, construct and dynamically update its capability tags.

[0049] Specifically, the capability tags include: attribute tags, including the professional field category and geographical scope of practice or service of the subject of the disposal.

[0050] Historical case handling records primarily originate from the case closure feedback interface connected to this system; publicly available data is obtained through legitimate web scraping or official interfaces, such as information on lawyer practice licenses published by the Department of Justice, law firm annual inspection reports, and lists of court-appointed mediation organizations. Attribute tags are relatively static basic profiles, such as "professional field": contract disputes, debt disputes, intellectual property disputes; "geographical scope": specifically, it can be a prefecture-level administrative region (e.g., the case-handling agency is the Futian District People's Court of Shenzhen) or a nationwide online service.

[0051] The capability tags also include performance tags, including efficiency and quality indicators obtained from historical data statistics.

[0052] Performance tags are dynamically calculated numerical indicators, with the rolling window typically set to the past 6 or 12 months. Efficiency indicators include, but are not limited to, the average case closure period T. close (Average number of calendar days from case acceptance to filing); First contact time (median time for the first contact with the debtor after case acceptance); Enforcement and control feedback time (average interval from submitting a seizure application to receiving a receipt). Quality indicators include: Success rate R. win (Percentage of effective judgments supporting the plaintiff's claims); Actual success rate of disposal R recovery (Amount recovered / Amount supported by judgment); Mediation success rate (Number of cases reaching a mediation agreement / Total number of cases accepted for mediation). These indicators are converted into quantifiable values ​​in the 0-1 range after min-max normalization or Z-score standardization.

[0053] The capability tag also includes a status tag, which is used to characterize the real-time availability status of the disposal entity, including the current case load, the number of pending execution cases, or the available resource capacity.

[0054] Status tags have the highest timeliness requirements and must be updated hourly. Current case load refers to the total number of cases received but not yet archived by the law firm / mediation organization; the number of pending enforcement cases specifically refers to the number of enforcement cases under the judge's name that are in the "filed but not yet closed" status at the enforcement bureau level; available resource capacity is a relatively static staffing number used to calculate the average workload per judge. This step uses an asynchronous message queue (such as RabbitMQ) to listen for case archiving events pushed by various business systems. Once the case status changes to "closed" or "termination of this enforcement procedure," the load count of the corresponding entity is immediately decremented.

[0055] As an optional rather than limiting option, this embodiment can introduce a load balancing weighting coefficient γ, allowing business administrators to set differentiated ideal load thresholds for entities in different jurisdictions based on the varying pressure levels of courts in different regions. This coefficient is used in the load deviation term D of the subsequent allocation algorithm. j It plays a regulatory role.

[0056] S132, at least based on the attribute tags, performance tags and status tags, generate the capability feature vector of the currently available disposal subject.

[0057] Steps S131-S132 above construct a three-layer dynamic tagging system of attributes, performance, and status. This transforms the information of the handling entities, scattered across publicly available judicial data, historical case closure reports, and real-time case handling processes, into a machine-readable, comparable, and aggregateable structured capability feature vector. In traditional case allocation, business personnel's understanding of law firm capabilities is often lagging and one-sided (e.g., "only knowing that a certain firm is a large firm, but unaware of its recent change in win rate"). This step, through a rolling update mechanism for performance tags, enables the entity's capability profile to be automatically updated; the real-time access to status tags allows the allocation system to have load awareness for the first time, avoiding continuous case assignment to overloaded handling entities and fundamentally solving the resource mismatch problem caused by the huge difference in case volume between different law firms.

[0058] S14. Based on the structured case feature vector and the capability feature vector, determine the assignment scheme for each case to the corresponding handling entity according to preset rules.

[0059] The preset rules for this step can be preset formulas or a configurable, optimizable rule engine instance (such as Drools or EasyRules). The assignment scheme is a set of many-to-many mapping relationships, which can be mathematically represented as an assignment matrix, where each element indicates whether a case is assigned to a specific subject, and each case is assigned to only one subject.

[0060] Please refer to Figure 4 As shown, this step S14 includes S141-S142, wherein: S141, Based on the structured case feature vector and the capability feature vector, a preliminary assignment scheme is calculated using a multi-objective optimization model.

[0061] This step formalizes the allocation decision as a multi-objective integer programming problem. The multi-objective optimization model aims to maximize the overall disposal benefit, and the optimization objective function is defined as follows: ; in: C represents the set of cases to be assigned, and P represents the set of available entities for handling such cases. (Expected recovery value): This is the expected amount to be recovered when case i is handled by entity j. It can be predicted through the structured case feature vector and capability feature vector mentioned above, such as the asset enforceability dimension feature and performance label. This embodiment does not impose any restrictions.

[0062] (Estimated processing time): Subject to the subject's historical case closure time T close In addition to the current load status, the higher the load, the longer the expected cycle is. (Resource load deviation): Defined as the absolute difference between the current case volume and the ideal load. The larger the deviation, the stronger the penalty. , , Weights can be configured for corresponding business processes, for example, increasing them when there is high pressure to collect payments. When regulatory assessment efficiency is increased .

[0063] As a preferred option rather than a limitation, in the initial stage of system launch or in scenarios with limited computing power, step S141 can be degenerated into weighted scorecard matching, that is, calculating a comprehensive score for each case i and candidate subject j:

[0064] in This is the normalized load deviation. The system directly selects... The highest-level entity is responsible for allocation. This method is logically transparent, easy to debug, and suitable for the cold start phase of a business.

[0065] As a preferred option rather than a limitation, when both the set of cases to be assigned C and the set of available subjects P are finite sets, and batch assignment of "one-to-one" or "one-to-a-limited quantity" needs to be achieved, step S141 can treat the set of assigned cases and the set of subjects as two sets of nodes in a bipartite graph, with edge weights set to... or The KM algorithm (Kuhn-Munkres Algorithm, an optimized variant of the Hungarian algorithm for finding the maximum weight matching) is used to solve the global optimal matching, ensuring that the overall value is maximized rather than the local optimal value of a single case.

[0066] S142, compare and verify the preliminary allocation scheme with the mandatory business rules. If there is a conflict, modify it according to the mandatory business rules to output the final allocation scheme.

[0067] As a specific approach, the aforementioned mandatory business rules have higher priority than the objective function of the optimization model. Please refer to... Figure 5 As shown, this step S142 includes S1421-S1426, wherein: S1421, Obtain the jurisdiction information of the cases to be assigned and the professional license or service coverage information of each disposal entity; if the jurisdiction of the cases to be assigned is not within the professional license or service coverage of the disposal entity, the pairing is determined to not meet the territorial compliance constraints, and the subsequent calculation of the assignment scheme is terminated.

[0068] For example, this step first extracts the administrative division code of the location of the court with jurisdiction from the structured fields of the case profile, and simultaneously reads the "service area" field from the attribute tags in the disposal entity capability tag library (stored as a list of administrative division codes or a list of courts with jurisdiction). Then, it performs a set affiliation determination: let r be the jurisdiction of case i. i The service coverage set of subject j is R. j If r i ∉R j If the region is deemed non-compliant, the system will immediately terminate all subsequent calculations for the pairing of (i,j) and record the reason code "GEO-001" in the log.

[0069] It should be noted that, according to the Lawyers Law and judicial practice, this step allows business administrators to configure a "Cross-Regional Practice Exception List" in the rules engine. For law firms that have already completed off-site registration with the courts in that region, even if their practice license does not specify that region, they can still be exempted from this restriction through a whitelist.

[0070] S1422, Obtain the historical conflict of interest record of the disposing entity; if the disposing entity has a history of conflict of interest with the debtor of the case, the pairing is determined not to meet the avoidance rule constraint, and the subsequent calculation of the allocation plan is terminated.

[0071] For example, this step first queries the conflict history list of subject j from the system's conflict of interest management module. This list is constructed through methods such as reverse lookup of historical case representation relationships, a list of permanent legal counsel units proactively declared by law firms, and disciplinary records obtained through connection with the "Lawyer Practice Integrity Information Platform" of the judicial administration department. Conflict records are indexed by the hash value of the debtor's ID number or the unified social credit code to ensure conflict detection without touching plaintext sensitive information. Subsequently, the debtor's identity identifier (after anonymization) of case i is matched with the conflict list of subject j. If a match is successful, it is determined to be a conflict of interest, and allocation is forcibly prohibited in accordance with Article 39 of the Lawyers Law and Article 51 of the Code of Conduct for Lawyers, and a conflict warning report is generated.

[0072] S1423, obtain the legal complexity dimension features of the case.

[0073] This step directly reads the legal complexity L field from the structured feature vector of the case profile. For example, this field is a binary variable (0 or 1), and its assignment logic is as follows: L=1 if the case meets any of the following conditions: (1) the number of parties involved (number of defendants > 1), or (2) an objection to jurisdiction has been raised, or (3) there is a dispute over the ownership of multiple layers of security or mortgage. Otherwise, L=0.

[0074] S1424, Obtain the professional qualification information and historical winning records of the entity handling the case.

[0075] This step retrieves the "Specialty Field" list from the attribute tags and the "Win Rate R" list from the performance tags from the entity's capability tag library. win (The rolling value for the past 12 months). Professional qualification information is not limited to lawyer's license, but also includes special access permits such as bankruptcy administrator qualification, intellectual property litigation agent qualification, and foreign-related legal service filing.

[0076] S1425, the legal complexity dimension of the case was determined to be high complexity (L). i =1), and if the disposing entity does not have the corresponding professional qualifications and its historical winning record is lower than the preset threshold, then the pairing is determined not to meet the qualification threshold constraint, and the subsequent calculation of the allocation plan is terminated.

[0077] This ruling aims to prevent cases with complex legal relationships and high difficulty in providing evidence from being assigned to entities whose professional capabilities have not been verified, and to avoid losing cases or being unable to enforce judgments due to insufficient representation capabilities.

[0078] S1426, obtain the total number of outstanding agency cases of the disposal entity and the preset upper limit of the remaining processing capacity; if the total number of outstanding agency cases of the disposal entity exceeds the preset upper limit of the remaining processing capacity, it is determined that the pairing does not meet the upper limit constraint, and the subsequent calculation of the allocation scheme is terminated.

[0079] This step first reads the "Current Case Load" status tag from the entity's capacity tag library, and the capacity limit threshold from the entity configuration table. The total number of pending cases refers to the total number of cases the entity has received but not yet archived; the remaining processing capacity limit is the recommended maximum number of cases being processed, calculated by the business department based on per capita capacity and the SLA promised in the service agreement. Then, a numerical comparison is performed: Let the current case load of entity j be L. j The remaining processing capacity is up to C. j If L j ≥C j If the system determines that the capacity is saturated, it will automatically freeze the candidate status of the subject in the current batch of assignments until the subject completes the archiving of some cases and the load drops below the threshold.

[0080] It is particularly important to emphasize that this embodiment deliberately avoids using "black box" deep learning models (such as unconstrained neural networks) for the final assignment decision in the multi-objective optimization model and constraint verification. All assignment results can be traced back to specific feature indicators (such as V). ij T ij D j The system provides clear business rules and terms. When outputting the allocation plan, the system mandates the inclusion of a "summary of the allocation reason," such as: "Case i has a high asset enforceability score, entity j has access to the vehicle management office for investigation, and its current load is below the threshold; therefore, it is given priority allocation." This mechanism aims to meet the regulatory requirements of Article 12 of the "Regulations on the Management of Algorithm Recommendations for Internet Information Services" regarding algorithm transparency and explainability, ensuring that the automated decision-making process is auditable, appealable, and remediable.

[0081] Steps S141-S142 above, through a two-tiered decision-making architecture of multi-objective optimization and hard-constraint filtering, achieve for the first time in the field of integrated legal proceedings and enforcement a quantitative matching of case characteristics and subject capabilities. Compared to traditional manual assignment relying on subjective impressions, this step uses V... ij The project directly links the recovery potential of a case to the entity's historical recovery ability; through T ij With D j The system links timeliness requirements with the workload and efficiency of the task force; it also directly blocks legal compliance risks at the source of task allocation through mandatory business rules. Empirical data shows that this mechanism can increase the matching rate of high-priority tasks with high-quality resources by more than 40% without human intervention, while reducing procedural backflows caused by jurisdictional errors or conflicts of interest to zero.

[0082] Example 2 Please see Figure 6 The present invention also provides a case assignment system 100 based on data fusion, the system comprising: The acquisition module 110 is used to acquire multi-source case data from multiple heterogeneous business systems, wherein the multi-source case data includes structured data and unstructured data. The fusion module 120 is used to perform fusion processing on the multi-source case data to generate a structured case feature vector containing multiple dimensions of quantitative features for each case. The generation module 130 is used to generate capability feature vectors for currently available disposal entities based on a preset disposal entity capability tag library. The capability feature vectors include geographical coverage information and efficiency indicators. The determination module 140 is used to determine, based on the structured case feature vector and the capability feature vector, an allocation scheme for assigning each case to the corresponding handling entity according to preset rules.

[0083] The modules in this embodiment are the same as the corresponding steps in the first embodiment described above, and will not be repeated here.

[0084] All modules / units in this embodiment are the same as the corresponding steps in the above method embodiments, and their logical relationships and working principles are also the same, so they will not be repeated here. Those skilled in the art can learn the corresponding virtual modules or units from the above method embodiments to make them correspond to the steps of the above method embodiments. Virtual modules / units not disclosed in this embodiment should also be regarded as the part of the content disclosed in this invention.

[0085] This invention acquires and fuses multi-source structured and unstructured data from multiple heterogeneous business systems, transforming the original fragmented data into structured case feature vectors containing multi-dimensional quantitative features. Simultaneously, based on a pre-defined capability tag library for handling entities, it generates capability feature vectors for currently available handling entities, including geographical coverage information and efficiency indicators. Then, based on these two types of feature vectors and pre-defined rules, it determines the allocation scheme. This achieves automated parsing and unified representation of multi-source heterogeneous data, transforming previously scattered, inconsistently formatted, and machine-unusable unstructured data into calculable and comparable numerical coordinates, breaking down data barriers between heterogeneous systems. By quantifying both case features and entity capabilities into vector form, it lays the data foundation for subsequent efficient matrix operations for matching, avoiding the high computational complexity of traditional text-based conditional matching. Furthermore, by introducing capability feature vectors containing real-time status information, the allocation decision-making process gains the ability to dynamically perceive the load status of processing units, enabling adaptive balancing of system resources during batch allocation. Ultimately, this invention improves the overall performance of automated task assignment in a multi-source heterogeneous data environment by enhancing data fusion efficiency, matching calculation efficiency, and system load balancing capabilities.

[0086] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0087] This invention also provides a computer storage medium storing a computer program that, when executed by a processor, implements the data fusion-based case assignment method as described in the above embodiments.

[0088] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above embodiments of the data fusion-based case assignment methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0089] Alternatively, if the integrated units of the present invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present invention, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, terminal, or network device, etc.) to execute all or part of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, RAM, ROM, magnetic disks, or optical disks.

[0090] Corresponding to the computer storage medium described above, one embodiment also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the case assignment method based on data fusion as described in the above embodiments.

[0091] This computer device can be a terminal, and its internal structure diagram can be as follows: Figure 7As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a case assignment method based on data fusion. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0092] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0093] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A case allocation method based on data fusion, characterized in that, include: Acquire multi-source case data from multiple heterogeneous business systems, including structured data and unstructured data; The multi-source case data is fused to generate a structured case feature vector for each case, which contains quantitative features of multiple dimensions. Based on a pre-set capability tag library for disposal entities, capability feature vectors are generated for currently available disposal entities. The capability feature vectors include geographic coverage information and efficiency indicators. Based on the structured case feature vector and the capability feature vector, an allocation scheme is determined according to preset rules to assign each case to the corresponding handling entity.

2. The method according to claim 1, characterized in that, The process of fusing the multi-source case data to generate a structured case feature vector containing multiple dimensions of quantitative features for each case includes: A text recognition module is used to extract first text information from the first unstructured data, the first unstructured data including image files and legal documents, and the first text information including case number, parties or amount of litigation claim; The semantic analysis module is used to extract the second text information from the second unstructured data, which includes the transcribed text of the call recording. The second text information includes the parties involved, the amount in dispute, the cause of action, the legal status, or preset keyword information. Based on the extracted first and second text information, the subjective repayment dimension features, asset enforceability dimension features, and legal complexity dimension features of the case are constructed and calculated to generate the structured case feature vector.

3. The method according to claim 2, characterized in that, The preset capability tag library for disposal entities generates capability feature vectors for currently available disposal entities, including: Based on the historical disposal records and publicly available data of the disposal entity, its capability tags are constructed and dynamically updated. These capability tags include: Attribute tags include the professional field category and geographical scope of practice or service of the entity handling the matter; Performance metrics include efficiency and quality indicators derived from historical data; and, Status tags are used to characterize the real-time availability status of the handling entity, including the current case load, the number of pending enforcement cases, or the available resource capacity; Based at least on the attribute tags, performance tags, and status tags, generate the capability feature vector of the currently available disposal subject.

4. The method according to claim 1, characterized in that, The allocation scheme, which determines the assignment of each case to the corresponding handling entity based on the structured case feature vector and the capability feature vector according to preset rules, includes: Based on the structured case feature vector and the capability feature vector, a preliminary assignment scheme is calculated using a multi-objective optimization model; The preliminary allocation scheme is compared and verified with the mandatory business rules. If there is a conflict, it is corrected according to the mandatory business rules to output the final allocation scheme.

5. The method according to claim 4, characterized in that, The multi-objective optimization model aims to maximize the overall disposal benefit, and the optimization objective function is defined as follows: ； Where C is the set of cases to be assigned, and P is the set of available entities for disposal; Assigning variables to represent cases Whether to assign to the main body , For the case By the main body The expected value of the proceeds from the disposal; This is the estimated disposal period; The resource load deviation of the main body after allocation is defined as the absolute difference between the current case volume and the ideal load; , , Weights can be configured for the corresponding business functions; The objective function is globally optimized in a single batch assignment to avoid local optima.

6. The method according to claim 5, characterized in that, The calculation of the preliminary allocation scheme using a multi-objective optimization model includes: Obtain jurisdictional information and legal complexity dimension characteristics of cases to be assigned; Obtain information on the business licenses or service coverage of each disposal entity, historical records of conflicts of interest, professional qualifications, historical winning records, and the preset upper limit of remaining processing capacity; If the jurisdiction of the case to be assigned is not within the scope of the entity's professional license or service coverage, the assignment plan will be deemed not to meet the geographical compliance constraints, and the subsequent calculation of the assignment plan will be terminated. If the disposing entity has a history of conflict of interest with the debtor in the case, the allocation plan will be deemed not to meet the constraints of the avoidance rule, and the subsequent calculation of the allocation plan will be terminated. If the legal complexity dimension of a case is determined to be high, and the handling entity does not possess the corresponding professional qualifications and its historical winning record is below a preset threshold, then the allocation plan is determined to not meet the qualification threshold constraints, and the subsequent calculation of the allocation plan is terminated. If the total number of pending cases handled by the entity exceeds the preset remaining processing capacity, the pairing is determined to not meet the capacity limit constraint, and the subsequent calculation of the allocation scheme is terminated.

7. The method according to claim 5, characterized in that, The calculation of the preliminary allocation scheme using a multi-objective optimization model includes: According to the formula: ； For each case to be assigned and the candidate handling entity, a comprehensive score is calculated, and the entity with the highest score is selected for assignment.

8. The method according to claim 4 or 5, characterized in that, Consider the set of cases to be assigned and the set of entities to be handled as two parts of a bipartite graph, with edge weights set as follows: or The Hungarian algorithm is used to find the globally optimal one-to-one matching.

9. A case assignment system based on data fusion, characterized in that, The system includes: The acquisition module is used to acquire multi-source case data from multiple heterogeneous business systems, including structured data and unstructured data. The fusion module is used to fuse the multi-source case data and generate a structured case feature vector containing multiple dimensions of quantitative features for each case. The generation module is used to generate capability feature vectors for currently available disposal entities based on a preset disposal entity capability tag library. The capability feature vectors include geographical coverage information and efficiency indicators. The determination module is used to determine, based on the structured case feature vector and the capability feature vector, an allocation scheme for assigning each case to the corresponding handling entity according to preset rules.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the case assignment method based on data fusion as described in any one of claims 1 to 8.

Images