A small and micro loan approval method based on large language model

Abstract

The present invention provides an expert system and method for micro-credit approval officers based on a large language model, which belongs to the technical field of approval management and specifically includes: a problem point recording module responsible for recording problem points of a credit report during the approval process; a potential problem point output module responsible for identifying and processing the credit report and the problem points of the credit report using the large language model as a basis to obtain potential problem points; a retrieval processing module responsible for performing retrieval processing based on the potential problem points to obtain supplementary retrieval information of the potential problem points; and a conclusion output module responsible for outputting an evaluation conclusion of the credit report based on the potential problem points and the corresponding supplementary retrieval information, thereby improving the efficiency of the approval process of the credit report.

Description

Technical Field

[0001] The present invention belongs to the technical field of approval management, and in particular relates to an expert system and method for micro-loan approval officers based on a large language model. Background Art

[0002] In the field of micro-loans, the review of credit reports is difficult and labor-intensive. It involves thousands of industries and clients with complex and ever-changing situations. The data obtained through expert judgment and credit investigation is limited, making it difficult to guarantee the effectiveness of approval.

[0003] In order to realize the approval and processing of credit reports, for example, in the invention patent application CN202011435869.0 "A method for intelligent analysis of credit customers", a credit customer model is established based on account settlement transaction flow data, and a credit report is generated according to the credit customer model, so that the data in the credit report is more unified, comprehensive, detailed and accurate, and the credit report is more standardized, which is conducive to improving approval efficiency.

[0004] During the credit report approval process, the credit report involves a large amount of data. Existing technical solutions often rely on the personal experience of the reviewer. Therefore, not only is the accuracy of identifying and processing problem items difficult to meet the requirements, but also targeted data retrieval cannot be performed during the approval process to identify and process problem items, resulting in the failure to meet the requirements for accuracy and efficiency of credit report approval processing.

[0005] In order to solve the above technical problems, the present application provides an expert system and method for micro-loan approval officers based on a large language model. Summary of the Invention

[0006] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present application provides a micro-loan expert system based on a large language model, specifically comprising:

[0008] Problem point recording module, potential problem point output module, retrieval processing module, conclusion output module;

[0009] The problem point recording module is responsible for recording the problem points of the credit report during the approval process;

[0010] The potential problem point output module is responsible for identifying and processing the credit report and the problem points in the credit report using a large language model to obtain potential problem points;

[0011] The retrieval processing module performs retrieval processing based on the potential problem points to obtain supplementary retrieval information of the potential problem points;

[0012] The conclusion output module is responsible for outputting the evaluation conclusion of the credit report based on potential problem points and corresponding supplementary search information.

[0013] The beneficial effects of the present invention are:

[0014] Based on the credit report and the problem points in the credit report, the large language model is used to identify and process potential problem points, thereby avoiding the technical problem of the original review and processing efficiency and accuracy not meeting the requirements due to relying solely on the personal experience of the examiner. Taking full consideration of the advantages of the large language model in data analysis and reasoning processing, it also lays the foundation for further retrieval, analysis and processing based on potential problem points.

[0015] The evaluation conclusion of the credit report is output based on potential problem points and corresponding supplementary search information, thereby avoiding the technical problem that the reliability of the approval processing results does not meet the requirements due to the single consideration of potential problem points and credit report data, improving the accuracy and reliability of the overall credit report approval processing, and also ensuring the efficiency of the credit report approval processing.

[0016] A further technical solution is that the problem points include credit data, credit usage data, production and operation indicators and financial indicators.

[0017] A further technical solution is that the supplementary search information is a search result of data associated with the potential problem point.

[0018] A further technical solution is to output the evaluation conclusion of the credit report, specifically including:

[0019] The potential risk points and the supplementary search information of the potential risk points are used as input, and the output results of the large model are used to determine the evaluation conclusion of the credit report.

[0020] In a second aspect, the present application provides a small and micro loan approval method based on a large language model, which is applied to the above-mentioned small and micro loan expert system based on a large language model, specifically comprising:

[0021] S1 records the problem points of the credit report during the approval process, determines the related report content of different problem points based on the type of the problem points, and proceeds to the next step when it is determined that the risk of the credit report meets the requirements based on the related report content of different problem points;

[0022] S2 uses the credit report and the problem points in the credit report as a basis, uses a large language model to perform identification processing, obtains potential problem points, and determines the target search source based on the information association and update status of different search sources and potential problem points;

[0023] S3: Determining deviations in the supplementary search information for different potential risk points based on the search processing results of the target search source, and when it is determined that the search plan needs to be adjusted based on the deviations, obtaining supplementary search information for the credit report based on all search sources;

[0024] S4 outputs the evaluation conclusion of the credit report based on the potential problem points and the corresponding supplementary search information.

[0025] A further technical solution is that the problem points in the approval process are determined based on the problems in the approval process of the relevant personnel of the credit report.

[0026] A further technical solution is that the content of the associated report of the problem point is determined according to the content of the report in which the problem point exists.

[0027] A further technical solution is to determine whether the risk of the credit report meets the requirements, specifically including:

[0028] Determining the content of the associated reports in different subunits of the credit report based on the content of the associated reports of the problem points;

[0029] Determining problematic subunits of the associated report content based on constituent data of the associated report content in different subunits;

[0030] Whether the risk of the credit report meets the requirements is determined by the number of the problem sub-units.

[0031] A further technical solution is that when the risk of the credit report does not meet the requirements, supplementary search information of the credit report is obtained based on all indicators and all search sources.

[0032] A further technical solution is that the sub-units include a credit analysis sub-unit, a production and operation sub-unit and a financial analysis sub-unit.

[0033] A further technical solution is to determine whether the search scheme needs to be adjusted, specifically including:

[0034] Determine the deviation between the supplementary search information and the associated report content of the different potential risk points in the credit report based on the deviation between the supplementary search information of the target data source;

[0035] Determining the number of potential risk points in the deviated associated report content based on the deviation between the associated report content and the supplementary search information;

[0036] Determine whether the retrieval plan needs to be adjusted based on the number of potential risk points in the deviated associated report content.

[0037] A further technical solution is that when the number of potential risk points in the deviated associated report content is greater than a preset risk point number threshold, it is determined that the retrieval scheme needs to be adjusted.

[0038] A further technical solution is that, when there is no need to adjust the retrieval scheme, the potential risk points and the supplementary retrieval information of the potential risk points in the target data source are used as input, and the output results of the large model are used to determine the evaluation conclusion of the credit report.

[0039] A further technical solution is that the evaluation conclusion of the credit report includes pass and fail.

[0040] Other features and advantages will be described in the following description. The objectives and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and drawings.

[0041] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and other features and advantages of the present invention will become more apparent by describing in detail example embodiments thereof with reference to the accompanying drawings.

[0043] Figure 1 It is a framework diagram of a small micro-loan expert system based on a large language model;

[0044] Figure 2 It is a flowchart of a micro-credit approval method based on a large language model;

[0045] Figure 3 It is a flow chart to determine whether the risk of credit report meets the requirements;

[0046] Figure 4 is a flow chart of a method for determining a target retrieval source;

[0047] Figure 5 This is a flowchart for determining the need for adjustment of the search plan. DETAILED DESCRIPTION

[0048] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this specification without creative work should fall within the scope of protection of this specification.

[0049] In this application, by using a large model to identify potential risk points in the credit report, and further searching the potential risk points to obtain supplementary search information, the supplementary search information and potential risk points are used to output the evaluation conclusion of the credit report based on the large model, thereby improving the efficiency of the credit report approval process.

[0050] Example 1

[0051] like Figure 1 As shown, this application provides a small micro-loan expert system based on a large language model, specifically including:

[0052] Problem point recording module, potential problem point output module, retrieval processing module, conclusion output module;

[0053] The problem point recording module is responsible for recording the problem points of the credit report during the approval process;

[0054] The potential problem point output module is responsible for identifying and processing the credit report and the problem points in the credit report using a large language model to obtain potential problem points;

[0055] The retrieval processing module performs retrieval processing based on the potential problem points to obtain supplementary retrieval information of the potential problem points;

[0056] The conclusion output module is responsible for outputting the evaluation conclusion of the credit report based on potential problem points and corresponding supplementary search information.

[0057] Furthermore, the problem points include credit data, credit usage data, production and operation indicators, and financial indicators.

[0058] Specifically, the supplementary search information is a search result of data associated with the potential problem point.

[0059] It is understood that the output of the credit report assessment conclusion includes:

[0060] The potential risk points and the supplementary search information of the potential risk points are used as input, and the output results of the large model are used to determine the evaluation conclusion of the credit report.

[0061] Example 2

[0062] Second, as Figure 2 As shown, the present application provides a small and micro loan approval method based on a large language model, which is applied to the above-mentioned small and micro loan expert system based on a large language model, specifically including:

[0063] S1 records the problem points of the credit report during the approval process, determines the related report content of different problem points based on the type of the problem points, and proceeds to the next step when it is determined that the risk of the credit report meets the requirements based on the related report content of different problem points;

[0064] When the word count of the report content related to the problem points in the credit report accounts for more than 0.2, it is determined that the risk of the credit report does not meet the requirements.

[0065] S2 uses the credit report and the problem points in the credit report as a basis, uses a large language model to perform identification processing, obtains potential problem points, and determines the target search source based on the information association and update status of different search sources and potential problem points;

[0066] Based on the information association between the retrieval source and the potential problem points, it is determined that there are potential problem points in the retrieval source that supplement the retrieval information, and they are used as information-associated problem points. According to the data update cycle of the retrieval source, the timeliness coefficients of different information-associated problem points are determined. Based on the sum of the timeliness coefficients of different information-associated problem points, the matching coefficient of the target retrieval source is determined. The target retrieval source is the data source with the largest matching coefficient and all potential problem-associated points are information-associated problem points.

[0067] S3: Determining deviations in the supplementary search information for different potential risk points based on the search processing results of the target search source, and when it is determined that the search plan needs to be adjusted based on the deviations, obtaining supplementary search information for the credit report based on all search sources;

[0068] Based on the deviation between the associated report content and the supplementary retrieval information, the number of potential risk points in the deviated associated report content is determined. When the number of potential risk points in the deviated associated report content is greater than the preset risk point number threshold, it is determined that the retrieval scheme needs to be adjusted.

[0069] S4 outputs the evaluation conclusion of the credit report based on the potential problem points and the corresponding supplementary search information.

[0070] Furthermore, the problem points in the approval process are determined based on the problems in the approval process of the relevant personnel of the credit report.

[0071] It can be understood that the content of the associated report of the problem point is determined according to the content of the report in which the problem point exists.

[0072] Specifically, such as Figure 3 As shown, determining that the risk of the credit report meets the requirements includes:

[0073] Determining the content of the associated reports in different subunits of the credit report based on the content of the associated reports of the problem points;

[0074] Determining problematic subunits of the associated report content based on constituent data of the associated report content in different subunits;

[0075] Whether the risk of the credit report meets the requirements is determined by the number of the problem sub-units.

[0076] Furthermore, the problem sub-unit is a sub-unit in which the word count ratio of the associated report content is greater than a preset word count ratio.

[0077] It should be noted that when the number of the problematic sub-units is greater than a preset sub-unit number threshold, it is determined that the risk of the credit report does not meet the requirements.

[0078] It is understandable that when the risk of the credit report does not meet the requirements, supplementary search information of the credit report is obtained based on all indicators and the full amount of search sources.

[0079] Furthermore, the sub-units include a credit analysis sub-unit, a production and operation sub-unit, and a financial analysis sub-unit.

[0080] In another possible embodiment, determining that the risk of the credit report meets the requirements specifically includes:

[0081] Determining the word count ratio of the related report content in the credit report based on the related report content of the problem point;

[0082] Determine whether the risk of the credit report meets the requirements based on the word count ratio of the associated report content.

[0083] Furthermore, when the word count ratio of the associated report content is not within a preset word count ratio range, it is determined that the risk of the credit report does not meet the requirements.

[0084] In another possible embodiment, determining that the risk of the credit report meets the requirements specifically includes:

[0085] Obtaining problem points of the credit report, and when the number of problem points of the credit report does not meet the requirement, determining that the risk of the credit report does not meet the requirement;

[0086] When the number of problem points on the credit report meets the requirements:

[0087] When the number of problem points in different indicators is determined to be insufficient to meet the required indicators, the risk of the credit report is determined to be insufficient to meet the requirements;

[0088] When the number of problem points does not meet the required indicators:

[0089] Obtaining indicators with problematic points, and determining the indicator risk coefficient of the credit report based on the number of problematic points in different indicators; and determining that the risk of the credit report meets the requirements when the indicator risk coefficient of the credit report is within a preset indicator risk coefficient range;

[0090] When the indicator risk coefficient of the credit report is not within the preset indicator risk coefficient range:

[0091] Determining the associated report contents in different subunits of the credit report based on the associated report contents of the problem points, and determining that the risk of the credit report does not meet the requirements when the number of subunits with associated report contents does not meet the requirements;

[0092] When the number of subunits with associated report content meets the requirement,

[0093] When it is determined based on the constituent data of the associated report contents in different sub-units that the associated report contents do not contain problematic sub-units, it is determined that the risk of the credit report meets the requirements;

[0094] When there are problematic subunits in the associated report content:

[0095] Obtaining the number of problem subunits in the associated report content, and when the number of problem subunits in the associated report content does not meet the requirement, determining that the risk of the credit report does not meet the requirement;

[0096] When the number of problem subunits in the associated report content meets the requirements:

[0097] The report content risk coefficient of the credit report is determined based on the constituent data of the associated report content in different sub-units and the number of problem sub-units, and whether the risk of the credit report meets the requirements is determined based on the report content risk coefficient.

[0098] Furthermore, when the risk coefficient of the report content does not meet the requirements, it is determined that the risk of the credit report does not meet the requirements.

[0099] Specifically, the indicators include credit data, credit usage data, production and operation indicators, and financial indicators.

[0100] Specifically, the potential problem points are determined based on the inference results of the large model.

[0101] It should be noted that if Figure 4 As shown, the method for determining the target search source is:

[0102] Based on the information association between the search source and the potential problem point, determining that the search source has a potential problem point for supplementing the search information, and using it as an information-associated problem point;

[0103] Determining the timeliness coefficients of different information-related problem points based on the data update cycle of the retrieval source;

[0104] Based on the sum of the timeliness coefficients of different information-related problem points, a matching coefficient of the target retrieval source is determined, and based on the matching coefficient, it is determined whether the data source is the target retrieval source.

[0105] Furthermore, the timeliness coefficient of the information-related problem point is determined according to the ratio of the data update time of the retrieval source to the preset update time.

[0106] It can be understood that the target retrieval source is a data source with the largest matching coefficient and all potential problem-related points belonging to information-related problem points.

[0107] In another possible embodiment, the method for determining the target retrieval source is:

[0108] Based on the information association between the search source and the potential problem point, determining that the search source has a potential problem point for supplementing the search information, and using it as an information-associated problem point;

[0109] Determining the timeliness coefficients of different information-related problem points based on the data update cycle of the retrieval source;

[0110] The matching coefficient of the target retrieval source is determined based on the product of the ratio of the number of different information-related problem points to the potential problem points and the timeliness coefficient, and whether the data source is the target retrieval source is determined based on the matching coefficient.

[0111] In another possible embodiment, the method for determining the target retrieval source is:

[0112] Based on the information association between the search source and the potential problem point, determining that the search source has a potential problem point with supplementary search information, and using the potential problem point as the information association problem point; and determining information association coefficients with different information association problem points based on the amount of data of the supplementary search information for different information association problem points;

[0113] Determining the timeliness coefficients of different information-related problem points based on the data update cycle of the retrieval source;

[0114] Based on the average value of the product of information association coefficients and timeliness coefficients of different information association problem points, a matching coefficient of the target retrieval source is determined, and based on the matching coefficient, it is determined whether the data source is the target retrieval source.

[0115] Furthermore, the information relevance coefficient is determined according to the ratio of the amount of data of the supplementary search information of the information relevance problem point to the amount of information to be verified and processed of the information relevance problem point.

[0116] Optionally, the method for determining the target search source is:

[0117] Based on the information association between the search source and the potential problem point, determining that the search source has a potential problem point that supplements the search information and using it as an information-associated problem point; if the search source has a potential problem point that does not belong to the information-associated problem point, determining that the search source does not belong to the target search source;

[0118] When the search source does not have any potential problem points that are not information-related problem points:

[0119] obtaining an average update period of the search information of the search source, and determining that the search source does not belong to the target search source when the average update period does not meet the requirement;

[0120] When the average update period meets the requirements:

[0121] determining timeliness coefficients of different information-related problem points based on a data update period of the supplementary search information of the search source and the information-related problem point, and determining that the search source does not belong to the target search source when an average value of the timeliness coefficients of the different information-related problem points does not meet a requirement;

[0122] When the average value of the timeliness coefficients of different information-related problem points meets the requirements:

[0123] Obtaining the number of information-related problem points whose timeliness coefficients do not meet the requirements, and when the number of information-related problem points whose timeliness coefficients do not meet the requirements is greater than a preset number of related problem points, determining that the search source does not belong to the target search source;

[0124] When the number of information-related problem points that do not meet the requirements for the timeliness coefficient is not greater than the preset number of related problem points:

[0125] determining information correlation coefficients with different information correlation problem points based on the amount of data of the supplementary search information of different information correlation problem points, and determining that the search source does not belong to the target search source when the average value of the information correlation coefficients with the different information correlation problem points does not meet the requirements;

[0126] When the average value of the information correlation coefficients with different information correlation problem points meets the requirements:

[0127] The matching coefficient of the target retrieval source is determined based on the average value of the product of the information correlation coefficient and the timeliness coefficient of different information correlation problem points, and whether the data source is the target retrieval source is determined based on the matching coefficient.

[0128] Specifically, such as Figure 5 As shown, it is determined that the search plan needs to be adjusted, including:

[0129] Determine the deviation between the supplementary search information and the associated report content of the different potential risk points in the credit report based on the deviation between the supplementary search information of the target data source;

[0130] Determining the number of potential risk points in the deviated associated report content based on the deviation between the associated report content and the supplementary search information;

[0131] Determine whether the retrieval plan needs to be adjusted based on the number of potential risk points in the deviated associated report content.

[0132] Furthermore, when the number of potential risk points in the deviated associated report content is greater than a preset risk point number threshold, it is determined that the retrieval scheme needs to be adjusted.

[0133] It should be noted that when there is no need to adjust the retrieval plan, the potential risk points and the supplementary retrieval information of the potential risk points in the target data source are used as input, and the output results of the large model are used to determine the evaluation conclusion of the credit report.

[0134] It is understandable that the evaluation conclusions of the credit report include pass and fail.

[0135] In another possible embodiment, determining that the search scheme needs to be adjusted specifically includes:

[0136] S41 determines the deviation between the supplementary search information of the target data source and the associated report content of the different potential risk points in the credit report and the supplementary search information based on the deviation between the associated report content of the different potential risk points in the credit report and the supplementary search information, and determines the data deviation coefficient of the different potential risk points based on the deviation between the associated report content of the different potential risk points in the credit report and the supplementary search information;

[0137] S42 determines the risk correlation coefficient of the target data source at the potential risk point based on the data volume and data update cycle of the supplementary search information of the target data source at the potential risk point;

[0138] S43 determines the data deviation probability evaluation value of the credit report based on the risk correlation coefficient and data deviation coefficient of different potential risk points, and determines whether the retrieval plan needs to be adjusted according to the data deviation probability evaluation value.

[0139] Furthermore, when the data deviation probability evaluation value is greater than a preset deviation probability threshold, it is determined that there is no need to adjust the retrieval scheme.

[0140] Optionally, the above step S41 includes the following contents:

[0141] S411 determines the deviation between the content of the associated report of the different potential risk points in the credit report and the supplementary search information based on the deviation between the content of the associated report of the different potential risk points and the supplementary search information. If there is no deviation between the content of the associated report and the supplementary search information, it is determined that no adjustment of the search plan is required. If there is a deviation between the content of the associated report and the supplementary search information, the process proceeds to step S412.

[0142] Step S412 obtains the number of associated report contents that deviate from the supplementary search information. If the number of associated report contents that deviate from the supplementary search information does not meet the requirement, it is determined that the search plan needs to be adjusted. If the number of associated report contents that deviate from the supplementary search information meets the requirement, the process proceeds to step S413.

[0143] S413 determines the data deviation coefficients of different potential risk points based on the deviations between the associated report content in the credit report and the supplementary search information. If the sum of the data deviation coefficients of different potential risk points does not meet the requirements, it is determined that the search plan needs to be adjusted. If the sum of the data deviation coefficients of different potential risk points meets the requirements, the process proceeds to step S414.

[0144] S414: When there is a potential risk point where the data deviation coefficient does not meet the requirements, the process proceeds to step S415; when there is no potential risk point where the data deviation coefficient does not meet the requirements, the process proceeds to step S42;

[0145] S415 When the number of potential risk points that do not meet the requirements of the data deviation coefficient is greater than the preset value of the number of risk points, it is determined that the retrieval scheme needs to be adjusted. When the number of potential risk points that do not meet the requirements of the data deviation coefficient is not greater than the preset value of the number of risk points, proceed to step S42.

[0146] Optionally, the above step S42 includes the following contents:

[0147] S421 determines the risk correlation coefficient of the target data source at the potential risk point based on the data volume and data update cycle of the supplementary retrieval information of the target data source at the potential risk point, and regards the potential risk point whose risk correlation coefficient is less than a preset risk correlation coefficient threshold as a shallowly correlated risk point. When the number of the shallowly correlated risk points or the proportion of the shallowly correlated risk points to the number of potential risk points does not meet the requirements, it is determined that the retrieval scheme needs to be adjusted. When the number of the shallowly correlated risk points or the proportion of the shallowly correlated risk points to the number of potential risk points both meet the requirements, the process proceeds to step S422.

[0148] S422: When the number of shallowly associated risk points is within the preset shallowly associated risk point number range, the process proceeds to step S423; when the number of shallowly associated risk points is not within the preset shallowly associated risk point number range, the process proceeds to step S424;

[0149] S423: When the average value of the data deviation coefficient of the shallowly associated risk point is within the preset deviation coefficient range, it is determined that the retrieval scheme needs to be adjusted. When the average value of the data deviation coefficient of the shallowly associated risk point is not within the preset deviation coefficient range, the process proceeds to step S424.

[0150] S424 determines the data source deviation of different potential risk points by multiplying the risk correlation coefficient of the target data source at the potential risk point and the data deviation coefficient. When the average value of the data source deviation of different potential risk points does not meet the requirements, it is determined that the retrieval scheme needs to be adjusted. When the average value of the data source deviation of different potential risk points meets the requirements, the process proceeds to step S43.

[0151] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant details, refer to the descriptions of the method embodiments.

[0152] The foregoing description of specific embodiments of this specification has been provided. In some cases, the actions or steps described may be performed in an order different from that shown in the embodiments and still achieve the desired results. Additionally, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0153] The foregoing description is merely one or more embodiments of this specification and is not intended to limit this specification. It will be apparent to those skilled in the art that various modifications and variations may be made to one or more embodiments of this specification. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of one or more embodiments of this specification are intended to be within the scope of the claims of this specification.

Claims

1. A small and micro loan approval method based on a large language model, characterized by: Specifically include: Recording problems in the credit report during the approval process, determining the associated report content for different problem points based on the type of the problem point, and proceeding to the next step when it is determined that the risk of the credit report meets the requirements based on the associated report content for different problem points; Based on the credit report and the problem points in the credit report, a large language model is used to perform identification processing to obtain potential problem points, and a target search source is determined based on the information association between different search sources and the potential problem points and the update status of different search sources; Determining deviations in supplementary search information for different potential risk points based on the search processing results of the target search source, and when determining that the search plan needs to be adjusted based on the deviations, obtaining supplementary search information for the credit report based on the full set of search sources; An evaluation conclusion of the credit report is output based on the potential problem points and the corresponding supplementary search information.

2. The micro-credit approval method based on a large language model as claimed in claim 1, characterized in that: Problem points in the approval process are determined based on problems in the approval process of relevant personnel of the credit report.

3. The micro-credit approval method based on a large language model as claimed in claim 1, characterized in that: Determine that the risk of the credit report meets the requirements, including: Determining the content of the associated reports in different subunits of the credit report based on the content of the associated reports of the problem points; Determining problematic subunits of the associated report content based on constituent data of the associated report content in different subunits; Whether the risk of the credit report meets the requirements is determined by the number of the problem sub-units.

4. The micro-credit approval method based on a large language model as claimed in claim 1, characterized in that: When the risk of the credit report does not meet the requirements, supplementary search information of the credit report is obtained based on all indicators and all search sources.

5. The micro-loan approval method based on a large language model as claimed in claim 1, characterized in that: Determine if adjustments to the search plan are necessary, including: Determine the deviation between the supplementary search information and the associated report content of the different potential risk points in the credit report based on the deviation between the supplementary search information of the target data source; Determining the number of potential risk points in the deviated associated report content based on the deviation between the associated report content and the supplementary search information; Determine whether the retrieval plan needs to be adjusted based on the number of potential risk points in the deviated associated report content.

6. The micro-loan approval method based on a large language model as claimed in claim 5, characterized in that: When there is no need to adjust the retrieval plan, the potential risk points and the supplementary retrieval information of the potential risk points in the target data source are used as input, and the output results of the large model are used to determine the evaluation conclusion of the credit report.

Images