Default prediction method and device, electronic equipment and storage medium
By acquiring and processing different types of descriptive data, extracting structured data features and combining risk assessment information to generate default prediction prompt information, the problem of inaccurate default prediction in the existing technology is solved, and the accuracy and efficiency of default prediction are improved.
Patent Information
- Application Number
- CN202411218487.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-02
- Publication Date
- 2024-12-27
AI Technical Summary
The default prediction in the prior art is not accurate enough to effectively evaluate the default risk of the object.
By obtaining the first and second types of description data of the object to be predicted, structured data features are extracted, and default prediction is made based on these features, combining the initial default prediction results and the initial prompt information for risk assessment information processing, the target default prediction results are generated. To improve the accuracy of default prediction.
By guiding the prediction process with semantic information, the accuracy and efficiency of default prediction are significantly improved and the processing capability of complex data is enhanced.
Smart Images
Figure CN119204301A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a default prediction method, device, electronic device and storage medium. Background Art
[0002] Among the many banking businesses, it is usually necessary to predict the default of data content (such as public credit, etc.) applied by various objects (such as users or enterprises) to assess the default risk of the objects. In related technologies, the default risk is usually predicted based on some basic information provided by the objects. In this way, the default prediction is not accurate enough. Summary of the invention
[0003] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.
[0004] To this end, the first purpose of the present disclosure is to propose a default prediction method that can improve the accuracy of default prediction.
[0005] The second objective of the present disclosure is to provide a default prediction device.
[0006] A third objective of the present disclosure is to provide an electronic device.
[0007] A fourth objective of the present disclosure is to provide a computer-readable storage medium.
[0008] A fifth object of the present disclosure is to provide a computer program product.
[0009] To achieve the above-mentioned purpose, the first aspect of the present disclosure proposes a default prediction method, comprising: obtaining first-category description data and second-category description data of an object to be predicted, wherein the data types of the first-category description data and the second-category description data are different, and the second-category description data at least include: risk assessment information of the application data content of the object to be predicted; obtaining initial prompt information for default prediction; extracting structured data features of the object to be predicted from the first-category description data; performing default prediction on the application data content of the object to be predicted according to the structured data features, and obtaining an initial default prediction result; processing the initial prompt information according to the first-category description data, the second-category description data and the initial default prediction result, and obtaining default prediction prompt information; and performing default prediction on the application data content of the object to be predicted according to the default prediction prompt information, and obtaining a target default prediction result.
[0010] To achieve the above-mentioned purpose, the second aspect of the present disclosure proposes a default prediction device, including: a first acquisition module, used to acquire a first type of description data and a second type of description data of an object to be predicted, wherein the data types of the first type of description data and the second type of description data are different, and the second type of description data at least includes: risk assessment information of the application data content of the object to be predicted; a second acquisition module, used to acquire initial prompt information for default prediction; an extraction module, used to extract structured data features of the object to be predicted from the first type of description data; a first prediction module, used to perform default prediction on the application data content of the object to be predicted according to the structured data features, and obtain an initial default prediction result; a processing module, used to process the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result, and obtain default prediction prompt information; and a second prediction module, used to perform default prediction on the application data content of the object to be predicted according to the default prediction prompt information, and obtain a target default prediction result.
[0011] To achieve the above-mentioned purpose, the third aspect embodiment of the present disclosure proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory, so as to implement a default prediction method proposed in the first aspect embodiment of the present disclosure.
[0012] To achieve the above-mentioned purpose, the fourth aspect embodiment of the present disclosure proposes a computer-readable storage medium, in which computer execution instructions are stored. When the computer execution instructions are executed by a processor, they are used to implement a default prediction method proposed in the first aspect embodiment of the present disclosure.
[0013] To achieve the above-mentioned purpose, the fifth aspect of the present disclosure proposes a computer program product, including a computer program, which, when executed by a processor, implements the default prediction method proposed in the first aspect of the present disclosure.
[0014] The default prediction method, device, electronic device, non-transitory computer-readable storage medium storing computer instructions, and computer program product provided by the present disclosure obtain the first type of description data and the second type of description data of the object to be predicted, and the initial prompt information of the default prediction, and extract the structured data features of the object to be predicted from the first type of description data, perform default prediction on the application data content of the object to be predicted according to the structured data features, obtain the initial default prediction result, process the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result, obtain the default prediction prompt information, and perform default prediction on the application data content of the object to be predicted according to the default prediction prompt information, and obtain the target default prediction result. Since the default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, when the default prediction is guided by the default prediction prompt information to perform default prediction on the application data content of the object to be predicted, the accuracy of the default prediction can be improved.
[0015] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0017] Figure 1 A flowchart of a default prediction method provided by an embodiment of the present disclosure;
[0018] Figure 2 A flowchart of another default prediction method provided by an embodiment of the present disclosure;
[0019] Figure 3 is a schematic diagram of a default prediction process in an embodiment of the present disclosure;
[0020] Figure 4 A schematic diagram of the structure of a default prediction device provided by an embodiment of the present disclosure;
[0021] Figure 5 It is a schematic structural diagram of an electronic device shown in an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION
[0022] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.
[0023] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this disclosure are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions. The acquisition, transmission, storage, use and processing of data in the technical solution of this disclosure are in compliance with the relevant provisions of national laws and regulations.
[0024] It should be noted that in the embodiments of the present disclosure, certain software, components, models and other existing solutions in the industry may be mentioned, which should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present disclosure, but it does not mean that the applicant has or will necessarily use the solution.
[0025] Figure 1 A flowchart of a default prediction method provided in an embodiment of the present disclosure.
[0026] This embodiment takes the default prediction method being configured in a default prediction device as an example. In this embodiment, the default prediction method can be configured in the default prediction device, and the default prediction device can be set in a server or in an electronic device, without limitation.
[0027] This embodiment takes the default prediction method configured in an electronic device as an example, wherein the electronic device is a hardware device such as a smart phone, a tablet computer, a personal digital assistant, an e-book, etc. having various operating systems.
[0028] It should be noted that the execution subject of the embodiment of the present disclosure may be, for example, a central processing unit (CPU) in a server or electronic device in terms of hardware, or may be, for example, a related background service in a server or electronic device in terms of software, without limitation.
[0029] like Figure 1 As shown, the default prediction method includes:
[0030] S101: Acquire first-category description data and second-category description data of an object to be predicted, wherein the data types of the first-category description data and the second-category description data are different, and the second-category description data at least includes: risk assessment information of application data content of the object to be predicted.
[0031] The object for which default prediction is to be performed may be referred to as an object to be predicted, such as a user, an enterprise, and the like.
[0032] The first type of description data and the second type of description data are respectively used to describe the object to be predicted. The first type of description data and the second type of description data can be used to describe the object to be predicted based on different dimensions. The data types of the first type of description data and the second type of description data are different. Thus, the object to be predicted can be described based on description data of different data types.
[0033] The second type of description data at least includes: risk assessment information of the application data content of the object to be predicted. The risk assessment information is used to describe the risk situation that may be caused by the application data content of the object to be predicted. Of course, the second type of description data can also include any other possible forms of description data, such as the default prediction results of the historical application of the data content, the risk assessment information of other objects associated with the object to be predicted, etc., without limitation.
[0034] In some embodiments, the first type of description data is structured data, and the second type of description data is text format data. Thus, it is supported to combine structured data and text format data to predict the default risk of the object to be predicted.
[0035] In some embodiments, in the process of obtaining the first category of descriptive data of the object to be predicted, application information submitted by the object to be predicted can be received, and the application information is used for the application data content of the object to be predicted, and the first category of descriptive data can be parsed from the application information; or an identifier of the object to be predicted can be obtained, and pre-stored first category of descriptive data can be obtained based on the identifier; or the first category of descriptive data can be obtained based on any other possible method.
[0036] In some embodiments, in the process of obtaining the second type of descriptive data of the object to be predicted, descriptive semantic features can be extracted from the first type of descriptive data, and the risk of the application data content of the object to be predicted can be evaluated based on the descriptive semantic features to obtain risk assessment information, and the second type of descriptive data can be generated based on the risk assessment information, thereby improving the description accuracy of the second type of descriptive data and supporting the improvement of default prediction accuracy.
[0037] Of course, the second type of description data of the object to be predicted can also be obtained based on any other possible method, such as manually referring to the first type of description data to conduct a risk assessment on the object to be predicted, obtaining risk assessment information, and generating the second type of description data based on the risk assessment information. There is no limitation to this.
[0038] Optionally, in some embodiments, the first category of descriptive data includes at least one of the following: basic attribute information, industry classification information, historical behavior characteristics, usage information of the applied data content, and report information, without limitation.
[0039] Optionally, in some embodiments, the second type of description data may be, for example, approval opinions, approval opinion text data, risk assessment text, which is used to describe the text description or text content of the risk of the object to be predicted, without limitation.
[0040] S102: Obtaining initial prompt information for default prediction.
[0041] The initial prompt information may be pre-configured and may be regarded as a configuration template for generating default prediction prompt information.
[0042] In some embodiments, in the process of obtaining the initial prompt information of the default prediction, the pre-defined initial prompt information can be obtained; or the initial prompt information can be dynamically configured; or the initial prompt information of the default prediction can be obtained based on any other possible method, such as modeling method, labeling method, etc., without limitation.
[0043] S103: Extracting structured data features of the object to be predicted from the first type of description data.
[0044] After obtaining the first type of description data, the structured data features of the object to be predicted can be extracted from the first type of description data. For example, various structured fields can be provided from the first type of description data, and the field values of the structured fields can be extracted. Then, the structured data features can be obtained by combining the field values of multiple structured fields; or the first type of description data can be converted into a structured vector to obtain a structured feature vector, and the structured feature vector can be used as a structured data feature, and there is no restriction on this.
[0045] In the above, structured data features of the object to be predicted are extracted from the first category of descriptive data. Since the extracted structured data features can not only carry the semantic information of the first category of descriptive data, but the data volume of the structured data features is relatively small compared to the first category of descriptive data, while supporting the improvement of the accuracy of default prediction, it can also reduce the amount of data processed by default prediction, and support the improvement of the processing efficiency of default prediction.
[0046] S104: Perform default prediction on the application data content of the prediction object according to the structured data characteristics to obtain an initial default prediction result.
[0047] Among them, the initial default prediction result can be a default prediction result obtained based on the prediction of the first type of descriptive data. The initial default prediction result can be, for example, an initial default probability. The higher the initial default probability, the higher the risk of default of the application data content of the object to be predicted. Correspondingly, the lower the initial default probability, the lower the risk of default of the application data content of the object to be predicted. Of course, the initial default prediction result can also be any other possible form of default prediction result, without limitation.
[0048] After obtaining the first type of description data of the object to be predicted, the structured data features of the object to be predicted can be determined by referring to the first type of description data, and then the structured data features are used to perform default prediction on the application data content of the object to be predicted to obtain an initial default prediction result.
[0049] In some embodiments, in the process of implementing default prediction for the application data content of the prediction object based on structured data features and obtaining an initial default prediction result, the default prediction for the application data content of the prediction object can be implemented by, for example, referring to the structured data features and default prediction analysis rules; or the structured data features can be processed based on artificial intelligence to predict the default of the application data content of the prediction object, or any other possible method can be used to implement default prediction for the application data content of the prediction object based on structured data features to obtain an initial default prediction result, without limitation.
[0050] S105: Processing the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain default prediction prompt information.
[0051] After obtaining the second type of description data and obtaining the initial default prediction result based on the first type of description data, the initial prompt information can be processed in combination with the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information. In other words, the default prediction prompt information is obtained by processing the initial prompt information based on the first type of description data, the second type of description data and the initial default prediction result. The default prediction prompt information can carry the respective semantic information of the first type of description data, the second type of description data and the initial default prediction result, so that when the default prediction of the application data content of the prediction object is guided based on the default prediction prompt information, the accuracy of the default prediction can be improved.
[0052] In some embodiments, in the process of processing the initial prompt information according to the first category of description data, the second category of description data and the initial default prediction result, the description semantic features can be extracted from the first category of description data, and the description semantic features can be extracted from the second category of description data, and the initial prompt information can be described using each description semantic feature and the initial default prediction result to obtain the default prediction prompt information.
[0053] S106: Perform default prediction on the application data content of the prediction object according to the default prediction prompt information to obtain a target default prediction result.
[0054] After the default prediction prompt information is generated, the default prediction can be performed on the application data content of the prediction object according to the default prediction prompt information to obtain the target default prediction result. In the process of realizing the default prediction of the application data content of the prediction object according to the default prediction prompt information, the default prediction of the application data content of the prediction object can be performed by referring to the default prediction prompt information and the default prediction analysis rules, or the default prediction prompt information can be processed based on artificial intelligence to perform the default prediction of the application data content of the prediction object, or any other possible method can be used to realize the default prediction of the application data content of the prediction object according to the default prediction prompt information to obtain the target default prediction result, without limitation.
[0055] In this embodiment, by obtaining the first type of description data and the second type of description data of the object to be predicted, and the initial prompt information of the default prediction, the structured data features of the object to be predicted are extracted from the first type of description data, and the default prediction is performed on the application data content of the object to be predicted according to the structured data features to obtain the initial default prediction result, the initial prompt information is processed according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information, and the default prediction is performed on the application data content of the object to be predicted according to the default prediction prompt information to obtain the target default prediction result. Since the default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, when the default prediction is performed on the application data content of the object to be predicted based on the default prediction prompt information, the accuracy of the default prediction can be improved.
[0056] Figure 2 A flowchart of another default prediction method provided by an embodiment of the present disclosure.
[0057] like Figure 2 As shown, the default prediction method includes:
[0058] S201: Acquire first-category description data and second-category description data of the object to be predicted, wherein the data types of the first-category description data and the second-category description data are different, and the second-category description data at least includes: risk assessment information of the application data content of the object to be predicted.
[0059] For the description of S201, please refer to the above embodiment, which will not be repeated here.
[0060] S202: Obtaining initial prompt information for default prediction.
[0061] S203: Extracting structured data features of the object to be predicted from the first type of description data.
[0062] After obtaining the first type of description data, the structured data features of the object to be predicted can be extracted from the first type of description data. For example, various structured fields can be provided from the first type of description data, and the field values of the structured fields can be extracted. Then, the structured data features can be obtained by combining the field values of multiple structured fields; or the first type of description data can be converted into a structured vector to obtain a structured feature vector, and the structured feature vector can be used as a structured data feature, and there is no restriction on this.
[0063] In the above, structured data features of the object to be predicted are extracted from the first category of descriptive data. Since the extracted structured data features can not only carry the semantic information of the first category of descriptive data, but the data volume of the structured data features is relatively small compared to the first category of descriptive data, while supporting the improvement of the accuracy of default prediction, it can also reduce the amount of data processed by default prediction, and support the improvement of the processing efficiency of default prediction.
[0064] S204: Processing structured data features based on the default prediction model to perform default prediction on the application data content of the prediction object to obtain an initial default prediction result; wherein the default prediction model has modeled and learned the mapping relationship between the structured data features and the initial default prediction result.
[0065] After extracting the structured data features of the object to be predicted from the first type of descriptive data, the structured data features can be processed based on the default prediction model to perform default prediction on the application data content of the object to be predicted and obtain an initial default prediction result.
[0066] Among them, the default prediction model has modeled and learned the mapping relationship between structured data features and initial default prediction results.
[0067] In some embodiments, the structured data features of the sample (which can be extracted from the first category of descriptive data of the sample) and the annotated default prediction results corresponding to the structured data features of the sample can be obtained to train the initial prediction model. The prediction model has the ability to analyze the structured data features of the sample to obtain the default prediction results. The initial prediction model can be iteratively trained multiple times until the prediction model obtained by the training converges. The prediction model obtained by the training is used as the default prediction model, and there is no restriction on this.
[0068] In some embodiments, structured data features may be input into a default prediction model so that the default prediction model performs default prediction on the application data content of the prediction object based on the structured data features to obtain an initial default prediction result, without limitation.
[0069] S205: Processing the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain default prediction prompt information.
[0070] After obtaining the second type of description data and obtaining the initial default prediction result based on the first type of description data, the initial prompt information can be processed according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information. In other words, the default prediction prompt information is synthesized based on the first type of description data, the second type of description data and the initial default prediction result. The default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, so that when the default prediction of the application data content of the prediction object is guided based on the default prediction prompt information, the accuracy of the default prediction can be improved.
[0071] In some embodiments, in the process of processing the initial prompt information according to the first category of description data, the second category of description data and the initial default prediction result, the description semantic features can be extracted from the first category of description data, and the description semantic features can be extracted from the second category of description data, and the initial prompt information can be described using each description semantic feature and the initial default prediction result to obtain the default prediction prompt information.
[0072] Optionally, in some embodiments, in the process of processing the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information, the first prompt part and the first prompt area of the first prompt part corresponding to the first type of description data, the second prompt part and the second prompt area of the second prompt part corresponding to the second type of description data, and the third prompt part and the third prompt area of the third prompt part corresponding to the initial default prediction result may be determined from the initial prompt information, and the first type of description data is filled into the first prompt area of the initial prompt information, and the second type of description data is filled into the second prompt area of the initial prompt information, and the initial default prediction result is filled into the third prompt area of the initial prompt information to obtain the default prediction prompt information. Thus, the default prediction prompt information is generated quickly and accurately.
[0073] Among them, the initial prompt information can be pre-configured. The initial prompt information can be regarded as a configuration template for generating default prediction prompt information, and the configuration template can include a pre-set first prompt part, which refers to the prompt part corresponding to the first type of description data, and the first prompt part has a corresponding first prompt area, and the first prompt area can be a text area for filling the first type of description data. Correspondingly, the configuration template can also include a pre-set second prompt part and a third prompt part, and the second prompt part refers to the prompt part corresponding to the second type of description data, and the second prompt part has a corresponding second prompt area, and the second prompt area can be a text area for filling the second type of description data. The third prompt part refers to the prompt part corresponding to the initial default prediction result, and the third prompt part has a corresponding third prompt area, and the third prompt area can be a text area for filling the initial default prediction result.
[0074] Therefore, after obtaining the first type of description data, the second type of description data and the initial default prediction result, the first type of description data, the second type of description data and the initial default prediction result can be filled into the corresponding prompt area in the initial prompt information, and after filling, the filled prompt information is used as the default prediction prompt information. Since the initial prompt information is pre-configured, after filling, the default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, which supports improving the accuracy and generation efficiency of the default prediction prompt information, thereby greatly supporting improving the overall default prediction efficiency and accuracy.
[0075] S206: Inputting the default prediction prompt information into the big model, so that the big model performs default prediction on the application data content of the prediction object based on the default prediction prompt information, and obtains the target default prediction result; wherein, the big model has modeled and learned the mapping relationship between the default prediction prompt information and the target default prediction result.
[0076] After generating the default prediction prompt information, the default prediction prompt information can be processed in combination with a large model (also called a large language model) to perform default prediction on the application data content of the prediction object to obtain a target default prediction result.
[0077] In some embodiments, the big model can be pre-trained, and sample prediction prompt information can be obtained (the generation process of the sample prediction prompt information can be based on the first category description data of the sample, the second category description data of the sample, and the initial default prediction result of the sample), and the labeled default prediction results corresponding to the sample prediction prompt information are used to train the initial big model. The initial big model has the ability to analyze the sample prediction prompt information to obtain the default prediction results, and the initial big model can be iteratively trained multiple times until the trained big model converges, and there is no restriction on this.
[0078] After obtaining the default prediction prompt information, the default prediction prompt information can be input into the large model. The large model can perform a default prediction on the application data content of the object to be predicted based on the default prediction prompt information to obtain a target default prediction result. The target default prediction result can be a more accurate prediction probability value. The higher the prediction probability value, the higher the risk of default of the application data content of the object to be predicted. Correspondingly, the lower the prediction probability value, the lower the risk of default of the application data content of the object to be predicted. Of course, the target default prediction result can also be any other possible form of default prediction result. For example, the target default prediction result is 0 or 1, 0 indicates that the application data content of the object to be predicted has a default risk, and 1 indicates that the application data content of the object to be predicted does not have a default risk. There is no restriction on this.
[0079] In this embodiment, by obtaining the first type of description data and the second type of description data of the object to be predicted, and the initial prompt information of the default prediction, the structured data features of the object to be predicted are extracted from the first type of description data, the default prediction is performed on the application data content of the object to be predicted according to the structured data features, and the initial default prediction result is obtained, the initial prompt information is processed according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information, and the default prediction is performed on the application data content of the object to be predicted according to the default prediction prompt information to obtain the target default prediction result. Since the default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, when the default prediction is guided by the default prediction prompt information to perform the default prediction on the application data content of the object to be predicted, the accuracy of the default prediction can be improved. When the structured data features of the object to be predicted are extracted from the first type of description data, since the extracted structured data features can not only carry the semantic information of the first type of description data, but also the data volume of the structured data features is relatively small compared to the first type of description data, while supporting the improvement of the accuracy of the default prediction, the amount of data processed by the default prediction can also be reduced, and the processing efficiency of the default prediction can be improved. After obtaining the first type of description data, the second type of description data and the initial default prediction result, the first type of description data, the second type of description data and the initial default prediction result can be filled into the corresponding prompt area in the initial prompt information respectively. After filling, the filled prompt information is used as the default prediction prompt information. Since the initial prompt information is pre-configured, after filling, the default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, which supports improving the accuracy and generation efficiency of the default prediction prompt information, thereby greatly supporting improving the overall default prediction efficiency and accuracy.
[0080] The examples for the above embodiment are as follows:
[0081] In the disclosed embodiment, default prediction is achieved by combining a machine learning model and a large-scale pre-trained language model (large model), improving the ability of text information (an optional example of the second type of descriptive data) in the approval process, thereby more comprehensively evaluating the financial risk of the enterprise (an optional example of the object to be predicted). It is possible to more effectively integrate and utilize text information (an optional example of the second type of descriptive data) in the approval process to improve the accuracy of default prediction. It not only improves the efficiency and accuracy of credit approval, but also enhances the model's ability to process complex data.
[0082] In the disclosed embodiments, based on the results of an existing machine learning model (an optional example of a default prediction model) (an optional example of an initial default prediction result), combined with a large-scale pre-trained language model (large model), prompts (an optional example of default prediction prompt information) can be constructed and fine-tuned using low-rank adaptation (LoRA) (LoRA is a new method for adapting large pre-trained language models to downstream tasks such as text classification and sentiment analysis), and fully combined with text information in the approval process, data such as the basic industry of the enterprise, the scale of the enterprise, and the type of enterprise, so as to improve the success rate of predicting enterprise defaults in the public credit approval process.
[0083] like Figure 3 As shown, Figure 3 It is a schematic diagram of the default prediction process in the embodiment of the present disclosure.
[0084] Part I: Data preparation and preprocessing.
[0085] Preparation of structured data: Collect structured data such as the company's basic information, industry, historical performance, credit usage and financial statements (an optional example of the first category of descriptive data).
[0086] Preparation of text data: Collect text information generated during the approval process of the enterprise's credit approval process. The approver mentioned in the text information during the approval process that "as of the end of March 2022, the owner's equity of the enterprise was -20.1351 million yuan, and the compliance of the source of the deposit for the low credit risk limit of 200 million yuan reported this time is questionable."
[0087] Data preprocessing: For structured data, data cleaning and standardization steps can be performed.
[0088] Part 2: Feature Engineering.
[0089] For the above-mentioned enterprise, key indicators such as debt-to-asset ratio and operating income growth rate can be extracted from its financial statements (an optional example of structured data features). Other features include enterprise revenue, etc.
[0090] Part III: Model training and fine-tuning.
[0091] Machine learning model training: Use machine learning algorithms such as xgboost or lightgbm, combined with structured data features, to train the default prediction model. Assume that for the above case company, the xgboost model (an optional example of a default prediction model) gives a default probability score of 0.65 (an optional example of the initial default prediction result).
[0092] Fine-tuning the large model: Fine-tune the large-scale pre-trained language model using historical default data (an optional example of the default prediction result of the sample) and approval opinion data (an optional example of the default prediction prompt information of the sample). This improves the performance of the large model in processing text information, especially in parsing the risk prompts in the approval opinions (an optional example of risk assessment information). The corresponding prompts (an optional example of the default prediction prompt information of the sample) can be constructed:
[0093] "You are performing a default prediction task for corporate credit approval. The default probability score of the company according to the existing xgboost model is 0.65. At the same time, the approval comment of the company is: As of the end of March 2022, the owner's equity of the company is -20.1351 million yuan. The compliance of the source of the deposit for the low credit risk limit of 200 million yuan reported this time is questionable. The name of the company is: ...; the industry it belongs to is: ...; the basic information of the company is: ...; please combine all the above information to judge whether the company will default. If yes, please reply 1, if no, please reply 0."
[0094] In addition, LoRA can be used to fine-tune the large model in combination with historical default data. The core idea of LoRA is to inject a trainable rank decomposition matrix into each layer of the pre-trained large model. The benefit is that the number of trainable parameters for downstream tasks is significantly reduced. The fine-tuned large model lora weight matrix is obtained and can be used for prediction. During the fine-tuning process, the rank of the LoRA adapter can be set to 16, and 160,000 sample data are used for training, with a training cycle of 30 epochs.
[0095] Part 4: Model prediction.
[0096] When predicting the model, first use the xgboost / lightglm model (an optional example of the default prediction model) to run the score, and at the same time input the natural language text of the company's basic information to construct the same prompts, and then use the fine-tuned large model to directly give the prediction results. Assuming that the xgboost prediction probability is x (an optional example of the initial default prediction result), construct the same prompts: "You are doing a corporate credit approval default prediction task. The default probability score of the company after the existing xgboost model is x, and the approval comment of the company is: {comment}. The name of the company is:...; the industry it belongs to is:...; the basic information of the company is:...; please combine all the above information to determine whether the user will default. If yes, please reply 1, if not, please reply 0." Go to query the fine-tuned large model to get the default prediction result. After experiments, the performance of the large model combined with the xgboost model on the validation set is 3% higher than that of using xgboost alone, from 89% to 91%.
[0097] Figure 4 A schematic diagram of the structure of a default prediction device provided in an embodiment of the present disclosure.
[0098] like Figure 4 As shown, the default prediction device 40 comprises:
[0099] The first acquisition module 401 is used to acquire the first type of description data and the second type of description data of the object to be predicted, wherein the data types of the first type of description data and the second type of description data are different, and the second type of description data at least includes: risk assessment information of the application data content of the object to be predicted.
[0100] The second acquisition module 402 is used to acquire initial prompt information of default prediction.
[0101] The extraction module 403 is used to extract the structured data features of the object to be predicted from the first type of description data.
[0102] The first prediction module 404 is used to perform default prediction on the application data content of the prediction object according to the structured data characteristics to obtain an initial default prediction result.
[0103] The processing module 405 is used to process the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information.
[0104] The second prediction module 406 is used to perform default prediction on the application data content of the prediction object according to the default prediction prompt information to obtain a target default prediction result.
[0105] It should be noted that the above explanation of the default prediction method is also applicable to the default prediction device of this embodiment and will not be repeated here.
[0106] In this embodiment, by obtaining the first type of description data and the second type of description data of the object to be predicted, and the initial prompt information of the default prediction, the structured data features of the object to be predicted are extracted from the first type of description data, and the default prediction is performed on the application data content of the object to be predicted according to the structured data features to obtain the initial default prediction result, the initial prompt information is processed according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information, and the default prediction is performed on the application data content of the object to be predicted according to the default prediction prompt information to obtain the target default prediction result. Since the default prediction prompt information can carry the semantic information of the first type of description data, the second type of description data and the initial default prediction result, when the default prediction is performed on the application data content of the object to be predicted based on the default prediction prompt information, the accuracy of the default prediction can be improved.
[0107] As an example, Figure 5 is a schematic diagram of the structure of an electronic device shown in an exemplary embodiment of the present disclosure, such as Figure 5 As shown, the electronic device 500 may further include:
[0108] The memory 510 and the processor 520, a bus 530 connecting different components (including the memory 510 and the processor 520), the memory 510 stores a computer program, and when the processor 520 executes the program, the default prediction method described in the embodiment of the present disclosure is implemented.
[0109] Bus 530 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnect (PCI) bus.
[0110] The electronic device 500 typically includes a variety of electronic device readable media. These media can be any available media that can be accessed by the electronic device 500, including volatile and non-volatile media, removable and non-removable media.
[0111] The memory 510 may also include computer system readable media in the form of volatile memory, such as random access memory (RAM) 540 and / or cache memory 550. The electronic device 500 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 560 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, usually called a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 530 via one or more data medium interfaces. The memory 510 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present disclosure.
[0112] A program / utility 580 having a set (at least one) of program modules 570 may be stored, for example, in the memory 510, such program modules 570 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 570 generally perform the functions and / or methods of the embodiments described in the present disclosure.
[0113] The electronic device 500 may also communicate with one or more external devices 590 (e.g., keyboard, pointing device, display 591, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 500, and / or communicate with any device that enables the electronic device 500 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed via an input / output (I / O) interface 592. In addition, the electronic device 500 may also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN) and / or public network, such as the Internet) via a network adapter 593. As shown, the network adapter 593 communicates with other modules of the electronic device 500 via a bus 530. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0114] The processor 520 executes various functional applications and data processing by running the programs stored in the memory 510 .
[0115] It should be noted that the implementation process and technical principles of the electronic device of this embodiment refer to the aforementioned explanation of the default prediction method of the embodiment of the present disclosure, and will not be repeated here.
[0116] In order to implement the above embodiments, the present disclosure also proposes an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.
[0117] In order to implement the above embodiments, the present disclosure further proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.
[0118] In order to implement the above embodiments, the present disclosure also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.
[0119] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this disclosure shall comply with the relevant laws and regulations and shall not violate public order and good morals.
[0120] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.
[0121] The present disclosure anticipates providing implementation schemes for users to selectively block the use or access of personal information data. That is, the present disclosure anticipates providing hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, risks can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.
[0122] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.
[0123] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0124] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.
[0125] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.
[0126] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0127] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0128] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0129] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.
Claims
1. A default prediction method, characterized in that: The method comprises: Acquire first-category description data and second-category description data of the object to be predicted, wherein the data types of the first-category description data and the second-category description data are different, and the second-category description data at least includes: risk assessment information of the application data content of the object to be predicted; Get initial alert information for default prediction; Extracting structured data features of the object to be predicted from the first type of description data; Performing default prediction on the application data content of the object to be predicted according to the structured data features to obtain an initial default prediction result; Processing the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain default prediction prompt information; and A default prediction is performed on the application data content of the object to be predicted according to the default prediction prompt information to obtain a target default prediction result.
2. The method according to claim 1, characterized in that The first type of description data is structured data, and the second type of description data is text format data.
3. The method according to claim 1, characterized in that The step of performing default prediction on the application data content of the object to be predicted according to the structured data features to obtain an initial default prediction result includes: The structured data features are processed based on a default prediction model to perform a default prediction on the application data content of the object to be predicted, thereby obtaining the initial default prediction result; wherein the default prediction model has modeled and learned the mapping relationship between the structured data features and the initial default prediction result.
4. The method according to claim 1, characterized in that: The step of processing the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain the default prediction prompt information includes: Determine from the initial prompt information a first prompt part corresponding to the first type of description data and a first prompt area of the first prompt part, a second prompt part corresponding to the second type of description data and a second prompt area of the second prompt part, and a third prompt part corresponding to the initial default prediction result and a third prompt area of the third prompt part; The first type of description data is filled into the first prompt area of the initial prompt information, and the second type of description data is filled into the second prompt area of the initial prompt information, and the initial default prediction result is filled into the third prompt area of the initial prompt information to obtain the default prediction prompt information.
5. The method according to claim 1, characterized in that The step of performing default prediction on the application data content of the object to be predicted according to the default prediction prompt information to obtain a target default prediction result includes: The default prediction prompt information is input into the big model so that the big model performs a default prediction on the application data content of the object to be predicted based on the default prediction prompt information to obtain a target default prediction result; wherein, the big model has modeled and learned the mapping relationship between the default prediction prompt information and the target default prediction result.
6. The method according to claim 1, characterized in that The first category of descriptive data includes at least one of the following: basic attribute information, industry classification information, historical behavior characteristics, usage information of the applied data content, and report information.
7. The method according to claim 1, characterized in that Obtain the second type of descriptive data of the object to be predicted, including: Extracting descriptive semantic features from the first type of descriptive data; Based on the descriptive semantic features, the risk of the application data content of the object to be predicted is evaluated to obtain the risk evaluation information; Based on the risk assessment information, the second type of description data is generated.
8. A default prediction device, characterized in that: The device comprises: A first acquisition module is used to acquire first-category description data and second-category description data of the object to be predicted, wherein the data types of the first-category description data and the second-category description data are different, and the second-category description data at least includes: risk assessment information of the application data content of the object to be predicted; The second acquisition module is used to obtain initial prompt information of default prediction; An extraction module, used to extract the structured data features of the object to be predicted from the first type of description data; A first prediction module, configured to perform default prediction on the application data content of the object to be predicted according to the structured data features, and obtain an initial default prediction result; a processing module, configured to process the initial prompt information according to the first type of description data, the second type of description data and the initial default prediction result to obtain default prediction prompt information; and The second prediction module is used to perform default prediction on the application data content of the object to be predicted according to the default prediction prompt information to obtain a target default prediction result.
9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.
11. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 7 when being executed by a processor.