A financial integration platform and a data processing method for application thereof

Abstract

The application discloses a financial integration platform and a data processing method for application of the same, relates to the technical field of financial management, and comprises the following modules: a project library construction module, which establishes a unified project management model, defines project attributes through a hierarchical structure, and associates full life cycle information of budget preparation, execution and performance evaluation for each project node; a budget association generation module, which generates a department budget draft according to distribution rules, and adjusts a mechanism to associate budget preparation and project execution progress; an index dynamic management module, which decomposes a ratified budget result into executable indexes in stages, and locks, adjusts and recycles the executable indexes according to project dimensions; a payment control module, which checks payment conditions based on project budget balances and contract performance states; and an emergency and data maintenance module, which performs local emergency payment when a network is interrupted, maintains cross-year project data, and realizes automatic association of budget execution data of historical projects and newly-built projects. The application improves the efficiency of financial management.

Description

Technical Field

[0001] This application relates to the field of fiscal management technology, and in particular to a data processing method for an integrated fiscal platform and its application. Background Technology

[0002] In traditional fiscal management, various subsystems such as budget preparation, payment, and non-tax revenue management operate independently, lacking a unified project management model, making it difficult to dynamically link business data. Especially in the budget execution phase, the matching degree between fund allocation and project progress is low, making it difficult to achieve refined management. As a result, existing fiscal management platforms often suffer from problems such as loose coordination between budget and execution and inconsistent data standards, leading to low fiscal management efficiency and an inability to adapt to increasingly complex fiscal business needs. Summary of the Invention

[0003] In order to improve the efficiency of fiscal management and provide an integrated fiscal platform that can effectively adapt to the increasingly complex needs of fiscal business, this application provides an integrated fiscal platform and a data processing method for its application.

[0004] Firstly, the objective of this invention is achieved through the following technical solution:

[0005] An integrated fiscal platform, comprising:

[0006] The project library construction module establishes a unified project management model based on the needs of fiscal operations. It defines project attributes through a hierarchical structure, and links each project node with information on the entire lifecycle of budget preparation, execution, and performance evaluation.

[0007] The budget generation module generates a departmental budget draft based on the hierarchical relationship and funding requirements of projects in the project management model and according to preset allocation rules. It also uses a dynamic adjustment mechanism for the corresponding indicators to link budget preparation with project execution progress.

[0008] The indicator dynamic management module decomposes the approved budget results into actionable indicators at each level, calculates the available funds based on the project progress, and performs indicator locking, adjustment, and recovery operations on the actionable indicators according to the project dimension.

[0009] The payment control module obtains the fund application information of the budget unit, verifies the payment conditions based on the project budget balance and contract performance status, generates direct payment vouchers or authorized payment limits, and updates the project fund flow.

[0010] The emergency and data maintenance module provides stand-alone emergency payment functionality, generating emergency payment instructions locally when the network is interrupted; it uses version stamp technology to maintain cross-year project data, enabling automatic association of budget execution data between historical and new projects.

[0011] By adopting the above technical solution, project attributes include basic expenditure items, special expenditure items, and multi-year projects; the emergency and data maintenance module provides emergency handling mechanisms and data consistency maintenance methods. The stand-alone emergency payment function can be implemented as a stand-alone system, generating emergency payment instructions through locally cached business data. After network recovery, budget execution data and project execution status are automatically synchronized and updated; specifically, by defining project attributes hierarchically and associating them with full lifecycle information, dynamic tracking of the entire project process from initiation, budget preparation, execution to performance evaluation is achieved, solving the technical problems of scattered and difficult-to-trace project information in the traditional model; and by automatically generating budget drafts based on project hierarchy and funding needs... The application addresses the technical challenge of budget preparation and execution being disconnected by implementing a dynamic adjustment mechanism for relevant indicators. This mechanism achieves real-time correlation between budget preparation and project execution progress. Furthermore, it optimizes fund allocation through hierarchical indicator decomposition, dynamic calculation of available funds, and multi-dimensional indicator locking / adjustment / recovery functions, ensuring precise matching of fund usage with project progress. Finally, by real-time verification of payment conditions, automatic generation of payment vouchers, and synchronous updating of fund flow, combined with stand-alone emergency payment and version stamp technology, it effectively prevents overpayment risks while ensuring business continuity during network interruptions. Therefore, this application improves fiscal management efficiency and provides an integrated fiscal platform that effectively adapts to the increasingly complex needs of fiscal operations.

[0012] In a preferred embodiment, this application further includes a data interface subsystem, which comprises:

[0013] The data exchange module, based on a predefined fiscal business data model, achieves bidirectional data synchronization with designated key systems through a Web Service interface, and supports conversion of different data formats.

[0014] The log monitoring module uses a distributed log framework to record the timestamps, operator IDs, and data verification results of all data exchange operations in real time, and supports multi-dimensional retrieval operations based on business type and anomaly level.

[0015] The network security isolation module deploys a physical gateway to achieve physical isolation between the internal and external networks of the finance and government affairs system, and transmits data through encrypted channels using the national cryptographic SM4 algorithm;

[0016] The version compatibility adaptation layer provides a unified API gateway and a built-in version manager, supporting dynamic adaptation of interface protocols between new and old systems.

[0017] By adopting the above technical solutions, the Web Service interface enables bidirectional data synchronization with other key systems (such as the fiscal accounting system and the non-tax revenue system), and supports the conversion of different data formats; the log monitoring module supports multi-dimensional retrieval, enhancing the transparency of the integrated fiscal platform and facilitating problem investigation and auditing; the network security isolation module improves the security of data transmission and protects sensitive information from leakage; and the version compatibility adaptation layer supports dynamic adaptation of interface protocols between new and old systems. This ensures the system's backward compatibility and reduces the impact of system version upgrades to the integrated fiscal platform.

[0018] In a preferred embodiment of this application: the emergency response and data maintenance module implements a standalone emergency payment function through a standalone emergency subsystem; the platform also includes:

[0019] The distributed service mesh uses a microservice framework to decompose the core business modules of the integrated finance platform, shortening the response time of service calls between various core business modules to a specified time range, and supporting circuit breaking, degradation, and rate limiting mechanisms.

[0020] The intelligent routing engine dynamically allocates multiple payment request nodes based on the consistent hashing algorithm, combines machine learning to predict the load pressure of each core business module, obtains the application load pressure prediction result, and performs node weight adjustment operation based on the application load pressure prediction result.

[0021] The time-series database establishes composite indexes for frequently queried fields and adopts an LSM tree storage structure; it uses open-source container orchestration components to achieve containerized deployment, automatically scaling up Pod nodes when the CPU utilization of the integrated financial platform exceeds a preset CPU utilization threshold, and controlling the cold start time to be less than a preset cold start duration threshold.

[0022] By adopting the above technical solutions, the distributed service mesh uses a microservice framework to decompose core business modules, shortening service call response time and supporting circuit breaking, degradation, and rate limiting mechanisms, which helps improve the response speed and stability of the integrated financial platform. The intelligent routing engine combines machine learning to predict load pressure and adjust node weights, optimizing resource allocation and improving system processing capabilities. The time-series database uses open-source container orchestration components (such as Kubernetes) for containerized deployment, automatically scaling up Pod nodes when CPU utilization exceeds a threshold. This improves query efficiency and system scalability.

[0023] In a preferred embodiment, this application also includes:

[0024] Based on the needs of fiscal operations, query condition combination rules are defined, supporting free combination of multi-dimensional parameters and generating dynamic query templates; the multi-dimensional parameters include project attributes, funding sources, time periods, and responsible units;

[0025] The entire lifecycle information and the budget execution data are dynamically linked to enable penetrating query capabilities from business process nodes such as budget preparation, indicator decomposition, payment progress, and contract performance, so as to support in-depth traceability query of cross-year projects;

[0026] Online analytical processing (OLAP) algorithms are used to perform multidimensional analysis on query results to generate standardized visual reports.

[0027] By adopting the above technical solutions, the query and statistics subsystem supports the free combination of query conditions with various parameters and generates dynamic query templates, thereby providing a highly flexible query method, meeting the diverse query needs of users, and improving user experience; by dynamically linking the project's entire lifecycle information with budget execution data, it achieves penetrating queries and supports in-depth tracing of cross-year projects; then, it uses online analytical processing algorithms (such as OLAP technology) to perform multidimensional analysis on the query results and generate standardized visual reports, improving the intuitiveness and readability of data presentation.

[0028] In a preferred embodiment, this application also includes:

[0029] A role-permission matrix is ​​established based on the RBAC model. The role-permission matrix is ​​configured with access permissions according to the dimensions of business department, funding nature, and project type.

[0030] Configure the business process model and corresponding business process nodes in the project management model, and trigger dynamic adjustment of approval nodes, node return for re-review, and reminder information based on the business process nodes;

[0031] In the project management model, operation and maintenance monitoring index parameters are collected in real time, and preset threshold alarm rules are associated with these operation and maintenance monitoring index parameters. When the preset threshold alarm rules meet the alarm conditions, alarm prompt instructions are triggered.

[0032] By adopting the above technical solution and configuring data access permissions according to dimensions such as business department, funding nature, and project type through the RBAC (Role-Based Access Control) model, fine-grained data access control is achieved. By configuring the business process model and corresponding business process nodes in the project management model, functions such as dynamically adjusting approval nodes, returning for resubmission, and reminders are allowed, making the business process more flexible and efficient. At the same time, it will collect operation and maintenance monitoring indicators (such as key performance indicators such as CPU, memory, and disk I / O) in real time and associate them with preset threshold alarm rules. When alarm conditions are met, alarm prompts will be automatically triggered to help operation and maintenance managers discover and solve problems in a timely manner.

[0033] Secondly, the objective of this invention is achieved through the following technical solution:

[0034] A data processing method applied to an integrated fiscal platform includes:

[0035] A unified project management model is established based on the needs of fiscal operations. Project attributes are defined through a hierarchical structure, and each project node is associated with information throughout the entire lifecycle of budget preparation, execution, and performance evaluation.

[0036] Based on the hierarchical relationship and funding requirements of projects in the project management model, a draft departmental budget is generated according to the preset allocation rules, and the budget preparation and project execution progress are linked by the dynamic adjustment mechanism of the target indicators.

[0037] The approved budget results are broken down into actionable indicators at each level. The available funds are calculated in conjunction with the project progress. The actionable indicators are locked, adjusted, and recovered according to the project dimension.

[0038] Obtain information on fund application from budget units, verify payment conditions based on project budget balance and contract performance status, generate direct payment vouchers or authorized payment limits, and update project cash flow.

[0039] By adopting the above technical solutions, project attributes are defined based on a hierarchical structure, and information such as budget preparation, execution, and performance evaluation are linked together to form a comprehensive project lifecycle management system. The system automatically generates a draft budget using hierarchical relationships and funding requirements, and links budget preparation with project execution progress in real time through a dynamic adjustment mechanism to improve the adaptability and accuracy of budget preparation.

[0040] In a preferred embodiment of this application, the method further includes:

[0041] A sample dataset is constructed based on fiscal business data within a preset time range. The constructed fund allocation optimization model is trained using the sample dataset, and the parameters of the fund allocation optimization model are adjusted to obtain the trained fund allocation optimization model.

[0042] The trained fund allocation optimization model is used to determine several target fund allocation models corresponding to several preset quantiles, and the target fund allocation models are used to determine several fund allocation benchmark values ​​corresponding to the several preset quantiles.

[0043] A fund allocation efficiency scoring formula is constructed using the aforementioned several fund allocation benchmark values. The fund allocation efficiency score of fiscal operations is determined based on the fund allocation efficiency scoring formula and real-time fiscal business data. The early warning mechanism is then determined based on the fund allocation efficiency score.

[0044] By adopting the above technical solution, a fund allocation optimization model is trained using historical data, and its parameters are adjusted to obtain the trained fund allocation optimization model. At the same time, by setting fund allocation models with different quantiles, optimization suggestions can be provided for different fund allocation scenarios. Different quantiles correspond to different risk levels (e.g., 0.2 quantile corresponds to low-risk conservative allocation, and 0.9 quantile corresponds to high-risk aggressive allocation). Based on the scoring formula, an early warning is triggered to indicate that the fund allocation deviates from the benchmark, so as to identify abnormal allocation behavior (such as excessive favoritism towards a certain project) and promote the fairness of fund allocation.

[0045] In a preferred embodiment of this application: the step of determining several target fund allocation models corresponding to several preset quantiles using the trained fund allocation optimization model, and determining several fund allocation benchmark values ​​corresponding to the several preset quantiles through the several target fund allocation models, includes:

[0046] The quantiles of the trained fund allocation optimization model are set to 0.2, 0.5, 0.7 and 0.9 respectively, and the fund allocation optimization models corresponding to the 0.2, 0.5, 0.7 and 0.9 quantiles are trained using the sample dataset to obtain the first target fund allocation model corresponding to the 0.2 quantile, the second target fund allocation model corresponding to the 0.5 quantile, the third target fund allocation model corresponding to the 0.7 quantile, and the fourth target fund allocation model corresponding to the 0.9 quantile.

[0047] A single-project dataset is constructed based on single-project data within a preset second time range, and the single-project dataset is input into the first target fund allocation model, the second target fund allocation model, the third target fund allocation model, and the fourth target fund allocation model respectively to obtain the corresponding first benchmark value, second benchmark value, third benchmark value, and fourth benchmark value.

[0048] By adopting the above technical solution, fund allocation optimization models with different quantiles are set and trained separately. By setting four quantiles of 0.2, 0.5, 0.7, and 0.9, the corresponding fund allocation models are trained respectively, thereby obtaining multiple benchmark values. By covering the entire scenario from conservative (0.2) to aggressive (0.9), different fiscal objectives are met. The refined analysis results intuitively reflect the model's sensitivity to specific scenarios. An efficiency scoring formula is constructed through the benchmark values ​​to quantify the fund allocation effect, so as to transform subjective allocation standards into quantifiable scoring indicators (such as fund utilization rate and project completion rate). The model parameters are continuously optimized based on the scoring results to improve long-term allocation efficiency.

[0049] In a preferred embodiment of this application: the step of constructing a sample dataset based on fiscal business data within a preset time range, training a constructed fund allocation optimization model using the sample dataset, and adjusting the parameters of the fund allocation optimization model to obtain a trained fund allocation optimization model includes:

[0050] The quantile parameter of the fund allocation optimization model is set to 0.5, and the fund allocation optimization model is trained using the sample dataset to obtain the optimization model to be determined.

[0051] The degree of fit between the optimization model to be determined and the fund allocation trend is determined based on the coefficient of determination, so as to determine whether the model accuracy of the optimization model to be determined reaches the preset model accuracy threshold based on the degree of fit.

[0052] If the model accuracy of the model to be optimized does not reach the preset model accuracy threshold, the parameters of the model to be optimized are adjusted and retrained until the model accuracy of the model to be optimized reaches the preset model accuracy threshold, so as to determine the model to be optimized as the post-trained fund allocation optimization model.

[0053] By adopting the above technical solution, the quantile is fixed at 0.5, the model to be optimized is trained, and the degree of fit is evaluated by the coefficient of determination (R²). The quantile of 0.5 corresponds to the median model, which is suitable for capturing the central tendency of fund allocation and avoiding the interference of extreme values. The coefficient of determination quantifies the degree of agreement between the model and the actual fund allocation trend, ensuring that the model's explanatory power meets the standard (e.g., R²≥0.85). If the model accuracy does not reach the threshold (e.g., R²<0.85), parameter adjustment and retraining are triggered until the standard is met. Iterative optimization avoids local optima, improves the model's adaptability to complex financial scenarios, and enhances the model's robustness.

[0054] In a preferred example, the construction of the sample dataset based on fiscal business data within a preset time range includes:

[0055] The unified data collection interface is used to traverse several sets of fiscal business data within a preset time range on the integrated fiscal platform.

[0056] Based on the timestamp, the several sets of fiscal business data are synchronously calibrated, and missing values ​​and outliers in the several sets of fiscal business data are removed to obtain several sets of preprocessed fiscal business data.

[0057] Remove the project number information from the preprocessed fiscal business data, summarize the preprocessed fiscal business data to obtain summarized fiscal business data, and construct a sample dataset based on the summarized fiscal business data.

[0058] By adopting the above technical solutions, and traversing fiscal business data based on a unified data interface, the problem of data silos across multiple systems is solved. This application connects subsystems such as budget preparation, execution, and payment, achieving seamless cross-module data integration. Data synchronization is calibrated based on timestamps, and missing and outlier values ​​are removed to improve data quality. By removing project number information and aggregating data, a de-identified sample set is constructed for data anonymization and aggregation optimization. Through aggregation, individual noise is eliminated, group characteristics (such as the average execution progress of departments) are preserved, and the generalization ability of the model is improved.

[0059] In summary, this application includes at least one of the following beneficial technical effects:

[0060] 1. Break down the "data silos" of scattered project information storage in the traditional fiscal system, and achieve dynamic linkage of budget, execution, and evaluation data; support cross-year project progress tracking and historical data backtracking through full lifecycle correlation; achieve flexible allocation of funds through dynamic adjustment of indicators (such as setting adjustment priorities), reducing the risk of human operation;

[0061] 2. Through refined access control (RBAC model), dynamic process adaptation (node ​​adjustment and expediting), intelligent operation and maintenance monitoring (threshold alarms), and machine learning-driven fund allocation optimization (quantile model and efficiency scoring). Attached Figure Description

[0062] Figure 1 This is a schematic diagram of the framework of an integrated fiscal platform according to one embodiment of this application;

[0063] Figure 2 This is a practical application example diagram of the dynamic indicator management module in a fiscal integrated platform according to one embodiment of this application;

[0064] Figure 3 This is a practical application example diagram of the direct payment function of the payment control module in a fiscal integrated platform according to one embodiment of this application;

[0065] Figure 4 This is a practical application example diagram of authorized payment in a payment control module of a fiscal integrated platform according to one embodiment of this application;

[0066] Figure 5 This is an example diagram illustrating the application of a payment adjustment function in a fiscal integration platform according to one embodiment of this application;

[0067] Figure 6 This is a flowchart of a data processing method applied to an integrated fiscal platform according to an embodiment of this application. Detailed Implementation

[0068] The present application will be further described in detail below with reference to the accompanying drawings.

[0069] In one embodiment, such as Figures 1 to 5 As shown, this application discloses an integrated fiscal platform, which includes a project database construction module, a budget linkage generation module, an indicator dynamic management module, a payment control module, and an emergency and data maintenance module. The project database construction module establishes a unified project management model based on fiscal business needs; this project management model is the project database. Project attributes are defined through a hierarchical structure, with each project node associated with the entire lifecycle information of budget preparation, execution, and performance evaluation. The hierarchical structure divides projects into "Level 1 Projects (Strategic), Level 2 Projects (Business), and Level 3 Projects (Execution)." The entire lifecycle information includes stage data such as project initiation basis, budget preparation, execution progress, acceptance report, and performance evaluation results. Specifically, by surveying the needs of fiscal business departments (such as the budget department), project attribute fields (such as project name, type, total budget, and responsible unit) are defined, and a hierarchical structure model is designed. For example, Level 1 projects are divided by field (such as education and healthcare), Level 2 projects are subdivided by business line (such as compulsory education → teacher training), and Level 3 projects correspond to specific execution plans (such as a middle school teacher training course). An API interface is developed to allow the budget preparation system and performance evaluation system to push data to the project database in real time.

[0070] The budget-linked generation module generates a departmental budget draft based on the hierarchical relationship and funding needs of projects in the project management model, according to preset allocation rules. It uses a dynamic adjustment mechanism of the budget allocation indicator to link budget preparation with project execution progress. The preset allocation rules include "allocation according to staffing ratio" and "allocation according to historical expenditure ratio". Specifically, it captures the latest progress of the project in real time to obtain the project execution progress, and records and tracks the use of funds to obtain fund usage data. It links the project execution progress with the project progress corresponding to the budget preparation (here, the planned progress). The budget allocation indicator refers to the amount of funds that can actually be used for expenditure within the total budget approved by the People's Congress. It is the detailed result of the approved budget and must meet two conditions: conforming to the project progress and contract performance status, and being the executable amount determined after dynamic adjustment (such as adjustment and recovery).

[0071] Specifically, the dynamic adjustment mechanism for performance indicators includes multi-source data collection and data verification mechanisms. Multi-source data collection involves capturing progress update events from project execution systems (such as contract management systems and ERP systems) in real time using the WebSocket protocol to obtain executable indicators, and using Kafka message queues for asynchronous data synchronization. The data verification mechanism utilizes distributed transactions (such as the Seata framework) to ensure consistency between budget data and execution data, and uses CRC checks to ensure data transmission integrity. The dynamic adjustment mechanism for performance indicators applies a dynamic calculation model for dynamic hierarchical control, where the calculation formula is: Where Q(t) refers to the executable index at time t; Here, is the risk adjustment coefficient (range 0.8-1.2); B is the approved total budget; P(t) is the contract performance progress at time t; E(t) is the cumulative actual expenditure at time t; the dynamic hierarchical control strategy is as follows:

[0072] ① When the execution progress P(t) < 50%, the risk adjustment coefficient is 0.8, and the fund release ratio (i.e., adjustment strategy) is 20% frozen;

[0073] ②When 50%≤execution progress P(t)≤80%, the risk adjustment coefficient is 0.9, and the fund release ratio is 10% frozen;

[0074] ③When P(t)≥80%, the risk adjustment coefficient is 0.9, and the fund release ratio is full release.

[0075] The fund allocation scheme is optimized based on genetic algorithms, and a real-time feedback channel is established: the dynamic adjustment mechanism of the target indicators supports the parallel adjustment of multiple projects, and the adjustment results are fed back to the budget preparation module in real time through RESTful API. The Webhook mechanism is used to trigger the automatic adjustment of the approval process.

[0076] like Figure 2 As shown, the indicator dynamic management module decomposes the approved budget results into executable indicators at each level, calculates the available funds based on the project progress, and performs indicator locking, adjustment, and recovery operations on an indicator-by-project basis. Executable indicators refer to the approved budget amount that can be directly used for payment; adjustment and recovery operations are manifested as the transfer or cancellation of indicators between different projects.

[0077] like Figure 3 and Figure 4As shown, the payment control module obtains the fund application information from the budget unit, verifies the payment conditions based on the project budget balance and contract performance status, generates a direct payment voucher or authorized payment limit, and updates the project's cash flow. Contract performance status includes the contract signing date, payment milestones (e.g., "30% payment upon equipment delivery"), and performance progress. Direct payment is when the finance department directly pays the supplier; authorized payment is when the budget unit makes its own payment (subject to limit control). In this embodiment, the fund application verification is based on receiving the fund application submitted by the budget unit (e.g., "Pay 1 million yuan for equipment to a supplier"). First, the budget balance is checked. If the current project's available limit is ≥ 1 million yuan, the contract matches. If the application amount is ≤ total contract amount × achieved performance progress (e.g., total contract price 10 million yuan, 30% delivered, allowable payment ≤ 3 million yuan), a "direct payment voucher" is generated when the payment conditions are met; otherwise, the application is rejected, a message is pushed to the budget unit's personnel, and the cash flow is updated simultaneously. Figure 5 As shown, in practical applications, the integrated fiscal platform also provides payment adjustment functions.

[0078] The emergency response and data maintenance module provides a standalone emergency payment function, generating emergency payment instructions locally when the network is interrupted. This module implements the standalone emergency payment function through a standalone emergency subsystem. A standalone system is designed to generate emergency payment instructions using locally cached business data during network interruptions. After network recovery, a Diff algorithm is used to compare the differences between the locally cached data and the core system data. Unsynchronized emergency instructions are then submitted in batches to the core system, updating the project fund flow table and payment status flags. This application also employs version stamp technology to maintain cross-year project data, enabling automatic association between historical and newly created project budget execution data. Version stamp technology adds a unique timestamp and version number to each data record, resolving data conflict issues.

[0079] The integrated fiscal platform also includes a data interface subsystem, which comprises a data exchange module, a log monitoring module, a network security isolation module, and a version compatibility adaptation layer. The data exchange module, based on a predefined fiscal business data model, enables data transmission and synchronization between the fiscal system and external systems (such as accounting systems and non-tax revenue systems). It achieves bidirectional data synchronization with designated key systems through a Web Service interface, supporting different data format conversions. The Web Service interface includes standardized communication interfaces based on SOAP protocols or RESTful APIs, supporting XML / JSON data formats. The log monitoring module uses a distributed log framework to record the timestamps, operator IDs, and data verification results of all data exchange operations in real time. It supports multi-dimensional retrieval operations based on business type and anomaly level, using a distributed log framework such as ELK Stack. The network security isolation module deploys physical gateways to achieve physical isolation between the fiscal administration intranet and the external network. These physical gateways are hardware-level network isolation devices. Data is transmitted through encrypted channels using the national cryptographic SM4 algorithm. The version compatibility adaptation layer provides a unified API gateway (such as a RESTful API gateway) with a built-in version manager, supporting dynamic adaptation of interface protocols between new and old systems.

[0080] The integrated fiscal platform also includes a distributed service mesh, an intelligent routing engine, and a time-series database. Figure 1 (Not shown in the image) The core business modules of the integrated finance platform are split into microservices, which shortens the service call response time between the core business modules to a specified time range and supports circuit breaking, degradation and rate limiting mechanisms; microservices frameworks such as Dubbo microservices framework.

[0081] The intelligent routing engine dynamically allocates multiple payment request nodes based on the consistent hashing algorithm, and combines machine learning to predict the load pressure of each core business module to obtain the application load pressure prediction result. Based on the application load pressure prediction result, it performs node weight adjustment operations. The time-series database establishes composite indexes for high-frequency query fields and adopts an LSM tree storage structure. It uses open-source container orchestration components (such as Kubernetes, or K8S) to achieve containerized deployment. When the CPU utilization of the integrated financial platform is greater than the preset CPU utilization threshold, it automatically expands the Pod nodes and controls the cold start time of the integrated financial platform to be less than the preset cold start duration threshold.

[0082] The integrated fiscal platform also includes rules for defining query condition combinations based on fiscal business needs, supporting free combination of multi-dimensional parameters and generating dynamic query templates. Multi-dimensional parameters include project attributes (such as name and type), funding sources (such as government appropriations and self-raised funds), time periods (such as annual and quarterly), and responsible units (such as departments and sections). It dynamically links full lifecycle information and budget execution data, enabling penetrating query capabilities across business process nodes from budget preparation, indicator decomposition, payment progress, and contract performance. This supports in-depth traceability queries for cross-year projects. This application uses recursive common expressions (CTEs) to achieve hierarchical penetrating queries. Penetrating queries refer to the ability to expand data details level by level through hierarchical relationships (such as from project overview to budget preparation, payment progress, and contract performance). It can establish cross-system data relationship graphs, using the project ID as the primary key to link budget tables, payment flow tables, and contract ledger tables. Online analytical processing (OLAP) algorithms are used to perform multi-dimensional analysis on the query results to generate standardized visual reports, such as PDF, Excel, and PPT.

[0083] The integrated fiscal platform also includes a role-permission matrix based on the RBAC model. This matrix configures access permissions according to dimensions such as business department, funding nature, and project type. Business departments are categorized into functional departments within the finance department; funding nature refers to the source of budgetary funds (e.g., government appropriations, non-tax revenue, government bonds); and project type refers to the attribute classification of fiscal projects (e.g., infrastructure, public welfare, information technology). Roles are first defined based on business department, funding nature, and project type, and a mapping table between roles and permissions is established. A visual interface is provided, supporting permission configuration by dimension: for example, at the business department level, selecting "Budget Department" grants access to the budget preparation module; at the funding nature level, "non-tax revenue" funds are restricted to approval only by the Finance Department; and at the project type level, infrastructure projects require an additional feasibility report. Upon user login, a permission token is dynamically generated based on the user's role, business department, funding nature, and project type.

[0084] Configure the business process model and corresponding business process nodes in the project management model. Based on these business process nodes, trigger dynamic adjustments to approval nodes, node resubmission for review, and reminder messages. The business process model includes the flow rules of financial business (e.g., "budget preparation - indicator decomposition - payment application - contract performance"). Approval nodes refer to the steps in the process that require manual review (e.g., "approval by the supervisor" or "financial review"). Resubmission for review refers to resubmission to the previous node due to incomplete materials or data errors. Dynamic node configuration includes configuring node attributes and node approval rules in the process engine. Node types include approval nodes, automatic nodes, and branch nodes.

[0085] In the project management model, operation and maintenance monitoring indicators are collected in real time, including CPU, memory, disk, and I / O indicators. Based on the operation and maintenance monitoring indicators, preset threshold alarm rules are associated. The preset threshold alarm rules trigger alarm prompts when alarm conditions are met. The threshold alarm rules refer to preset abnormal judgment conditions (such as triggering an alarm when CPU utilization is ≥85%).

[0086] In one embodiment, such as Figure 6 As shown, a data processing method for an integrated fiscal platform is provided. This method, applied to an integrated fiscal platform, specifically includes the following steps:

[0087] S1: Establish a unified project management model based on the needs of fiscal operations, define project attributes through a hierarchical structure, and associate each project node with information on the entire lifecycle of budget preparation, execution, and performance evaluation.

[0088] Specifically, by surveying financial departments (such as the budget department), we clarify the project attribute requirements, define a hierarchical structure (Level 1 projects → Level 2 projects → Level 3 projects), design a database table structure, and link data tables throughout the entire lifecycle of budget preparation, execution, and performance evaluation through "project ID".

[0089] S2: Based on the hierarchical relationship and funding requirements of projects in the project management model, generate a draft departmental budget according to the preset allocation rules, and use the dynamic adjustment mechanism of the target indicators to link budget preparation with project execution progress.

[0090] Specifically, based on the hierarchical relationships in the project management model, the budget amount is allocated according to preset rules (such as staffing ratio and historical expenditure ratio); for example, a bureau's total budget is 100 million yuan, which is allocated according to "personnel expenses: project expenses = 5:5", generating a draft budget for secondary projects.

[0091] S3: Decompose the approved budget results into actionable indicators at each level, calculate the available funds in conjunction with the project progress, and lock, adjust and recover the actionable indicators according to the project dimension.

[0092] Specifically, the approved budget results are broken down into departmental indicators, project indicators, etc.; the real-time available quota is calculated in conjunction with the project progress (such as the contract performance ratio), for example, available quota = approved budget × contract performance progress - amount already paid; the locking operation refers to freezing part of the quota for projects that are about to exceed the budget, which can only be unfrozen after approval; the adjustment operation refers to adjusting surplus funds across projects, which requires recording the reason for the adjustment and the approval path; the recovery operation refers to automatically recovering the unused quota for projects that have not been started for a long time (such as those that have been inactive for more than 1 year).

[0093] S4: Obtain the fund application information of the budget unit, verify the payment conditions based on the project budget balance and contract performance status, generate direct payment vouchers or authorized payment limits, and update the project fund flow.

[0094] In one embodiment, a data processing method applied to an integrated fiscal platform further includes:

[0095] S10: Construct a sample dataset based on fiscal business data within a preset time range, use the sample dataset to train the constructed fund allocation optimization model, and adjust the parameters of the fund allocation optimization model to obtain the trained fund allocation optimization model.

[0096] In this embodiment, the sample dataset is a training dataset built based on historical financial business data (such as budget preparation, execution progress, and contract performance status); the fund allocation optimization model is a machine learning model used to predict the optimal fund allocation scheme.

[0097] Specifically, a sample dataset is constructed based on fiscal business data within a preset time range. The constructed fund allocation optimization model is trained using this sample dataset, and the parameters of the fund allocation optimization model are adjusted to obtain the trained fund allocation optimization model, including:

[0098] S101: Set the quantile parameter of the capital allocation optimization model to 0.5, and train the capital allocation optimization model using the sample dataset to obtain the optimization model to be determined.

[0099] In this embodiment, during the model initialization phase of the fund allocation optimization model, the model quantile parameter is set to 0.5.

[0100] S102: Determine the degree of fit between the optimization model to be determined and the capital allocation trend based on the coefficient of determination, and determine whether the model accuracy of the optimization model to be determined reaches the preset model accuracy threshold based on the degree of fit.

[0101] In this embodiment, the coefficient of determination (R²) is used to measure the degree of fit between the model's predicted value and the actual value. The value ranges from 0 to 1, and the closer the value is to 1, the higher the degree of fit. The preset model accuracy threshold refers to the pre-set model accuracy standard (such as R²≥0.85), which is used to determine whether the model meets the business requirements.

[0102] Specifically, the coefficient of determination (R²) of the model is calculated using the test set data:

[0103] Where i is the index of the sample in the dataset (from 1 to n); The actual value allocated to funds; These are model predictions, which can be calculated and predicted using a linear regression model. This is the average value of the actual funds allocated.

[0104] S103: If the model accuracy of the model to be optimized does not reach the preset model accuracy threshold, the parameters of the model to be optimized are adjusted and retrained until the model accuracy of the model to be optimized reaches the preset model accuracy threshold, so as to determine the model to be optimized as the optimized model for fund allocation after training.

[0105] Specifically, a preset model accuracy threshold is set (e.g., R² ≥ 0.85). If the calculated R² value is ≥ 0.85, the model accuracy meets the standard; otherwise, the parameter adjustment process begins. The parameter adjustment strategy includes: if the model accuracy does not meet the standard, the parameters are adjusted using a grid search method: traverse preset parameter combinations (e.g., number of trees: 100~500, maximum depth: 5~15); retrain the model using the adjusted parameters and recalculate the coefficient of determination; repeat step S102 until the model R² value is ≥ 0.85, and then stop the iteration.

[0106] Furthermore, in this embodiment, a sample dataset is constructed based on fiscal business data within a preset time range, including:

[0107] S111: Traverse several sets of fiscal business data within a preset time range through a unified data collection interface.

[0108] In this embodiment, a RESTful API interface is developed, and data collection specifications (such as field names, data formats, and update frequencies) are defined. Through paginated queries or streaming processing, all financial business data (such as data from the past 3 years) within a preset time range are traversed. This application supports access to multiple data sources, such as relational databases, non-relational databases, and file systems (CSV / Excel).

[0109] S112: Based on timestamps, synchronize and calibrate several sets of fiscal business data, remove missing and outlier values ​​from several sets of fiscal business data, and obtain several sets of preprocessed fiscal business data.

[0110] In this embodiment, the null value rate of statistical fields (e.g., if the null value rate of a field is >20%, it is marked as a missing field) and records in time series data that are missing for more than 3 consecutive time points are marked as invalid to detect missing values. Forward fill or mean / median is used to fill numeric fields. Missing values ​​in text fields are marked as NULL or default values ​​to complete the missing value filling operation. The 3σ principle is used to remove values ​​that exceed the mean ± 3 times the standard deviation. In practical applications, an anomaly detection model can also be trained using an Autoencoder to mark abnormal records.

[0111] S113: Remove the project number information from several sets of preprocessed fiscal business data, summarize the several sets of preprocessed fiscal business data to obtain summarized fiscal business data, and construct a sample dataset based on the summarized fiscal business data.

[0112] In this embodiment, sensitive fields (such as project numbers) are replaced with hash values ​​or unique identifiers while preserving field relationships. For example, a mapping table is used to ensure that the desensitized project numbers are traceable to the original data. Data is aggregated according to multiple dimensions, such as aggregating the total budget by project type (infrastructure / people's livelihood) and summarizing the payment progress by quarter. Derivative fields in the data are calculated, such as budget execution rate = actual expenditure / budget amount; payment delay days = actual payment date - contractual date. Then, the data is stored according to structured and unstructured data, and the data is partitioned, such as by time (e.g., by month) or by project type (e.g., infrastructure, people's livelihood), and a sample dataset is constructed.

[0113] S20: Use the trained capital allocation optimization model to determine several target capital allocation models corresponding to several preset quantiles, so as to determine several capital allocation benchmark values ​​corresponding to several preset quantiles through several target capital allocation models.

[0114] In this embodiment, the target funding allocation model is a specialized model trained for different quantiles to generate differentiated benchmark values. A quantile is a numerical point that divides a set of data into several equal parts according to its numerical value, representing the location characteristics of the data distribution. The funding allocation benchmark value is a reasonable funding allocation amount predicted by the target funding allocation model based on historical data and project characteristics (such as budget amount and contract performance progress). It is a recommended funding allocation standard value used to measure the rationality of the actual allocation.

[0115] Specifically, step S20 includes:

[0116] S201: Set the quantiles of the trained capital allocation optimization model to 0.2, 0.5, 0.7 and 0.9 respectively, and train the capital allocation optimization models corresponding to the 0.2, 0.5, 0.7 and 0.9 quantiles using the sample dataset to obtain the first objective capital allocation model corresponding to the 0.2 quantile, the second objective capital allocation model corresponding to the 0.5 quantile, the third objective capital allocation model corresponding to the 0.7 quantile, and the fourth objective capital allocation model corresponding to the 0.9 quantile.

[0117] In this embodiment, the quantile is a quantile value set according to business needs (such as 0.2, 0.5, 0.7, 0.9), which corresponds to conservative, balanced, aggressive, and radical strategies, respectively; the target fund allocation model is an independent model trained for a single quantile, which outputs the benchmark value of fund allocation under that quantile.

[0118] Specifically, in the post-trained capital allocation optimization model, the quantile parameters are set to 0.2, 0.5, 0.7, and 0.9 respectively, generating four independent sub-models. For example, the quantile 0.2 model predicts the lower limit of capital allocation (conservative strategy); the quantile 0.5 model predicts the median value of capital allocation (balanced strategy); the quantile 0.7 model predicts the upper limit of capital allocation (aggressive strategy); and the quantile 0.9 model predicts the high-risk value of capital allocation (aggressive strategy).

[0119] S202: Construct a single-project dataset based on single-project data within a preset second time range, and input the single-project dataset into the first target fund allocation model, the second target fund allocation model, the third target fund allocation model, and the fourth target fund allocation model respectively to obtain the corresponding first benchmark value, second benchmark value, third benchmark value, and fourth benchmark value.

[0120] In this embodiment, the single-project dataset is historical data for a single fiscal project (such as budget preparation, execution progress, and contract performance status); the benchmark value is the recommended fund allocation value output by the quantile model, reflecting a reasonable allocation scheme under different risk preferences.

[0121] For example, historical data (such as budget execution records for the past three years) of a project is extracted from the integrated fiscal platform. The data is then filtered based on a preset time range (such as 2021-2023) to ensure time continuity. To facilitate the demonstration of the application process, a simple tabular data example is shown below:

[0122] The single-item dataset is input into four quantile models to calculate the baseline value. The output results include:

[0123] First baseline value (Q0.2): The conservative allocation value predicted by the model (e.g., 5.5 million yuan).

[0124] Second baseline value (Q0.5): The balanced allocation value predicted by the model (e.g., 6 million yuan).

[0125] Third baseline value (Q0.7): The aggressive allocation value predicted by the model (e.g., 6.5 million yuan);

[0126] Fourth baseline value (Q0.9): The high-risk allocation value predicted by the model (e.g., 7 million yuan).

[0127] The calculation formula is Where q is the quantile (0.2, 0.5, 0.7, 0.9). This is the trained model for the corresponding quantile.

[0128] Four benchmark values ​​are associated with project numbers and stored in the "Fund Allocation Benchmark Library" of the database. When making fund allocation decisions, the benchmark value of the corresponding quantile is selected according to the project risk level. The quantile model of this application is not sensitive to outliers, the prediction results are more stable, and the four quantiles cover all scenarios from conservative to aggressive, meeting the needs of different projects. Thus, the integrated fiscal platform realizes the generation of multi-dimensional and dynamic fund allocation benchmark values.

[0129] S30: Construct a fund allocation efficiency scoring formula through several fund allocation benchmark values, and determine the fund allocation efficiency score of fiscal business based on the fund allocation efficiency scoring formula and real-time fiscal business data; determine whether to trigger the early warning mechanism based on the fund allocation efficiency score.

[0130] In this embodiment, the fund allocation efficiency score is a quantitative indicator that comprehensively evaluates the rationality of fund allocation and the matching degree of execution progress (e.g., the score range is 0~100); the early warning mechanism refers to triggering an alarm when the score is lower than a preset threshold (e.g., 60 points), prompting manual intervention.

[0131] Specifically, the formula for scoring the efficiency of fund allocation includes:

[0132] in, Assign a score to the efficiency of fund allocation; The actual fund allocation value is the allocation amount in real-time business data; , , , The quantile benchmarks are, in order, the extreme values ​​of the aggressive strategy, the upper limit of the aggressive strategy, the median of the balanced strategy, and the lower limit of the conservative strategy; 20, 15, 10, and 5 are penalty coefficients used to determine the severity of penalties for allocations exceeding the benchmark values. The higher the quantile (the greater the risk), the larger the penalty coefficient, reflecting risk sensitivity. This calculation formula imposes the most severe penalty (coefficient 20) on high-risk allocations (0.9 quantile) to strengthen fiscal discipline; and only a slight penalty (coefficient 5) is imposed on low-risk allocations (0.2 quantile) to avoid excessively restricting grassroots flexibility.

[0133] Furthermore, compliance factors can be added outside the formula. For example, the final fund allocation efficiency score = When there are over-budget payments, assuming =0.7 (a direct reduction of 30%).

[0134] Furthermore, the application scenarios and response actions of the formula include the following situations:

[0135] ① In the scenario of aggressive allocation, the actual allocation value The scoring range is ≤80 points, and the response action is a red alert, which suspends payment and initiates an audit.

[0136] ② In aggressive allocation scenarios, the actual allocation value The scoring range is 80 to 90 points; the response action is a yellow warning, and the matter is submitted to the supervisor for review.

[0137] ③ In a balanced allocation scenario, the actual allocation value The scoring range is 90 to 100 points; the response action is executed normally and no intervention is required.

[0138] ④ Conservative allocation scenario, actual allocation value The scoring range is 100 points. A green response indicates approval, and additional budget adjustments can be requested.

[0139] In one embodiment, the process of training a fund allocation optimization model using a sample dataset includes:

[0140] The fund allocation optimization model is trained based on a sample dataset. The model is solved by minimizing an asymmetric absolute value residual loss function to fit the changing trends of fund allocation parameters. Specifically, a higher penalty weight (e.g., coefficient) is assigned to deviations where the actual fund allocation value exceeds the budget benchmark value. The deviation of the actual fund allocation value from the budget benchmark value is given a lower penalty weight (such as a coefficient). ), and satisfy This reflects the business characteristic of prioritizing the management of overspending risks in the allocation of fiscal funds.

[0141] Furthermore, this application also divides dynamic threshold ranges based on a fund allocation efficiency score, setting at least three health status levels:

[0142] High-efficiency zone (score ≥ 90): Fund allocation is highly aligned with policy objectives;

[0143] Warning zone (70 points ≤ score < 90 points): Risk of allocation deviation exists;

[0144] Crisis Zone (Score < 70): Funding allocation is severely unbalanced or suspected of being irregular.

[0145] The triggering conditions include: if the fund allocation efficiency score falls into the warning zone or crisis zone, and meets any of the following conditions, a tiered alarm will be automatically triggered:

[0146] Condition 1: The score continues to decline for N consecutive periods;

[0147] Condition 2: The score is below the threshold and is accompanied by related risk events such as abnormal contract performance and budget execution delays.

[0148] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0149] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0150] S1: Establish a unified project management model based on the needs of fiscal operations, define project attributes through a hierarchical structure, and link each project node with information on the entire lifecycle of budget preparation, execution, and performance evaluation;

[0151] S2: Based on the hierarchical relationship and funding requirements of projects in the project management model, generate a draft departmental budget according to the preset allocation rules, and use the dynamic adjustment mechanism of the target indicators to link budget preparation with project execution progress;

[0152] S3: Decompose the approved budget results into actionable indicators at each level, calculate the available funds in combination with the project progress, and lock, adjust and recover the actionable indicators according to the project dimension.

[0153] S4: Obtain the fund application information of the budget unit, verify the payment conditions based on the project budget balance and contract performance status, generate direct payment vouchers or authorized payment limits, and update the project fund flow.

[0154] Those skilled in the art will understand that all or part of the processes in the methods of 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 embodiments of the above 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.

[0155] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0156] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A data processing method applied to an integrated fiscal platform, characterized in that, include: A unified project management model is established based on the needs of fiscal operations. Project attributes are defined through a hierarchical structure, and each project node is associated with information throughout the entire lifecycle of budget preparation, execution, and performance evaluation. Based on the hierarchical relationship and funding requirements of projects in the project management model, a draft departmental budget is generated according to the preset allocation rules, and the budget preparation and project execution progress are linked by the dynamic adjustment mechanism of the target indicators. The approved budget results are broken down into actionable indicators at each level. The available funds are calculated in conjunction with the project progress. The actionable indicators are locked, adjusted, and recovered according to the project dimension. Obtain information on fund application from budget units, verify payment conditions based on project budget balance and contract performance status, generate direct payment vouchers or authorized payment limits, and update project cash flow. The method further includes: A sample dataset is constructed based on fiscal business data within a preset time range. The constructed fund allocation optimization model is trained using the sample dataset, and the parameters of the fund allocation optimization model are adjusted to obtain the trained fund allocation optimization model. The trained fund allocation optimization model is used to determine several target fund allocation models corresponding to several preset quantiles, and the target fund allocation models are used to determine several fund allocation benchmark values ​​corresponding to the several preset quantiles. A fund allocation efficiency scoring formula is constructed using the aforementioned several fund allocation benchmark values. The fund allocation efficiency scoring formula and real-time fiscal business data are used to determine the fund allocation efficiency score of fiscal business. The fund allocation efficiency score is used to determine whether to trigger an early warning mechanism. The step of using the trained capital allocation optimization model to determine several target capital allocation models corresponding to several preset quantiles, and then using these target capital allocation models to determine several capital allocation benchmark values ​​corresponding to the several preset quantiles, includes: The quantiles of the trained fund allocation optimization model are set to 0.2, 0.5, 0.7 and 0.9 respectively, and the fund allocation optimization models corresponding to the 0.2, 0.5, 0.7 and 0.9 quantiles are trained using the sample dataset to obtain the first target fund allocation model corresponding to the 0.2 quantile, the second target fund allocation model corresponding to the 0.5 quantile, the third target fund allocation model corresponding to the 0.7 quantile, and the fourth target fund allocation model corresponding to the 0.9 quantile. A single-project dataset is constructed based on single-project data within a preset second time range, and the single-project dataset is input into the first target fund allocation model, the second target fund allocation model, the third target fund allocation model, and the fourth target fund allocation model respectively to obtain the corresponding first benchmark value, second benchmark value, third benchmark value, and fourth benchmark value.

2. The data processing method applied to an integrated fiscal platform according to claim 1, characterized in that, The integrated fiscal platform includes: The project library construction module establishes a unified project management model based on the needs of fiscal operations. It defines project attributes through a hierarchical structure, and links each project node with information on the entire lifecycle of budget preparation, execution, and performance evaluation. The budget generation module generates a departmental budget draft based on the hierarchical relationship and funding requirements of projects in the project management model and according to preset allocation rules. It also uses a dynamic adjustment mechanism for the corresponding indicators to link budget preparation with project execution progress. The indicator dynamic management module decomposes the approved budget results into actionable indicators at each level, calculates the available funds based on the project progress, and performs indicator locking, adjustment, and recovery operations on the actionable indicators according to the project dimension. The payment control module obtains the fund application information of the budget unit, verifies the payment conditions based on the project budget balance and contract performance status, generates direct payment vouchers or authorized payment limits, and updates the project fund flow. The emergency and data maintenance module provides stand-alone emergency payment functionality, generating emergency payment instructions locally when the network is interrupted; it uses version stamp technology to maintain cross-year project data, enabling automatic association of budget execution data between historical and new projects.

3. The data processing method applied to an integrated fiscal platform according to claim 2, characterized in that, It also includes a data interface subsystem, which comprises: The data exchange module, based on a predefined fiscal business data model, achieves bidirectional data synchronization with designated key systems through a Web Service interface, and supports conversion of different data formats. The log monitoring module uses a distributed log framework to record the timestamps, operator IDs, and data verification results of all data exchange operations in real time, and supports multi-dimensional retrieval operations based on business type and anomaly level. The network security isolation module deploys a physical gateway to achieve physical isolation between the internal and external networks of the finance and government affairs system, and transmits data through encrypted channels using the national cryptographic SM4 algorithm; The version compatibility adaptation layer provides a unified API gateway and a built-in version manager, supporting dynamic adaptation of interface protocols between new and old systems.

4. The data processing method applied to an integrated fiscal platform according to claim 2, characterized in that, The emergency response and data maintenance module implements stand-alone emergency payment functionality through a stand-alone emergency subsystem; the platform also includes: The distributed service mesh uses a microservice framework to decompose the core business modules of the integrated finance platform, shortening the response time of service calls between various core business modules to a specified time range, and supporting circuit breaking, degradation, and rate limiting mechanisms. The intelligent routing engine dynamically allocates multiple payment request nodes based on the consistent hashing algorithm, combines machine learning to predict the load pressure of each core business module, obtains the application load pressure prediction result, and performs node weight adjustment operation based on the application load pressure prediction result. The time-series database establishes composite indexes for frequently queried fields and adopts an LSM tree storage structure; it uses open-source container orchestration components to achieve containerized deployment, automatically scaling up Pod nodes when the CPU utilization of the integrated financial platform exceeds a preset CPU utilization threshold, and controlling the cold start time to be less than a preset cold start duration threshold.

5. The data processing method applied to an integrated fiscal platform according to claim 2 or 4 further includes: Based on the needs of fiscal business, the query condition combination rules are defined, supporting free combination of multi-dimensional parameters and generating dynamic query templates; The multi-dimensional parameters include project attributes, funding sources, timeframe, and responsible entity; The entire lifecycle information and the budget execution data are dynamically linked to enable penetrating query capabilities from business process nodes such as budget preparation, indicator decomposition, payment progress, and contract performance, so as to support in-depth traceability query of cross-year projects; Online analytical processing (OLAP) algorithms are used to perform multidimensional analysis on query results to generate standardized visual reports.

6. The data processing method applied to an integrated fiscal platform according to claim 2, characterized in that, Also includes: A role-permission matrix is ​​established based on the RBAC model. The role-permission matrix is ​​configured with access permissions according to the dimensions of business department, funding nature, and project type. Configure the business process model and corresponding business process nodes in the project management model, and trigger dynamic adjustment of approval nodes, node return for re-review, and reminder information based on the business process nodes; In the project management model, operation and maintenance monitoring index parameters are collected in real time, and preset threshold alarm rules are associated with these operation and maintenance monitoring index parameters. When the preset threshold alarm rules meet the alarm conditions, alarm prompt instructions are triggered.

7. The data processing method applied to an integrated fiscal platform according to claim 1, characterized in that, The process of constructing a sample dataset based on fiscal business data within a preset time range, training a fund allocation optimization model using the sample dataset, and adjusting the parameters of the fund allocation optimization model to obtain a trained fund allocation optimization model includes: The quantile parameter of the fund allocation optimization model is set to 0.5, and the fund allocation optimization model is trained using the sample dataset to obtain the optimization model to be determined. The degree of fit between the optimization model to be determined and the fund allocation trend is determined based on the coefficient of determination, so as to determine whether the model accuracy of the optimization model to be determined reaches the preset model accuracy threshold based on the degree of fit. If the model accuracy of the model to be optimized does not reach the preset model accuracy threshold, the parameters of the model to be optimized are adjusted and retrained until the model accuracy of the model to be optimized reaches the preset model accuracy threshold, so as to determine the model to be optimized as the post-trained fund allocation optimization model.

8. The data processing method applied to an integrated fiscal platform according to claim 1, characterized in that, The sample dataset constructed based on fiscal business data within a preset time range includes: The unified data collection interface is used to traverse several sets of fiscal business data within a preset time range on the integrated fiscal platform. Based on the timestamp, the several sets of fiscal business data are synchronously calibrated, and missing values ​​and outliers in the several sets of fiscal business data are removed to obtain several sets of preprocessed fiscal business data. Remove the project number information from the preprocessed fiscal business data, summarize the preprocessed fiscal business data to obtain summarized fiscal business data, and construct a sample dataset based on the summarized fiscal business data.

Images