Fee difference analysis method and system based on two-stage penetration processing

Abstract

The invention discloses an intelligent report generation method and system based on two-stage penetration processing, and relates to the crossing field of insurance finance and artificial intelligence. According to the method, through a collaborative architecture of a non-intelligent calculation tool, a multi-system interface and a rule adjustment large model, firstly, the non-intelligent calculation tool automatically calculates multi-source cost data based on a preset actuarial template, and outputs standardized cost difference integration data; pushing to a large model which is subjected to prompt engineering and small sample learning adjustment and calibration through a RESTful API (Application Program Interface) protocol; and finally, the large model executes multi-dimensional penetration analysis according to an insurance financial fee difference business rule knowledge base, accurately locates an abnormal fee difference cause and generates a structured report containing a natural language conclusion. According to the method, the problems of high data processing time consumption, insufficient penetration dimension and manual report interpretation in the prior art are solved, and the cost difference analysis efficiency and precision are greatly improved.

Description

Technical Field

[0001] This invention relates to the fields of insurance finance and artificial intelligence technology, and in particular to a cost difference analysis method and system based on two-stage penetration processing. Background Technology

[0002] In the process of digital transformation in the insurance industry, expense variance analysis, as a core component for measuring institutional operating efficiency and optimizing resource allocation, directly impacts the quality of management decision-making and the long-term development of the enterprise. With the expansion of insurance business, diversification of channels, and increasing complexity of expense structures, expense variance analysis needs to integrate multi-source data from business systems, financial systems, and other sources to complete complex tasks such as calculating expense variance achievement rates, tracing causes, and multi-dimensional breakdown. Traditional analysis methods relying on manual operation are no longer adequate for the industry's development needs, necessitating intelligent technology solutions to improve analytical efficiency and depth.

[0003] To address this challenge, existing technologies generally employ a combination of non-intelligent computing tools and human-assisted analysis. This involves using customized Excel plugins, localized calculation scripts, and other non-intelligent tools to process financial indicators, followed by staff combining their business experience to piece together data from multiple systems, trace the causes of expense discrepancies, and manually compile and generate analysis reports. This achieves the initial integration and analysis of expense discrepancy data.

[0004] However, the existing technology has a core flaw: the data splicing and processing of multiple systems is extremely time-consuming, the cost difference penetration analysis lacks sufficient dimensions and the root cause is not accurately located, and the report generation relies on manual interpretation. This results in problems such as poor computational collaboration, limited penetration depth, and insufficient intelligence in the overall analysis process, making it unable to quickly and accurately meet the insurance institutions' needs for efficient and refined cost difference analysis. Summary of the Invention

[0005] This invention provides a cost difference analysis method and system based on two-stage penetration processing, which can solve the problems of high time consumption in multi-system data splicing processing, insufficient cost difference penetration dimensions, and the need for manual interpretation. It can effectively improve the efficiency, depth and intelligence level of insurance financial cost difference analysis.

[0006] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions: Firstly, a two-stage penetration processing-based expense difference analysis method is provided, comprising: acquiring multi-source expense-related data from the insurance institution's business and financial systems, including budgeted expense data, actual expense data, and premium income data; performing indicator calculations on the multi-source expense-related data using non-intelligent tools to obtain standardized expense difference integrated data, which includes budgeted expense, expense difference achievement rate, and total expense difference achievement rate fields for each sub-item under each of the three dimensions of institution, channel, and expense item; encapsulating the standardized expense difference integrated data in JSON format using the RESTful API protocol, and pushing the encapsulated standardized expense difference integrated data to a preset rule model, which is obtained through prompting engineering and small sample learning; based on an insurance financial expense difference business rule knowledge base containing expense difference penetration dimension definitions and abnormal expense difference identification rules, performing penetration analysis on the standardized expense difference integrated data for each sub-item under each dimension through the preset rule model, identifying the causes of abnormal expense differences and calculating the impact weights, and generating expense difference analysis results, which include data statements, root cause interpretations, and solutions.

[0007] The method provided by this invention acquires multi-source data such as budgeted expenses, actual expenses, and premium income from the business and financial systems of insurance institutions. It uses non-intelligent tools to perform indicator calculations to obtain standardized expense difference integrated data containing three-dimensional sub-items of institution, channel, and expense items. Then, it pushes the data to a pre-set rule-based large model through a RESTful API protocol, which has been calibrated through prompting engineering and small-sample learning. Finally, it completes multi-dimensional penetrating analysis based on a business rule knowledge base and generates results including data statements, root cause interpretations, and solutions. This eliminates the need for manual splicing of data from multiple systems and manual tracing of causes, effectively solving the problems of high data processing time, insufficient penetrating dimensions, and the need for manual interpretation of reports in existing technologies. It achieves automation, intelligence, and accuracy in expense difference analysis.

[0008] In one possible implementation of the first aspect, each of the three dimensions of institution, channel, and expense item includes at least two sub-items; when the dimension is institution, the sub-items include different branches or subsidiaries of the insurance institution; when the dimension is channel, the sub-items include offline agency channels, online promotion channels, or bancassurance cooperation channels; when the dimension is expense item, the sub-items include premiums, human resource costs, daily expenses, or workplace expenses.

[0009] The method provided by this invention defines the sub-item types of three dimensions: organization, channel, and expense item. Each dimension contains at least two sub-items, making the dimensional division of standardized expense difference integration data more specific and clear. This provides accurate sub-data support for the subsequent penetrating analysis of the sub-items under each dimension in the execution of the large model with preset rules. It ensures that the cause of abnormal expense differences can be traced back to specific branch organizations, channel types, or expense items, avoiding the problem of inaccurate root cause location due to fuzzy dimensions.

[0010] In one possible implementation of the first aspect, the step of performing indicator calculations on the multi-source cost-related data using non-intelligent tools to obtain standardized cost difference integration data includes: mapping the multi-source cost-related data to obtain a dimensional data set, wherein the dimensional data set includes budgeted cost data, actual cost data, and premium income data corresponding to each sub-item under each dimension; automatically performing indicator calculations on the dimensional data set using non-intelligent tools according to a preset actuarial template to obtain the total cost difference amount, the total cost difference achievement rate, and the cost difference amount and cost difference contribution corresponding to each sub-item under each dimension; in the same When the sum of the cost difference amounts of all sub-items under one dimension equals the total cost difference amount, the budget cost data, total cost difference achievement rate, and cost difference achievement rate of each sub-item under each dimension are extracted to generate the standardized cost difference integrated data. Here, the total cost difference amount is the difference between the total budget cost data and the total actual cost data of all sub-items; the cost difference amount of each sub-item is the difference between the budget cost data and the actual cost data of each sub-item; the total cost difference achievement rate is the ratio of the actual cost difference amount to the planned cost difference amount; and the cost difference contribution of each sub-item is the ratio of the cost difference amount of each sub-item to the total cost difference amount.

[0011] The method provided by this invention forms a dimensional data set by mapping multi-source data together, and automatically performs indicator calculations by non-intelligent tools using a pre-set actuarial template. It clarifies the specific calculation formulas for indicators such as total cost difference amount and cost difference contribution, and ensures data accuracy by summing and verifying the cost difference amounts of sub-items under the same dimension. This avoids the errors and inefficiencies of manual calculations and ensures the calculation standardization and reliability of standardized cost difference integrated data, providing high-quality basic data for subsequent penetrating analysis.

[0012] In one possible implementation of the first aspect, the step of encapsulating the standardized fee difference integration data in JSON format using the RESTful API protocol and pushing the encapsulated standardized fee difference integration data to the preset rule big model includes: encapsulating the standardized fee difference integration data in JSON format using the RESTful API protocol, initiating transmission to the preset rule big model at a batch processing frequency of T+1 months, where T is the statistical base month; in the event that the standardized fee difference integration data transmission fails, retrying sequentially at preset time intervals; and triggering an exception alarm and recording the transmission log if the number of retries reaches a preset value and the transmission still fails.

[0013] The method provided by this invention uses the RESTful API protocol to encapsulate data in JSON format and pushes it in batches every T+1 months, which is in line with the business habits of insurance institutions to calculate expense differences on a monthly basis. At the same time, it sets up a retry mechanism and alarm recording function after transmission failure, which can effectively reduce the impact of data transmission interruption, ensure that the standardized expense difference integration data is pushed to the preset rule big model stably and in a timely manner, and ensure the continuity and timeliness of the expense difference analysis process.

[0014] In one possible implementation of the first aspect, the method further includes: when the preset rule big model identifies that the standardized fee difference integration data has missing key indicators, insufficient data accuracy, or incomplete dimension coverage, it generates a supplementary calculation instruction, the supplementary calculation instruction including the type of missing indicator, data accuracy requirements, and dimension supplementation range; the non-intelligent tool responds to the supplementary calculation instruction, obtains multi-source expense-related data from the insurance institution's business system and financial system, re-executes indicator calculation, and obtains updated standardized fee difference integration data; the updated standardized fee difference integration data is encapsulated in JSON format using the RESTful API protocol, and the encapsulated updated standardized fee difference integration data is pushed to the preset rule big model.

[0015] The method provided by this invention actively identifies defects in standardized cost difference integrated data through a large model with preset rules, generates supplementary calculation instructions with specific requirements, and then non-intelligent tools respond and re-acquire data to perform calculations, forming a closed loop of data verification, supplementary calculation, and re-push. This ensures that the data pushed to the large model is complete and accurate, avoids distortion of analysis results caused by missing data, insufficient accuracy, or incomplete dimensions, and further improves the reliability of cost difference analysis.

[0016] In one possible implementation of the first aspect, the step of using an insurance financial expense difference business rule knowledge base containing expense difference penetration dimension definitions and abnormal expense difference identification rules to perform penetration analysis on the standardized expense difference integrated data under each dimension through the preset rule big model, identifying the causes of abnormal expense differences and calculating their impact weights, and generating expense difference analysis results includes: calling the expense difference penetration dimension definition rules in the insurance financial expense difference business rule knowledge base through the preset rule big model to construct a penetration analysis framework, which includes multiple levels such as total dimension, single dimension, sub-item, and cross dimension; comparing the expense difference achievement rate and expense difference contribution of each sub-item with a preset threshold according to the abnormal expense difference identification rules in the knowledge base, marking abnormal sub-items that exceed the threshold range, and determining the abnormal sub-item. The degree of abnormal deviation is analyzed; a multi-dimensional penetrating analysis is performed on the abnormal sub-items to obtain the logic of expense difference changes, the synergistic influence relationship, the expense difference fluctuation pattern of continuous statistical periods, and the direct, indirect, and root causes of abnormal expense differences; based on the causal association rules in the knowledge base, a weighted scoring method is used to calculate the influence weight of each root cause, and root causes with an influence weight exceeding a preset proportion are identified as core root causes; expense difference analysis results are generated, which include data statements, root cause interpretations, and solutions. The data statements are used to present expense difference-related data and abnormal indicators of each dimension and sub-item in a hierarchical manner; the root cause interpretations are used to quantify the influence weight of core root causes and explain their causes; the solutions are to match the optimal practice solutions in the knowledge base and optimize the execution steps and expected effects in combination with the business scenarios of insurance institutions.

[0017] The method provided by this invention constructs a multi-level penetrating analysis framework by calling a business rule knowledge base through a pre-set rule-based large model. It marks abnormal sub-items by comparing thresholds, deeply deconstructs the logic, synergistic effects, and causes of expense difference changes, and uses a weighted scoring method to quantify the root cause weights and generate structured results. This achieves comprehensive deconstruction and accurate root cause location from the total dimension to the cross-dimensional dimension, making the analysis results both hierarchical and able to directly guide practice without the need for additional manual sorting, thus improving the professionalism and practicality of expense difference analysis.

[0018] In one possible implementation of the first aspect, the method further includes: importing core rules from the insurance financial expense difference business rule knowledge base into the basic model, so that the basic model can establish the business boundaries and core judgment criteria for expense difference analysis; constructing a dedicated template for insurance financial expense difference analysis through prompting engineering, wherein the dedicated template for insurance financial expense difference analysis is used to adjust the analysis logic framework and output format specifications of the basic model; and inputting multiple sets of historical expense difference analysis examples into the basic model using a few-sample learning method, wherein each set of historical expense difference analysis examples includes corresponding input data and standard output conclusions, so that the basic model can master the business logic and expression specifications of expense difference analysis and obtain the preset rule big model.

[0019] The method provided by this invention imports the core rules of the insurance financial expense difference business rule knowledge base into the basic model, constructs a dedicated template and inputs multiple sets of historical analysis examples, so that the basic model can gradually establish business boundaries, master analysis logic and expression norms, and finally form a large-scale model of preset rules adapted to the insurance financial expense difference analysis scenario. This solves the problem of traditional large-scale models lacking financial analysis professionalism and ensures that the large-scale model can output professional analysis results that meet business needs.

[0020] Secondly, this invention provides a expense difference analysis system based on two-stage penetration processing, comprising: an acquisition module for acquiring multi-source expense-related data from the insurance institution's business system and financial system, wherein the multi-source expense-related data includes budget expense data, actual expense data, and premium income data; a calculation module for performing indicator calculations on the multi-source expense-related data using non-intelligent tools to obtain standardized expense difference integrated data, wherein the standardized expense difference integrated data includes budget expense, expense difference achievement rate, and total expense difference achievement rate fields corresponding to each sub-item under each of the three dimensions of institution, channel, and expense item; and a push module for using RESTful... The API protocol encapsulates the standardized expense difference integration data in JSON format and pushes the encapsulated standardized expense difference integration data to a preset rule model. The preset rule model is obtained through prompting engineering and small sample learning. The analysis module is used to perform penetration analysis on the standardized expense difference integration data based on the insurance financial expense difference business rule knowledge base, which includes expense difference penetration dimension definitions and abnormal expense difference identification rules. The module identifies the causes of abnormal expense differences and calculates the impact weights, generating expense difference analysis results. The expense difference analysis results include data statements, root cause interpretations, and solutions.

[0021] Thirdly, an electronic device is provided, the electronic device including a memory and one or more processors; the memory is coupled to the processors; wherein the memory stores computer program code, the computer program code including computer instructions, which, when executed by the processor, cause the electronic device to perform the method as described in any implementation of the first aspect.

[0022] Fourthly, a computer-readable storage medium is provided, including computer instructions that, when executed on an electronic device, cause the electronic device to perform a method as described in any implementation of the first aspect.

[0023] Fifthly, a computer program product is provided that, when run on a computer, causes the computer to perform the method in any implementation of the first aspect.

[0024] Understandably, the beneficial effects achieved by the system of the second aspect, the electronic device of the third aspect, the computer-readable storage medium of the fourth aspect, and the computer program product of the fifth aspect provided above can be referred to with reference to the beneficial effects of the first aspect and any of its possible design embodiments, which will not be repeated here. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention; Figure 2 A flowchart of a cost difference analysis method based on two-stage penetration processing provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a cost difference analysis system provided in an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be described below with reference to the accompanying drawings. In the description of the present invention, unless otherwise stated, " / " indicates that the objects before and after are in an "or" relationship. For example, A / B can represent A or B. The "or" in the present invention is merely a description of the relationship between the related objects, indicating that three relationships can exist. For example, A or B can represent: A alone, A and B simultaneously, and B alone. A and B can be singular or plural. Furthermore, in the description of the present invention, unless otherwise stated, "multiple" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items.

[0027] Furthermore, to facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0028] In this embodiment of the invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in this embodiment of the invention should not be construed as superior or more advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner for ease of understanding.

[0029] In the process of digital transformation in the insurance industry, expense variance analysis, as a core component for measuring institutional operating efficiency and optimizing resource allocation, directly impacts the quality of management decision-making and the long-term development of the enterprise. With the expansion of insurance business, diversification of channels, and increasing complexity of expense structures, expense variance analysis needs to integrate multi-source data from business systems, financial systems, and other sources to complete complex tasks such as calculating expense variance achievement rates, tracing causes, and multi-dimensional breakdown. Traditional analysis methods relying on manual operation are no longer adequate for the industry's development needs, necessitating intelligent technology solutions to improve analytical efficiency and depth.

[0030] To address this challenge, existing technologies generally employ a combination of non-intelligent computing tools and human-assisted analysis. This involves using customized Excel plugins, localized calculation scripts, and other non-intelligent tools to process financial indicators, followed by staff combining their business experience to piece together data from multiple systems, trace the causes of expense discrepancies, and manually compile and generate analysis reports. This achieves the initial integration and analysis of expense discrepancy data.

[0031] However, the existing technology has a core flaw: the data splicing and processing of multiple systems is extremely time-consuming, the cost difference penetration analysis lacks sufficient dimensions and the root cause is not accurately located, and the report generation relies on manual interpretation. This results in problems such as poor computational collaboration, limited penetration depth, and insufficient intelligence in the overall analysis process, making it unable to quickly and accurately meet the insurance institutions' needs for efficient and refined cost difference analysis.

[0032] In view of this, embodiments of the present invention provide a method and system for expense difference analysis based on two-stage penetration processing, comprising: acquiring multi-source expense-related data from the business system and financial system of an insurance institution, wherein the multi-source expense-related data includes budget expense data, actual expense data, and premium income data; performing indicator calculations on the multi-source expense-related data using non-intelligent tools to obtain standardized expense difference integrated data, wherein the standardized expense difference integrated data includes budget expense, expense difference achievement rate, and total expense difference achievement rate fields corresponding to each sub-item under each of the three dimensions of institution, channel, and expense item; encapsulating the standardized expense difference integrated data in JSON format using the RESTful API protocol, and pushing the encapsulated standardized expense difference integrated data to a preset rule model, wherein the preset rule model is obtained through prompting engineering and small sample learning; based on an insurance financial expense difference business rule knowledge base containing expense difference penetration dimension definitions and abnormal expense difference identification rules, performing penetration analysis on the standardized expense difference integrated data corresponding to the sub-items under each dimension through the preset rule model, identifying the causes of abnormal expense differences and calculating the impact weights, and generating expense difference analysis results, wherein the expense difference analysis results include data statements, root cause interpretations, and solutions.

[0033] The method provided by this invention acquires multi-source data such as budgeted expenses, actual expenses, and premium income from the business and financial systems of insurance institutions. It uses non-intelligent tools to perform indicator calculations to obtain standardized expense difference integrated data containing three-dimensional sub-items of institution, channel, and expense items. Then, it pushes the data to a pre-set rule-based large model through a RESTful API protocol, which has been calibrated through prompting engineering and small-sample learning. Finally, it completes multi-dimensional penetrating analysis based on a business rule knowledge base and generates results including data statements, root cause interpretations, and solutions. This eliminates the need for manual splicing of data from multiple systems and manual tracing of causes, effectively solving the problems of high data processing time, insufficient penetrating dimensions, and the need for manual interpretation of reports in existing technologies. It achieves automation, intelligence, and accuracy in expense difference analysis.

[0034] In some embodiments, the cost difference analysis method based on two-stage penetration processing provided by the present invention can be executed by a cost difference analysis system 100 based on two-stage penetration processing (hereinafter referred to as cost difference analysis system 100).

[0035] As an example, the cost difference analysis system 100 can be any electronic device 200 with data processing capabilities, such as a general-purpose computer, personal computer, laptop computer, switch, or tablet computer. The specific implementation of the cost difference analysis system 100 is not limited here.

[0036] Figure 1 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention is shown. The electronic device 200 includes a processor 210, a memory 220, and a communication interface 230.

[0037] Processor 210 may include one or more processing cores. Processor 210 connects to various parts within electronic device 200 using various interfaces and lines, and performs various functions and processes data of electronic device 200 by running or executing instructions, programs, code sets, or instruction sets stored in memory 220, and by calling data stored in memory 220. Optionally, processor 210 may be implemented using at least one of the following hardware forms: Central Processing Unit (CPU), Graphics Processing Unit (GPU), Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA).

[0038] The memory 220 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 220 may include a non-transitory computer-readable storage medium. The memory 220 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 220 may include a program storage area. This program storage area may store instructions for implementing an operating system, instructions for implementing at least one function, instructions for implementing the various method embodiments described above, etc.

[0039] Communication interface 230 is used to communicate with other devices, equipment or communication networks, such as data storage devices, image processing devices or Ethernet, wireless access network (RAN), wireless local area network (WLAN), etc.

[0040] In terms of physical implementation, the aforementioned devices (such as processor 210, memory 220, and communication interface 230) can each be devices within the same device (such as a laptop computer). Alternatively, at least two of these devices can be located within the same device, i.e., as different devices within the same device, similar to the deployment of devices or components in a distributed system.

[0041] It is understood that the structure illustrated in this embodiment does not constitute a specific limitation on the electronic device 200. In other embodiments of the present invention, the electronic device 200 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0042] The following description, in conjunction with the accompanying drawings, illustrates a cost difference analysis method based on two-stage penetration processing provided by an embodiment of the present invention.

[0043] Figure 2 This is a flowchart illustrating a cost difference analysis method based on two-stage penetration processing, provided as an embodiment of the present invention. Optionally, this method can be... Figure 1 The illustrated electronic device 200 performs this method, which includes the following steps: S1. Obtain multi-source expense-related data from the insurance institution's business and financial systems.

[0044] Specifically, the multi-source cost-related data includes budgeted cost data, actual cost data, and premium income data; In one possible implementation, each of the three dimensions—organization, channel, and expense item—includes at least two sub-items; when the dimension is organization, the sub-items include different branches or subsidiaries of the insurance institution; when the dimension is channel, the sub-items include offline agency channels, online promotion channels, or bancassurance cooperation channels; when the dimension is expense item, the sub-items include premiums, human resource costs, daily expenses, or workplace expenses.

[0045] It should be noted that each expense item in the categories of premium, personnel costs, and daily or workplace expenses includes both a planned portion and an actual portion.

[0046] Specifically, branches can be regional operating outlets established by insurance institutions in different provinces and cities, such as a branch office of an insurance company in Shanghai or a central branch office in Hangzhou; subsidiaries refer to independent legal entities wholly owned or controlled by an insurance group, focusing on specific insurance business areas or regional markets; business units can be independent operating modules within an insurance company divided according to business type, such as individual insurance business units and group insurance business units. Offline agency channels can specifically include individual agent teams that have signed cooperation agreements with insurance institutions, professional insurance agency companies, etc.; online promotion channels include the insurance company's official APP, WeChat official account, third-party insurance e-commerce platforms, etc.; bancassurance cooperation channels refer to cooperative relationships with banks, selling insurance products through bank branches, mobile banking, and other channels. The expense items include premiums, which correspond to the expenses related to premium income collected by insurance institutions through various channels; human resources expenses, which include employee salaries, benefits, training expenses, and other human resources-related expenditures; daily expenses, which cover daily operating costs such as office supplies procurement, water and electricity bills, and communication fees; and workplace expenses, which include expenses related to the business premises such as office space rent, decoration and maintenance fees, and property management fees. The planned portion of each expense item is the pre-determined budget amount, while the actual portion is the actual settlement amount after the expense is incurred. Together, they constitute the complete data dimension of the expense item.

[0047] This invention, by clearly defining specific sub-items across three dimensions, allows previously vague expense difference data to be organized according to clear classification standards. Each sub-item becomes a concrete tool for expense difference analysis, enabling subsequent large-scale modeling based on pre-defined rules to accurately pinpoint the analysis object without needing to search within ambiguous dimensions. This explicit segmentation method helps the large-scale model quickly locate the scope of abnormal expense differences, directly linking them to specific operating entities, sales channels, or expense types. This significantly reduces the time cost of root cause tracing, making the analysis results more targeted and actionable, and providing strong support for insurance institutions to formulate precise expense control strategies.

[0048] S2. Perform index calculations on the multi-source cost-related data using non-intelligent tools to obtain standardized cost difference integrated data.

[0049] The standardized cost difference integration data includes the fields of budget cost, cost difference achievement rate, and total cost difference achievement rate for each sub-item under each of the three dimensions: institution, channel, and cost item.

[0050] In some embodiments, S2 specifically includes: mapping multi-source cost-related data to obtain a dimensional data set, wherein the dimensional data set includes budgeted cost data, actual cost data, and premium income data corresponding to each sub-item under each dimension; automatically performing indicator calculations on the dimensional data set using non-intelligent tools based on a preset actuarial template to obtain the total cost difference amount, the total cost difference achievement rate, and the cost difference amount and cost difference contribution rate corresponding to each sub-item under each dimension; when the sum of the cost difference amounts of all sub-items under the same dimension equals the total cost difference amount, extracting the budgeted cost data, the total cost difference achievement rate, and the cost difference achievement rate corresponding to each sub-item under each dimension to generate the standardized cost difference integrated data; wherein the total cost difference amount is the difference between the total budgeted cost data and the total actual cost data of all sub-items, the cost difference amount of each sub-item is the difference between the budgeted cost data and the actual cost data of each sub-item, the total cost difference achievement rate is the ratio of the actual cost difference amount to the planned cost difference amount, and the cost difference contribution rate of each sub-item is the ratio of the cost difference amount of each sub-item to the total cost difference amount.

[0051] It should be noted that non-intelligent tools can be customized Excel plugins, localized calculation scripts, or traditional financial calculation systems. These tools only execute preset indicator organization logic and calculation rules and do not have the ability to learn or make decisions independently.

[0052] Furthermore, the pre-built actuarial templates comply with the requirements of the "Guidelines for Expense Allocation of Insurance Companies," and include multiple standardized calculation formulas that can accurately adapt to the business specifications for calculating insurance financial expense differences. The correlation mapping process involves matching data from different sources, such as policy-related expense data from the insurance institution's business system and revenue and expenditure accounting data from the financial system, according to three dimensions and their sub-items: institution, channel, and expense item. This ensures that each sub-item corresponds to complete budgeted expenses, actual expenses, and premium income data. For example, the budgeted sales expense data, actual expenditure data, and corresponding premium income data generated by a branch through online promotion channels can be correlated into a single dimensional data record.

[0053] The indicator calculation process is completed automatically by non-intelligent tools without human intervention. For example, when calculating the total cost difference, the total budgeted cost and the total actual cost of all sub-items are first summarized, and then the result is obtained by the difference between the two. The cost difference contribution of each sub-item is obtained by dividing the cost difference amount of each sub-item by the total cost difference amount to get the specific proportion.

[0054] It should be understood that the summation and verification of the expense difference amounts of sub-items under the same dimension is an important quality control step. For example, after the expense difference amounts of all branches under the institutional dimension are added together, they must be completely consistent with the total expense difference amount. If there is a deviation, a data verification anomaly prompt will be triggered, and problems in the data association mapping or calculation process need to be investigated until the verification is passed before the standardized expense difference integrated data is generated.

[0055] This standardized calculation and verification process transforms cost data that was originally scattered across multiple systems and varied in format into integrated data with a unified structure and high accuracy. This not only reduces operational errors caused by human intervention but also significantly shortens the time cost of data processing. It allows subsequent penetrating analysis of large-scale models based on preset rules to be carried out directly on high-quality data without the need for additional time for data cleaning or correction, thus ensuring the efficiency and accuracy of cost difference analysis from the source.

[0056] S3. Using the RESTful API protocol, the standardized fee difference integration data is encapsulated in JSON format and pushed to the preset rule big model.

[0057] Specifically, the preset rule model is obtained through prompting engineering and small sample learning. In one possible implementation, the step of encapsulating the standardized fee difference integration data in JSON format using the RESTful API protocol and pushing the encapsulated standardized fee difference integration data to the preset rule model includes: encapsulating the standardized fee difference integration data in JSON format using the RESTful API protocol, initiating transmission to the preset rule model at a batch processing frequency of T+1 months, where T is the statistical base month; in the event that the standardized fee difference integration data transmission fails, retrying sequentially at preset time intervals; and triggering an exception alarm and recording the transmission log if the number of retries reaches a preset value and the transmission still fails.

[0058] Specifically, the statistical base month refers to the last day of the calendar month. For example, if T is June 30, 2024, the batch processing in T+1 month means that the encapsulation and transmission of the relevant standardized fee difference integration data for June will be completed within July 2024, which is in line with the regular business rhythm of insurance institutions to calculate fee differences on a monthly basis.

[0059] RESTful API protocol is the mainstream interface communication protocol, which has the characteristics of strong cross-platform compatibility and high transmission efficiency. JSON format, as a lightweight data exchange format, can concisely and clearly carry the field information of various dimensions in standardized data integration, ensuring that the data is not easily corrupted during transmission.

[0060] The preset time intervals are specifically set to an exponential backoff mechanism of 10 seconds, 30 seconds, and 60 seconds, with a preset retries of 3. This tiered retry strategy effectively addresses momentary transmission failures caused by network fluctuations and temporary high loads on large models, significantly improving the transmission success rate. Triggered anomaly alarms can be notified to relevant technical maintenance personnel via system pop-ups, emails, or SMS messages. The recorded transmission logs contain detailed information such as transmission time, data volume, failure reason, number of retries, and alarm trigger time, providing a complete basis for subsequent troubleshooting of transmission faults.

[0061] The method provided by this invention not only ensures that the data push process is highly aligned with the business cycle of insurance institutions, but also minimizes the interference of transmission interruptions on the cost difference analysis process through a scientific retry mechanism and alarm recording function. This allows standardized cost difference integration data to be delivered to the preset rule big model on time and completely, laying a stable foundation for the smooth implementation of subsequent penetration analysis, while also reducing the difficulty of troubleshooting for technical maintenance personnel.

[0062] Optionally, the method provided in this embodiment of the invention further includes: when the preset rule big model identifies that the standardized fee difference integration data has missing key indicators, insufficient data accuracy, or incomplete dimension coverage, generating a supplementary calculation instruction, the supplementary calculation instruction including the missing indicator type, data accuracy requirements, and dimension supplementation range; the non-intelligent tool responds to the supplementary calculation instruction, obtains multi-source expense-related data from the insurance institution's business system and financial system, re-executes indicator calculation, and obtains updated standardized fee difference integration data; the updated standardized fee difference integration data is encapsulated in JSON format using the RESTful API protocol, and the encapsulated updated standardized fee difference integration data is pushed to the preset rule big model.

[0063] Specifically, missing key indicators refer to the omission of core fields such as the achievement rate of a certain sub-item and budget expenses in the standardized expense difference integration data. For example, the achievement rate of sales expense difference for a certain branch was not included in the integrated data. Insufficient data accuracy refers to the fact that the number of decimal places retained in the data does not meet the analysis requirements. For example, the expense difference achievement rate is only retained to one decimal place, which cannot meet the needs of accurate analysis. Incomplete dimension coverage refers to the omission of one of the three dimensions of institution, channel, and expense item, or the absence of key sub-items under a certain dimension. For example, data on the sub-item of bancassurance cooperation channel is missing.

[0064] The pre-defined rule-based model automatically identifies the aforementioned data issues through built-in integrity and accuracy verification logic. The generated supplementary calculation instructions clearly indicate the specific problems, such as the missing contribution indicator for "Institutional Dimension - Beijing Branch - Management Expenses," requiring data precision to be retained to four decimal places, or the need to supplement budget expense data for all sub-items under the channel dimension. Non-intelligent tools, upon receiving these instructions, connect to the original data collection channels of the insurance institution's business and financial systems, re-extract the corresponding data, and perform indicator calculations to ensure that the updated standardized expense difference integration data can compensate for the original deficiencies. The updated data is still pushed using the RESTful API protocol in JSON format, maintaining consistency with the initial data transmission specifications and ensuring the continuity and consistency of data transmission.

[0065] The method provided by this invention proactively identifies defects in standardized expense difference integration data through a pre-defined rule-based large model. This ensures that data quality control is integrated throughout the entire expense difference analysis process, eliminating reliance on manual post-event verification. This reduces data analysis bias caused by human error and ensures the completeness and accuracy of each set of data pushed to the large model through clear instructions and standardized supplementary processes. This allows subsequent penetrating analysis to be conducted based on a reliable data foundation, making the final expense difference analysis results more credible and providing stronger support for the business decisions of insurance institutions.

[0066] S4. Based on the insurance financial expense difference business rule knowledge base, which includes the definition of expense difference penetration dimensions and the rules for identifying abnormal expense differences, the standardized expense difference integrated data is subjected to penetration analysis of the sub-items under each dimension through the preset rule big model, the causes of abnormal expense differences are identified and the influence weights are calculated, and expense difference analysis results are generated.

[0067] Specifically, the cost difference analysis results include data presentation, root cause analysis, and solutions.

[0068] In some embodiments, S4 above includes: calling the fee difference penetration dimension definition rules in the insurance financial fee difference business rule knowledge base through the preset rule big model to construct a penetration analysis framework, the penetration analysis framework including multiple levels of total dimension, single dimension, sub-item and cross dimension; comparing the fee difference achievement rate and fee difference contribution of each sub-item with the preset threshold according to the abnormal fee difference identification rules in the knowledge base, marking abnormal sub-items that exceed the threshold range, and determining the degree of abnormal deviation of each abnormal sub-item; performing multi-dimensional penetration decomposition on the abnormal sub-items to obtain the fee difference change logic, collaborative influence relationship, and fee difference fluctuation pattern of continuous statistical period, so as to... The analysis identifies the direct, indirect, and root causes of abnormal expense discrepancies. Based on the causal association rules in the knowledge base, a weighted scoring method is used to calculate the influence weight of each root cause. Root causes with an influence weight exceeding a preset proportion are identified as core root causes. Expense discrepancy analysis results are generated, including data statements, root cause interpretations, and solutions. The data statements are used to present expense discrepancy-related data and anomaly indicators for each dimension and sub-item in a hierarchical manner. The root cause interpretations are used to quantify the influence weight of core root causes and explain their causes. The solutions are based on the best practice solutions in the knowledge base, combined with the business scenarios of insurance institutions to optimize execution steps and expected results.

[0069] In one example, the Insurance Finance Expense Difference Business Rules Knowledge Base contains more than 2,000 business rules. Among them, the Expense Difference Penetration Dimension Definition Rules clarify the specific scope and correlation of each level of analysis. The total dimension corresponds to the overall expense difference situation of the entire organization, while the single dimension focuses on the expense difference performance of the organization, channel, and expense item respectively. The subdivision item goes into the specific branch, channel type, or expense expenditure item, and the cross dimension combines two or more dimensions for joint analysis, such as the sales expense difference situation of a branch through online promotion channels.

[0070] The preset threshold is specifically set as follows: the cost difference achievement rate is below -5% or above 10%, and the cost difference contribution ratio exceeds 30%. The degree of abnormal deviation is determined by calculating the ratio of the difference between the actual value and the threshold. For example, if the cost difference achievement rate of a certain sub-item is 15%, which exceeds the threshold of 10% by 5 percentage points, its degree of abnormal deviation is 50%.

[0071] During the multi-dimensional analysis, the logic of expense difference changes analyzes the difference between the budget and actual cost of each sub-item. The synergistic impact relationship explores the interaction between different sub-items, such as the impact of a channel's cost overrun on the overall expense difference of the relevant organization. The fluctuation pattern of expense difference over a continuous statistical period compares the trend of expense difference data over the past 3-12 months. A direct cause might be that the actual cost of a sub-item exceeds the budget; an indirect cause might involve changes in the market environment leading to increased channel promotion costs; and a fundamental cause might be that market fluctuations were not fully considered when setting the budget.

[0072] The weighted scoring method assigns weight coefficients based on factors such as the scope and severity of the root cause's impact. The preset ratio is set at 30%, meaning that root causes with an impact weight exceeding 30% are identified as core root causes. In the generated cost difference analysis results, the data presentation clearly displays the budgeted costs, actual costs, cost difference achievement rates, and other data for each dimension and sub-item in a hierarchical manner, and anomalies are marked.

[0073] Root cause analysis clarifies the specific impact weight of the core root cause, such as the impact weight of a branch's sales expense overrun being 45%, and details how this root cause leads to abnormal expense discrepancies. The solution will match corresponding best practices from the knowledge base, such as optimizing the budgeting process and strengthening channel expense control, and will refine the execution steps and expected goals based on the insurance institution's business scale, operating characteristics, and other scenario factors. For example, it will optimize the budgeting process within 3 months to keep the achievement rate of relevant sub-items within the threshold range.

[0074] The method provided by this invention constructs a multi-level penetrating analysis framework by calling a business rule knowledge base through a pre-set rule-based large model. It marks abnormal sub-items by comparing thresholds, deeply deconstructs the logic, synergistic effects, and causes of expense difference changes, and uses a weighted scoring method to quantify the root cause weights and generate structured results. This achieves comprehensive deconstruction and accurate root cause location from the total dimension to the cross-dimensional dimension, making the analysis results both hierarchical and able to directly guide practice without the need for additional manual sorting, thus improving the professionalism and practicality of expense difference analysis.

[0075] As can be seen from S1-S4 above, the method provided by the embodiments of the present invention obtains multi-source data such as budget expenses, actual expenses, and premium income from the business system and financial system of insurance institutions. It uses non-intelligent tools to calculate standardized expense difference integrated data containing three-dimensional sub-item data of institutions, channels, and expense items. Then, it pushes the data to a preset rule big model that has been calibrated through prompting engineering and small sample learning via RESTful API protocol. Finally, it completes multi-dimensional penetrating analysis based on the business rule knowledge base and generates results containing data statements, root cause interpretations, and solutions. It eliminates the need for manual splicing of data from multiple systems and manual tracing of causes, effectively solving the problems of high data processing time, insufficient penetrating dimensions, and the need for manual interpretation of reports in the prior art, and realizes the automation, intelligence, and accuracy of expense difference analysis.

[0076] In one possible implementation, the method further includes: importing core rules from the insurance financial expense difference business rule knowledge base into the base model, so that the base model can establish the business boundaries and core judgment criteria for expense difference analysis; constructing a dedicated template for insurance financial expense difference analysis through prompting engineering, the dedicated template for insurance financial expense difference analysis being used to adjust the analytical logic framework and output format specifications of the base model; and inputting multiple sets of historical expense difference analysis examples into the base model using a few-sample learning approach, each set of historical expense difference analysis examples including corresponding input data and standard output conclusions, so that the base model can master the business logic and expression specifications of expense difference analysis, thereby obtaining the preset rule big model.

[0077] In one example, the base model specifically uses DeepSeek v3, which has strong natural language processing capabilities and adaptability, and can quickly respond to rule insertion and template constraints.

[0078] In another example, the specific template for insurance financial expense difference analysis is as follows: "You are an insurance finance expert, specializing in expense differential knowledge and related data analysis, and able to organize and present financial data results in a reasonable and easy-to-understand manner; your language style should be professional yet easy to understand, using the most understandable language possible for analysis."

[0079] #Task: We've already provided an overview of the main indicator. Now, based on user questions and query results, and considering the main indicator, its parent indicators, and sibling indicators, we need to analyze the parent indicator's performance and how the main indicator influences it. The main text should be bulleted down (do not use the HTML detail section), and no main headings are needed. The relationships between the main indicator, parent indicators, and sibling indicators are derived from the formulas in the indicator definitions. Only analyze the platforms mentioned in the main indicator section; do not analyze other platforms. #Task Requirements: - The analysis should first state the data situation, then evaluate the data performance. The current platform, institution, and time should be clearly stated in the description and title. --First describe the deviation of the parent indicator, then use one sentence to explain how the main indicator and its sibling indicators affect the parent indicator; #Formatting Requirements: Please strictly follow the structure below and do not output any parts outside the structure; please note that each part needs to show the platform, organization, and data time of the current data; The title is: ## Analysis of the impact of {primary indicator name} on {parent indicator name} First, describe the deviation of the parent indicator, and then use one sentence to explain how the main indicator and its sibling indicators affect the parent indicator.

[0080] #Further Knowledge: The following information provides relevant metrics that may be used to answer the questions. Please use and refer to it appropriately.

[0081] Output format: 1. Reasons for anomalies in different metrics across different platforms, and anomaly analysis. 2. In-depth analysis of each root cause 3. Recommended Solutions 4. Few-Shot Learning: Provides 1-5 sets of excellent historical analysis reports as examples, each set including an input table and output conclusions.

[0082] 5. Model parameters: The LLM uses DeepSeek v3. It should be noted that the above-mentioned special template for insurance financial expense difference analysis is only for illustrative purposes, and the embodiments of the present invention do not impose any special restrictions on the specific implementation of the special template for insurance financial expense difference analysis.

[0083] Optionally, the Insurance Finance Expense Difference Business Rules Knowledge Base includes multiple business rules. Its core rules cover key content such as the definition of expense difference penetration dimensions, the threshold for identifying abnormal expense differences, the logic of cause correlation, and compliance analysis requirements. During the integration process, these rules will be transformed into structured instructions that the model can recognize and incorporated into the training logic of the basic model, thus defining clear business analysis boundaries for the model.

[0084] The dedicated template for insurance financial expense difference analysis is built through a prompt-based engineering process. The template clearly stipulates that the analysis logic must follow the order of "data presentation - anomaly identification - cause tracing - conclusion output". It must also link key information such as institutions, channels, expense item sub-items and statistical periods. The output format is fixed as a three-part structure of "anomaly cause - root cause interpretation - solution". At the same time, it is constrained to use a professional and easy-to-understand language style to avoid redundant expressions.

[0085] Furthermore, the number of historical expense difference analysis examples is set to 1-5 sets. Each example consists of input data and standard output conclusions. The input data is a standardized expense difference integrated data fragment, and the standard output conclusions are the corresponding structured analysis results. These examples help the basic model quickly learn the correlation between expense difference indicators, anomaly judgment criteria, and the expression norms of natural language conclusions. Throughout the calibration process, the model output results will be continuously monitored to ensure that it gradually masters the business logic and expression norms of expense difference analysis, ultimately forming a large-scale model with preset rules adapted to insurance financial expense difference analysis scenarios.

[0086] The method provided by this invention, through targeted tuning, endows the originally general basic model with professional capabilities for insurance financial expense difference analysis. It is no longer limited to simple natural language processing, but can accurately meet the business needs of the insurance industry, outputting professional, logical and practical analysis results. It fills the gap in the field of financial professional analysis of traditional large models and substantially improves the intelligence level of expense difference analysis.

[0087] To facilitate understanding of this solution, the following specific example will be used to further illustrate the method provided in the embodiments of the present invention.

[0088] In one example, the method provided in this embodiment of the invention is used to perform expense difference analysis for life insurance company A in June 2024. The statistical base month T is set as June 2024, and the batch processing in T+1 month, i.e., the entire process analysis is completed within July 2024. Specifically, it includes the following steps: First, multi-source data is extracted from the business and financial systems. The institutional dimension includes Beijing and Shanghai branches, the channel dimension includes offline agents, online promotion, and bancassurance cooperation channels, and the expense item dimension includes sales, management, and financial expenses. Each sub-item corresponds to complete budget, actual expenses, and premium income data. For example, the Beijing branch's offline agent channel sales expenses were budgeted at 1 million yuan, actually amounted to 1.1 million yuan, and the premium income was 5 million yuan.

[0089] Next, a customized Excel plugin (a non-intelligent tool) calculates the total expense difference amount (400,000 yuan - total budget of 8 million yuan - total actual expense difference of 8.4 million yuan) and the total expense difference achievement rate (1% - 400,000 yuan - actual expense difference of 40 million yuan) according to a pre-set actuarial template (compliant with the "Insurance Company Expense Allocation Guidelines"). After the association mapping forms a dimensional data set, the total expense difference amount and contribution of each sub-item are also calculated simultaneously. After passing the same dimension summation verification, standardized expense difference integrated data is generated. For example, the Beijing branch's sales expenses for this channel correspond to a budget of 1 million yuan, a total expense difference achievement rate of -1%, and its own expense difference achievement rate of 10%.

[0090] The data was then encapsulated in RESTful API protocol and JSON format, and pushed to the pre-defined rule-based large model on July 10th in a T+1 month batch processing frequency.

[0091] The large model validation revealed that the achievement rate of financial expense difference in the bancassurance cooperation channel of the Shanghai branch was missing. A supplementary instruction was generated, and the non-intelligent tool re-extracted the data of this sub-item (budget of 200,000 yuan, actual amount of 220,000 yuan, premium of 1 million yuan), calculated the achievement rate of 20.00%, updated the data, and then pushed it again.

[0092] After the data is validated, the large model calls upon a knowledge base containing multiple rules to construct a four-level penetration framework. Items with a fee difference achievement rate below -5% or above +10% are marked as abnormal (deviation from 100%). Penetration analysis reveals that the item has been overspent for three consecutive months, related to rising bank service fees. The core root cause is that the budget did not include fee adjustment factors (weight 40%). The final analysis results are generated, with data presented hierarchically across various dimensions and anomaly indicators. The root cause analysis quantifies the core weights and explains the causes. The solution matches best practices from the knowledge base, clarifying the execution steps and expected results for optimizing the budget process within three months.

[0093] The foregoing primarily describes the solutions of the embodiments of the present invention from a methodological perspective. It is understood that, in order to achieve the above-mentioned functions, the cost difference analysis system 100 includes at least one of the hardware structures and software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of the present invention.

[0094] In this embodiment of the invention, the cost difference analysis system 100 can be divided into functional units according to the above method example. For example, the cost difference analysis system 100 can be divided into functional units corresponding to each function, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this embodiment of the invention is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0095] For example, Figure 3This diagram illustrates the hardware structure of a cost difference analysis system according to an embodiment of the present invention. The cost difference analysis system 100 includes: an acquisition module 110, used to acquire multi-source cost-related data from the insurance institution's business system and financial system, the multi-source cost-related data including budgeted cost data, actual cost data, and premium income data; a calculation module 120, used to perform indicator calculations on the multi-source cost-related data using non-intelligent tools to obtain standardized cost difference integrated data, the standardized cost difference integrated data including budgeted cost, cost difference achievement rate, and total cost difference achievement rate fields corresponding to each sub-item under each of the three dimensions of institution, channel, and cost item; and a push module 130, used to push data using a RESTful API. The protocol encapsulates the standardized fee difference integration data in JSON format and pushes the encapsulated standardized fee difference integration data to a preset rule model. The preset rule model is obtained through prompting engineering and small sample learning. The analysis module 140 is used to perform penetration analysis on the standardized fee difference integration data according to the sub-items under each dimension through the preset rule model, based on the insurance financial fee difference business rule knowledge base containing fee difference penetration dimension definitions and abnormal fee difference identification rules, to identify the causes of abnormal fee differences and calculate the influence weights, and generate fee difference analysis results. The fee difference analysis results include data statements, root cause interpretations, and solutions.

[0096] It should be understood that specific descriptions of the above-mentioned optional methods can be found in the foregoing method embodiments, and will not be repeated here. Furthermore, explanations of any of the cost difference analysis systems 100 provided above, as well as descriptions of their beneficial effects, can be found in the corresponding method embodiments described above, and will not be repeated here.

[0097] This invention also provides a computer-readable storage medium storing at least one computer instruction, which is loaded and executed by a processor to implement the methods of the various embodiments described above. Explanations of the relevant content and descriptions of the beneficial effects of any of the computer-readable storage media provided above can be found in the corresponding embodiments described above, and will not be repeated here.

[0098] This invention also provides a chip. This chip integrates a control circuit for implementing the functions of the aforementioned cost difference analysis system 100 and one or more ports. Optionally, the functions supported by this chip are as described above and will not be repeated here.

[0099] Those skilled in the art will understand that the program for implementing all or part of the steps of the above embodiments, which can be executed by a program instructing related hardware, can be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a random access memory, etc. The processing unit or processor mentioned above can be a central processing unit, a general-purpose processor, an application-specific integrated circuit (ASIC), a microprocessor (DSP), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof.

[0100] This invention also provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform any of the methods described in the above embodiments. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., SSD), etc.

[0101] It should be noted that the devices for storing computer instructions or computer programs provided in the embodiments of the present invention, such as, but not limited to, the aforementioned memory, computer-readable storage medium, and communication chip, are all non-transitory. Those skilled in the art should recognize that the functions described in the embodiments of the present invention in one or more of the above examples can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium accessible to general-purpose or special-purpose computers.

[0102] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A cost difference analysis method based on two-stage penetration processing, characterized in that, include: Acquire multi-source expense-related data from the business and financial systems of insurance institutions, including budgeted expense data, actual expense data, and premium income data. By performing indicator calculations on the multi-source cost-related data using non-intelligent tools, standardized cost difference integration data is obtained. The standardized cost difference integration data includes the budget cost, cost difference achievement rate and total cost difference achievement rate fields corresponding to each sub-item under each of the three dimensions of institution, channel and cost item. The standardized fee difference integration data is encapsulated in JSON format using the RESTful API protocol, and the encapsulated standardized fee difference integration data is pushed to the preset rule large model. The preset rule large model is obtained through prompting engineering and small sample learning and tuning. Based on the insurance financial expense difference business rule knowledge base, which includes the definition of expense difference penetration dimensions and the rules for identifying abnormal expense differences, the standardized expense difference integrated data is subjected to penetration analysis of the sub-items under each dimension through the preset rule big model. The causes of abnormal expense differences are identified and the influence weights are calculated to generate expense difference analysis results. The expense difference analysis results include data statements, root cause interpretations and solutions.

2. The cost difference analysis method based on two-stage penetration processing according to claim 1, characterized in that, Each of the three dimensions—organization, channel, and expense item—includes at least two sub-items. When the dimension is organization, the sub-items include different branches or subsidiaries of the insurance institution. When the dimension is channel, the sub-items include offline agency channels, online promotion channels, or bancassurance cooperation channels. When the dimension is expense item, the sub-items include premiums, human resource costs, daily expenses, or workplace expenses.

3. The cost difference analysis method based on two-stage penetration processing according to claim 2, characterized in that, The process of performing indicator calculations on the multi-source cost-related data using non-intelligent tools to obtain standardized cost difference integrated data includes: Multi-source cost-related data are correlated and mapped to obtain a dimensional data set, which includes budgeted cost data, actual cost data and premium income data corresponding to each sub-item under each dimension; The non-intelligent tools automatically perform indicator calculations on the dimensional data set according to the preset actuarial template to obtain the total cost difference amount, the total cost difference achievement rate, and the cost difference amount and cost difference contribution of each sub-item under each dimension. When the sum of the cost difference amounts of all sub-items under the same dimension equals the total cost difference amount, extract the budget cost data, total cost difference achievement rate and cost difference achievement rate of each sub-item under each dimension to generate the standardized cost difference integrated data; Wherein, the total cost difference amount is the difference between the total budgeted cost data and the total actual cost data of all sub-items, the cost difference amount of each sub-item is the difference between the budgeted cost data and the actual cost data of each sub-item, the total cost difference achievement rate is the ratio of the actual cost difference amount to the planned cost difference amount, and the cost difference contribution of each sub-item is the ratio of the cost difference amount of each sub-item to the total cost difference amount.

4. The cost difference analysis method based on two-stage penetration processing according to claim 1, characterized in that, The process of encapsulating the standardized fee difference integration data in JSON format using the RESTful API protocol and pushing the encapsulated standardized fee difference integration data to the preset rule-based large model includes: The standardized cost difference integration data is encapsulated in JSON format using the RESTful API protocol and transmitted to the pre-defined rule-based large model at a batch push frequency of T+1 months, where T is the statistical base month. In the event of failure of the standardized cost difference integrated data transmission, retry is performed sequentially at preset time intervals; if the number of retries reaches the preset value and the transmission still fails, an abnormal alarm is triggered and the transmission log is recorded.

5. The cost difference analysis method based on two-stage penetration processing according to claim 4, characterized in that, The method further includes: When the preset rule big model identifies that the standardized cost difference integrated data has missing key indicators, insufficient data accuracy, or incomplete dimension coverage, it generates a supplementary calculation instruction. The supplementary calculation instruction includes the type of missing indicator, data accuracy requirements, and dimension supplementation range. The non-intelligent tool responds to the supplementary calculation instruction, obtains multi-source expense-related data from the insurance institution's business system and financial system, re-executes the indicator calculation, and obtains updated standardized expense difference integrated data. The updated standardized fee difference integration data is encapsulated in JSON format using the RESTful API protocol, and then pushed to the preset rule-based large model.

6. The cost difference analysis method based on two-stage penetration processing according to claim 1, characterized in that, The insurance financial expense difference business rule knowledge base, which includes the definition of expense difference penetration dimensions and rules for identifying abnormal expense differences, performs penetration analysis on the standardized expense difference integrated data under each dimension through the preset rule model. This identifies the causes of abnormal expense differences, calculates their impact weights, and generates expense difference analysis results, including: By calling the rules of the fee difference penetration dimension definition in the insurance financial fee difference business rule knowledge base through the preset rule big model, a penetration analysis framework is constructed. The penetration analysis framework includes multiple levels of total dimension, single dimension, sub-item and cross dimension. Based on the abnormal cost difference identification rules in the knowledge base, the cost difference achievement rate and cost difference contribution of each sub-item are compared with the preset threshold. Abnormal sub-items that exceed the threshold range are marked, and the degree of abnormal deviation of each abnormal sub-item is determined. By performing multi-dimensional penetrating analysis on the abnormal sub-items, we can obtain the logic of expense difference changes, the synergistic influence relationship, the expense difference fluctuation pattern in continuous statistical periods, and the direct, indirect and root causes of abnormal expense differences. Based on the causal association rules in the knowledge base, a weighted scoring method is used to calculate the influence weight of each root cause, and root causes whose influence weight exceeds a preset proportion are identified as core root causes. Generate expense difference analysis results, which include data statements, root cause interpretations, and solutions. The data statements are used to present expense difference-related data and anomaly indicators for each dimension and sub-item in a hierarchical manner. The root cause interpretations are used to quantify the impact weight of core root causes and explain their causes. The solutions are to match the best practice solutions in the knowledge base and optimize the execution steps and expected results in combination with the business scenarios of insurance institutions.

7. The cost difference analysis method based on two-stage penetration processing according to claim 1, characterized in that, The method further includes: Import the core rules of the insurance financial expense difference business rule knowledge base into the basic model so that the basic model can establish the business boundary and core judgment criteria for expense difference analysis. The project constructs a dedicated template for insurance financial expense difference analysis, which is used to adjust the analysis logic framework and output format specifications of the basic model. Multiple sets of historical cost difference analysis examples are input into the base model using a few-sample learning approach. Each set of historical cost difference analysis examples includes corresponding input data and standard output conclusions, so that the base model can master the business logic and expression specifications of cost difference analysis and obtain the preset rule big model.

8. A cost difference analysis system based on two-stage penetration processing, characterized in that, include: The acquisition module is used to acquire multi-source expense-related data from the business and financial systems of insurance institutions. The multi-source expense-related data includes budgeted expense data, actual expense data, and premium income data. The calculation module is used to perform indicator calculations on the multi-source cost-related data using non-intelligent tools to obtain standardized cost difference integration data. The standardized cost difference integration data includes the budget cost, cost difference achievement rate and total cost difference achievement rate fields corresponding to each sub-item under each of the three dimensions of institution, channel and cost item. The push module is used to encapsulate the standardized fee difference integration data in JSON format using the RESTful API protocol, and push the encapsulated standardized fee difference integration data to the preset rule big model. The preset rule big model is obtained through prompting engineering and small sample learning and tuning. The analysis module is used to perform penetration analysis on the standardized expense difference integrated data based on the insurance financial expense difference business rules knowledge base, which includes the definition of expense difference penetration dimensions and the rules for identifying abnormal expense differences. It identifies the causes of abnormal expense differences and calculates the impact weights, and generates expense difference analysis results, which include data statements, root cause interpretations and solutions.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the cost difference analysis method based on two-stage penetration processing as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to enable the computer to implement the cost difference analysis method based on two-stage penetration processing as described in any one of claims 1-7.

Images