Artificial intelligence-based discriminant risk control method, apparatus and device, and medium
Through a case-based risk control method based on artificial intelligence, data layering and document conversion of user information sets are generated, user data document sets are embedded in time and multi-dimensional similarity matching, which solves the problem of poor flexibility in traditional risk control approval methods and achieves efficient and flexible risk control approval.
Patent Information
- Application Number
- CN202510333744.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-20
- Publication Date
- 2025-07-29
AI Technical Summary
Due to the inconsistent rules of traditional risk control approval methods, poor flexibility and difficult to adapt to complex situations.
A case-based risk control method based on artificial intelligence is adopted, and data layering and document conversion of user information sets are generated, user data document sets are generated, and time-series embedding operations are carried out, a case library of approval is established, multi-dimensional similarity matching and autoregression analysis are performed, and risk control decision results are generated.
It improves the efficiency and flexibility of risk control approval, enhances the accuracy and scope of approval results, and reduces human resource consumption.
Smart Images
Figure CN120387666A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of data analysis and fintech, and in particular to a case-based risk control method, device, equipment and medium based on artificial intelligence. Background Art
[0002] Risk control approval is a process in the business process that, based on a risk assessment model and review rules, reviews, judges, and decides whether to pass relevant applications or behaviors such as loan, credit card approval, insurance underwriting, business cooperation, etc. The core of risk control approval lies in identifying, evaluating, controlling, and coping with risks to ensure the interests of the enterprise and the security of the business.
[0003] The traditional risk control approval method is a risk control approval method based on preset rules, that is, by setting a series of fixed rules and making approval decisions based on the degree of compliance of user data with the rules. In practical applications, due to the lack of uniformity of risk control rules, which often change, the risk control approval method based on preset rules has problems of poor flexibility and weak ability to adapt to complex situations. Summary of the Invention
[0004] The present invention provides an artificial intelligence-based case-based risk control method, device, equipment and medium to solve the technical problem of low flexibility in risk control approval in the financial system due to different risk control approval rules.
[0005] In a first aspect, there is provided a case-based risk control method based on artificial intelligence, including:
[0006] Performing data stratification and document conversion on a pre-acquired user information set to obtain a user data document set;
[0007] Performing time series embedding operation on the user data document set to obtain a user document feature set;
[0008] Performing data standardization and database archiving on pre-acquired historical approval data to obtain an approval case library;
[0009] Performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set;
[0010] Performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set;
[0011] Generating a risk control decision result set according to the matching decision result set and the analysis decision result set.
[0012] In a second aspect, there is provided a case-based risk control device based on artificial intelligence, including:
[0013] A document conversion module, configured to perform data stratification and document conversion on a pre-acquired user information set to obtain a user data document set;
[0014] A feature extraction module, configured to perform a time series embedding operation on the user data document set to obtain a user document feature set;
[0015] A data archiving module, configured to perform data standardization and database archiving on pre-acquired historical approval data to obtain an approval case library;
[0016] A matching decision module, configured to perform multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set;
[0017] An analysis decision module, configured to perform autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set;
[0018] A risk control decision module, configured to generate a risk control decision result set according to the matching decision result set and the analysis decision result set.
[0019] In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned case-based risk control method based on artificial intelligence is implemented.
[0020] In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned case-based risk control method based on artificial intelligence is implemented.
[0021] In the solutions implemented by the above-mentioned case-based risk control method, device, computer device, and storage medium based on artificial intelligence, data stratification and document conversion can be performed on a pre-acquired user information set to obtain a user data document set, which can complete primary data classification and sorting, and archive according to the importance of the data, realizing formatted data storage, facilitating subsequent data analysis, improving the efficiency of risk control approval. By performing a time series embedding operation on the user data document set to obtain a user document feature set, the user data document set can be converted into a feature form convenient for calculation, reducing the calculation dimension, improving the efficiency of risk control approval, and also facilitating subsequent feature matching. By performing data standardization and database archiving on pre-acquired historical approval data to obtain an approval case library, the archival storage of historical data can be realized, and subsequent rapid similarity retrieval and similarity calculation can be facilitated, improving the flexibility of risk control approval.
[0022] By performing multi-dimensional similarity matching on the user document feature set according to the approval case library, a matching decision result set can be obtained, which can perform similarity matching by combining features in three dimensions: core data, behavior features, and derivative data, thereby improving the accuracy of approval result matching, and can perform flexible approval result matching based on historical record data, improving the flexibility of risk control approval; by performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set, the risk control approval of the user document feature set can be realized by using the method of deep learning, thereby improving the flexibility of risk control approval. By generating a risk control decision result set according to the matching decision result set and the analysis decision result set, risk control approval can be performed based on feature matching and autoregressive models, expanding the scope of risk control approval, reducing human resource consumption, and improving the flexibility of risk control approval. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments of the present invention. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0024] Figure 1 is a schematic diagram of an application environment of a case-based risk control method based on artificial intelligence according to an embodiment of the present invention;
[0025] Figure 2 is a schematic flowchart of a case-based risk control method based on artificial intelligence according to an embodiment of the present invention;
[0026] Figure 3 is Figure 1 a schematic flowchart of a specific implementation manner of step S20 in
[0027] Figure 4 is Figure 1 a schematic flowchart of a specific implementation manner of step S30 in
[0028] Figure 5 is Figure 1 a schematic flowchart of a specific implementation manner of step S40 in
[0029] Figure 6 is a schematic structural diagram of a case-based risk control device based on artificial intelligence according to an embodiment of the present invention;
[0030] Figure 7 is a schematic structural diagram of a computer device according to an embodiment of the present invention;
[0031] It is another schematic structural diagram of a computer device in an embodiment of the present invention. Detailed implementation manners
[0032] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.
[0033] The case-based risk control method based on artificial intelligence provided by the embodiments of the present invention can be applied in, for example In the application environment, the client communicates with the server through the network. The server can perform data layering and document conversion on the pre-acquired user information set through the client to obtain a user data document set; perform a temporal embedding operation on the user data document set to obtain a user document feature set; perform data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library; perform multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set; perform autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set; generate a risk control decision result set according to the matching decision result set and the analysis decision result set. In the present invention, performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set includes: sequentially selecting the user document features in the user document feature set as target user document features, and sequentially selecting the historical document features in the historical document feature set in the approval case library as target historical document features; performing multi-dimensional similarity calculation on the target user document features and the target historical document features to obtain a multi-dimensional matching degree; aggregating the multi-dimensional matching degrees of all target historical document features in the approval case library into a multi-dimensional matching degree set; performing extreme value screening on the multi-dimensional matching degree set to obtain an extreme value matching degree, and taking the target historical document feature corresponding to the extreme value matching degree as a matching historical document feature; taking the user approval result corresponding to the matching historical document feature in the approval case library as a matching decision result; aggregating the matching decision results corresponding to all target user document features in the user document feature set into a matching decision result set, which can perform similarity matching by combining features in three dimensions of core data, behavioral features, and derivative data, thereby improving the accuracy of approval result matching, and can perform flexible approval result matching according to historical record data, improving the flexibility of risk control approval. Among them, the client can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of multiple servers. The present invention will be described in detail through specific embodiments below.
[0034] Please refer to as shown in FIG. is a flowchart of a case-based risk control method based on artificial intelligence provided by an embodiment of the present invention, including the following steps:
[0035] S10. Perform data layering and document conversion on the pre-acquired user information set to obtain a user data document set.
[0036] Specifically, the user information set refers to a set composed of collected user-related information. Each user information in the user information set includes the user's personal information, historical transaction records, and credit data. Among them, the personal information includes personal income information, the historical transaction records include repayment records, and the credit data includes information such as loan application history.
[0037] Specifically, each user data document in the user data document set is a document format storing the user information in the user information set in a hierarchical manner. The user data document includes three data layers: the core data layer, the behavior feature layer, and the derived data layer.
[0038] Specifically, the core data layer includes highly reliable data such as income certificates and asset certificates; key credit data such as credit records and repayment histories, and stability index data such as occupation information and working years. The data in the core data layer directly affects risk control decisions and requires strict structured storage and independent indexing.
[0039] Specifically, the behavior feature layer includes dynamic data such as consumption behavior and transaction frequency; behavior trajectory data such as device usage habits and location information, and social attribute data such as social networks and interest preferences. The data in the behavior feature layer is stored after being transformed into tags or indicators through feature engineering.
[0040] Specifically, the derived data layer includes various risk index data calculated based on the core data; user portrait tag data summarized based on behavior data, and decision support information data related to historical approvals. The data in the derived data layer is used for secondary processed data for assisting decisions.
[0041] In the embodiment of the present invention, the data layering and document conversion of the pre-acquired user information set to obtain the user data document set includes:
[0042] Performing data cleaning on the pre-acquired user information set to obtain a cleaned information set;
[0043] Selecting the cleaning information in the cleaned information set one by one as the target cleaning information, and performing core data matching on the target cleaning information to obtain target core data;
[0044] Performing behavior feature extraction on the target cleaning information to obtain target behavior features;
[0045] Performing risk index extraction on the target cleaning information to obtain target derived data;
[0046] Performing document conversion on the target core data, the target behavior features, and the target derived data to obtain a user data document;
[0047] Generate a user data document set based on the user data documents of all target cleaning information in the cleaning information set.
[0048] Specifically, the data cleaning includes removing duplicate or invalid data entries; filling in missing values using methods such as mean filling or pattern filling, unifying data units and formats, and ensuring data field consistency.
[0049] Specifically, the core data matching refers to matching high-credibility data such as income certificates and asset certificates, key credit data such as credit records and repayment histories, and stability index data such as occupation information and working years as target core data through keyword matching.
[0050] Specifically, the behavior feature extraction refers to extracting dynamic data such as consumption behavior and transaction frequency, behavior trajectory data such as device usage habits and location information, and social attribute data such as social networks and interest preferences as target behavior features through methods such as regular expressions combined with feature engineering.
[0051] Specifically, the risk index extraction refers to extracting various risk index data calculated based on core data, user portrait label data summarized based on behavior data, and decision support information data related to historical approvals as target derivative data through keyword matching.
[0052] Specifically, the document conversion refers to converting the target core data, the target behavior features, and the target derivative data into a document format with a hierarchical structure, such as converting into user data documents in document formats such as JSON, CSV, or data reports using tools such as Python Pandas.
[0053] In the embodiments of the present invention, through data stratification and document conversion of the pre-acquired user information set, a user data document set can be obtained to complete primary data classification and sorting, and file according to the importance of the data, realizing formatted data storage, facilitating subsequent data analysis, and improving the efficiency of risk control approval.
[0054] S20. Perform a time series embedding operation on the user data document set to obtain a user document feature set.
[0055] Specifically, each user document feature in the user document feature set is a feature vector corresponding to the user data document in the user data document set, used to reflect user data information.
[0056] In the embodiments of the present invention, referring to As shown, the performing a time series embedding operation on the user data document set to obtain a user document feature set includes:
[0057] S21. Select each user data document in the user data document set as the target data document, and perform time series splitting on the target data document to obtain a time series document sequence;
[0058] S22. Perform document word segmentation on the time series document sequence to obtain a time series document phrase sequence;
[0059] S23. Perform word vector conversion on the time series document phrase sequence to obtain a time series word vector group sequence;
[0060] S24. Perform position encoding on the time series word vector group sequence to obtain a time series word feature group sequence;
[0061] S25. Perform feature splicing according to the time series word feature group sequence to obtain a time series document feature sequence;
[0062] S26. Perform time series feature convolution on the time series document feature sequence to obtain user document features;
[0063] S27. Aggregate the user document features of all target data documents in the user data document set into a user document feature set.
[0064] Specifically, the document word segmentation refers to the process of decomposing a text into several smaller units. The word vector conversion is to map the words or sub-words obtained by word segmentation into a vector representation of a fixed dimension to capture semantic or syntactic information, so that the numerical representation that allows the machine to understand and operate on the text. The document word segmentation can be performed using the bidirectional maximum matching method, and the word vector conversion can be performed using encoding methods such as One Hot Encoding, Word2Vec, GloVe, and BERT.
[0065] Specifically, the position encoding is used to provide position information for each word in the sequence to make up for the deficiency of models such as Transformer in sensitivity to position information. The position encoding can be performed using methods such as ALiBi or dynamic position encoding.
[0066] Specifically, the feature splicing is the process of combining features from different sources or different dimensions to form a unified vector representation. The feature splicing can be performed using methods such as attention weighted splicing, weighted average splicing, or linear splicing.
[0067] Specifically, the time series feature convolution is the process of performing a convolution operation on data with a time dimension, aiming to capture local features and change patterns in the time series data. The time series feature convolution can be performed using methods such as pyramid convolution, dynamic convolution, or dilated convolution.
[0068] In the embodiments of the present invention, by performing a temporal embedding operation on the user data document set, a user document feature set is obtained, which can convert the user data document set into a feature form convenient for calculation, reduce the calculation dimension, improve the efficiency of risk control approval, and facilitate subsequent feature matching.
[0069] S30. Perform data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library.
[0070] Specifically, the historical approval data refers to user information processed by means such as manual approval in the past time period and the corresponding approval results, where the approval results include approval success, approval failure, etc.
[0071] In the embodiments of the present invention, as shown in performing data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library includes:
[0072] S31. Split the pre-acquired historical approval data by user to obtain a user approval data set;
[0073] S32. Extract a historical user data set and the user approval result set corresponding to the historical user data set from the user approval data set respectively;
[0074] S33. Perform data stratification and document conversion on the historical user data set to obtain a historical user document set;
[0075] S34. Perform a temporal embedding operation on the historical user document set to obtain a historical document feature set;
[0076] S35. Perform database archiving on the historical document feature set and the user approval result set to obtain an approval case library.
[0077] Specifically, the user split refers to splitting the historical approval data into the approval data corresponding to each user according to the unique identifier of the user, and each user approval data in the user approval data set refers to the user information corresponding to a user processed by means such as manual approval in the past time period and the corresponding approval result.
[0078] Specifically, each historical user data in the historical user data set corresponds to the user data in each user approval data in the user approval data set, and each user approval result in the user approval result set corresponds to the approval result of each user approval data in the user approval data set.
[0079] Specifically, the method for data layering and document conversion is the same as the method in step S10 above, and the method for the time series embedding operation is the same as the method in step S20 above, which will not be elaborated here.
[0080] Specifically, the database archiving means storing each historical document feature in the historical document feature set and the corresponding user approval result in the user approval result set in the database in a corresponding manner.
[0081] In the embodiment of the present invention, by performing data standardization and database archiving on the pre-acquired historical approval data, an approval case library is obtained, which can realize the archival storage of historical data, facilitate subsequent rapid similarity retrieval and similarity calculation, and improve the flexibility of risk control approval.
[0082] S40. Perform multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set.
[0083] Specifically, each matching decision result in the matching decision result set corresponds to the matching result of the risk control approval of the user corresponding to each user document feature in the user document feature set.
[0084] In the embodiment of the present invention, as shown in performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set includes:
[0085] S41. Select each user document feature in the user document feature set as a target user document feature one by one, and select each historical document feature in the historical document feature set in the approval case library as a target historical document feature one by one;
[0086] S42. Perform multi-dimensional similarity calculation on the target user document feature and the target historical document feature to obtain a multi-dimensional matching degree;
[0087] S43. Aggregate the multi-dimensional matching degrees of all target historical document features in the approval case library into a multi-dimensional matching degree set;
[0088] S44. Perform extreme value screening on the multi-dimensional matching degree set to obtain an extreme value matching degree, and use the target historical document feature corresponding to the extreme value matching degree as a matching historical document feature;
[0089] S45. Use the user approval result corresponding to the matching historical document feature in the approval case library as a matching decision result;
[0090] S46. Aggregate the matching decision results corresponding to all target user document features in the user document feature set into a matching decision result set.
[0091] Specifically, the extreme value screening refers to taking the multi-dimensional matching degree with the largest multi-dimensional matching degree value in the multi-dimensional matching degree set as the extreme value matching degree.
[0092] Specifically, the calculation of the multi-dimensional similarity between the target user document features and the target historical document features to obtain the multi-dimensional matching degree includes:
[0093] Performing feature-level splitting on the target user document features to obtain user core features, user behavior features, and user-derived features;
[0094] Calculating a matching weight combination according to the user core features, the user behavior features, and the user-derived features;
[0095] Performing feature-level splitting on the target historical document features to obtain historical core features, historical behavior features, and historical-derived features;
[0096] Calculating a core feature matching degree according to the user core features and the historical core features;
[0097] Calculating a behavior feature matching degree according to the user behavior features and the historical behavior features;
[0098] Calculating a derived feature matching degree according to the user-derived features and the historical-derived features;
[0099] Performing weighted summation on the core feature matching degree, the behavior feature matching degree, and the derived feature matching degree according to the matching weight combination to obtain the multi-dimensional matching degree.
[0100] Specifically, the feature-level splitting refers to splitting the features in the target user document features according to the hierarchical structure of the data stratification. The user core features correspond to some features corresponding to the core data layer in the target user document features, the user behavior features correspond to some features corresponding to the behavior feature layer in the target user document features, and the user-derived features correspond to some features corresponding to the derived data layer in the target user document features.
[0101] Specifically, the following formula can be used to calculate the matching weight combination:
[0102] w = softmax(MLP(Transformer([S core ,S behavior ,S derived ,S risk-context )))
[0103] where \(w\) is the matching weight group, softmax is the normalization function, MLP is the multi-layer perceptron network, Transformer refers to the Transformer attention neural network, \(S core refers to the user core feature, \(S behavior refers to the user behavior feature, \(S derived refers to the user-derived feature, \(S risk-context refers to the preset risk control feature, [S core ,S behavior ,S derived ,S risk-context refers to the vector concatenation operation on the user core feature, the user behavior feature, the user-derived feature, and the risk control feature.
[0104] Specifically, the historical core feature corresponds to the partial feature corresponding to the core data layer in the target historical document feature, the historical behavior feature corresponds to the partial feature corresponding to the behavior feature layer in the target historical document feature, and the historical derived feature corresponds to the partial feature corresponding to the derived data layer in the target historical document feature.
[0105] Specifically, the cosine similarity distance algorithm can be used to calculate the core feature matching degree according to the user core feature and the historical core feature, the dynamic time warping similarity algorithm (Dynamic Time Warping, abbreviated as DTW) can be used to calculate the behavior feature matching degree according to the user behavior feature and the historical behavior feature, and the Jaccard similarity algorithm can be used to calculate the derived feature matching degree according to the user-derived feature and the historical derived feature.
[0106] Specifically, the cosine similarity algorithm is an algorithm used to measure the similarity of the cosine value of the angle between two vectors, the dynamic time warping similarity algorithm is used to calculate the similarity between time series, that is, even if two time series have an offset in the time dimension, their optimal matching path can be found to make the overall similarity optimal, and the Jaccard similarity algorithm is used to measure the similarity between two sets.
[0107] Specifically, when performing weighted summation on the core feature matching degree, the behavior feature matching degree, and the derivative feature matching degree according to the matching weight recombination, it means extracting the matching core weight of the user core feature, the matching behavior weight of the user behavior feature, and the derivative matching weight of the user derivative feature from the matching weight recombination respectively, multiplying the matching core weight by the user core feature, and multiplying the result by the parameter weight value of the user core feature, multiplying the matching behavior weight by the user behavior feature, and multiplying the result by the parameter weight value of the user behavior feature, multiplying the matching derivative weight by the user derivative feature, and multiplying the result by the parameter weight value of the user derivative feature.
[0108] Specifically, the parameter weight value of the core feature matching degree is greater than the parameter weight value of the behavior feature matching degree, and the parameter weight value of the behavior feature matching degree is greater than the parameter weight value of the derivative feature matching degree. By default, the parameter weight value of the core feature matching degree is 0.6, the parameter weight value of the behavior feature matching degree is 0.3, and the parameter weight value of the derivative feature matching degree is 0.1.
[0109] In the embodiment of the present invention, by performing multi-dimensional similarity matching on the user document feature set according to the approval case library, a matching decision result set is obtained, which can perform similarity matching by combining the features of three dimensions: core data, behavior features, and derivative data, thereby improving the accuracy of the approval result matching, and can perform flexible approval result matching according to the historical record data, improving the flexibility of the risk control approval.
[0110] S50. Perform autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set.
[0111] Specifically, each analysis decision result in the analysis decision result set corresponds to the decision result obtained by performing approval analysis on the corresponding user document feature in the user document feature set according to the deep learning model.
[0112] In the embodiment of the present invention, performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set includes:
[0113] Performing approval analysis on the historical document feature set in the approval case library by using a preset autoregressive model to obtain an analysis approval result set;
[0114] Calculating the result loss value between the user approval result set in the approval case library and the analysis approval result set;
[0115] Performing iterative model training on the autoregressive model according to the result loss value to obtain an approval case model;
[0116] Use the approval case model to perform approval analysis on the user document feature set to obtain an analysis decision result set.
[0117] Specifically, the autoregressive model can be an autoregressive neural network model such as a long short-term memory neural network model, Transformer, GPT, etc.
[0118] In detail, the result loss value can be calculated by combining the loss value algorithm with the temporal variance of each user approval result in the user approval result set. Among them, the loss value algorithm can be a mean square error loss value algorithm, an average absolute error algorithm, a Huber loss value algorithm, or a logarithmic cosine loss value algorithm.
[0119] In detail, the gradient descent algorithm, the adaptive learning rate algorithm, or the particle swarm algorithm, the differential evolution algorithm can be used for iterative model training, that is, the gradient descent algorithm, the adaptive learning rate algorithm, or the particle swarm algorithm, the differential evolution algorithm is used to adjust the model parameters of the autoregressive model according to the result loss value, and return the step of using the preset autoregressive model to perform approval analysis on the historical document feature set in the approval case library to obtain an analysis approval result set, until the result loss value is less than the preset loss threshold or the number of iterations is greater than the preset iteration threshold, to obtain the approval case model.
[0120] In the embodiment of the present invention, by performing autoregressive approval analysis on the user document feature set according to the approval case library, an analysis decision result set is obtained, and the risk control approval of the user document feature set can be realized by using the method of deep learning, thereby improving the flexibility of risk control approval.
[0121] S60. Generate a risk control decision result set according to the matching decision result set and the analysis decision result set.
[0122] In the embodiment of the present invention, each risk control decision result in the risk control decision result set corresponds to the approval result of the risk control approval of each user information in the user information set.
[0123] In detail, generating the risk control decision result set according to the matching decision result set and the analysis decision result set means performing weighted summation of the confidence or matching degree according to each matching decision result in the matching decision result set and the corresponding analysis decision result in the analysis decision result set, so as to obtain the final decision result. For example, the matching decision result is approval failure, the weight is 0.4, and the matching degree is 60%, and the analysis decision result is approval success, the weight is 0.6, and the confidence is 90%, then the risk control decision result is approval success. Among them, when the weighted summation result is zero, it can be submitted to manual approval.
[0124] In an embodiment of the present invention, by generating a risk control decision result set according to the matching decision result set and the analysis decision result set, risk control approval can be performed based on feature matching and autoregressive models, expanding the scope of risk control approval, reducing human resource consumption, and improving the flexibility of risk control approval.
[0125] It can be seen that in the above solution, by performing data stratification and document conversion on the pre-acquired user information set, the user data document set can be obtained to complete primary data classification and sorting, and archived according to the importance of the data, realizing formatted data storage, facilitating subsequent data analysis, improving the efficiency of risk control approval. By performing a time series embedding operation on the user data document set to obtain a user document feature set, the user data document set can be converted into a feature form convenient for calculation, reducing the calculation dimension, improving the efficiency of risk control approval, and also facilitating subsequent feature matching. By performing data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library, the archived storage of historical data can be realized, and fast similarity retrieval and similarity calculation in the future can be facilitated, improving the flexibility of risk control approval.
[0126] By performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set, similarity matching can be performed by combining features in three dimensions of core data, behavioral characteristics, and derivative data, thereby improving the accuracy of approval result matching, and flexible approval result matching can be performed according to historical record data, improving the flexibility of risk control approval; by performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set, risk control approval of the user document feature set can be realized by using deep learning methods, thereby improving the flexibility of risk control approval. By generating a risk control decision result set according to the matching decision result set and the analysis decision result set, risk control approval can be performed based on feature matching and autoregressive models, expanding the scope of risk control approval, reducing human resource consumption, and improving the flexibility of risk control approval.
[0127] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
[0128] In one embodiment, a case-based risk control device based on artificial intelligence is provided. The case-based risk control device based on artificial intelligence corresponds one-to-one with the case-based risk control method based on artificial intelligence in the above embodiment. As As shown in the figure, the case-based risk control device based on artificial intelligence includes a document conversion module 101, a feature extraction module 102, a data archiving module 103, a matching decision module 104, an analysis decision module 105, and a risk control decision module 106. The detailed description of each functional module is as follows:
[0129] The document conversion module 101 is used to perform data stratification and document conversion on the pre-acquired user information set to obtain a user data document set;
[0130] The feature extraction module 102 is used to perform time series embedding operation on the user data document set to obtain a user document feature set;
[0131] The data archiving module 103 is used to perform data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library;
[0132] The matching decision module 104 is used to perform multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set;
[0133] The analysis decision module 105 is used to perform autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set;
[0134] The risk control decision module 106 is used to generate a risk control decision result set according to the matching decision result set and the analysis decision result set.
[0135] In an embodiment, when the document conversion module 101 performs data stratification and document conversion on the pre-acquired user information set to obtain a user data document set, it is used for:
[0136] Perform data cleaning on the pre-acquired user information set to obtain a cleaned information set;
[0137] Select the cleaning information in the cleaned information set one by one as the target cleaning information, perform core data matching on the target cleaning information to obtain target core data;
[0138] Extract the behavioral characteristics of the target cleaning information to obtain target behavioral characteristics;
[0139] Extract risk indicators from the target cleaning information to obtain target derivative data;
[0140] Perform document conversion on the target core data, the target behavioral characteristics, and the target derivative data to obtain a user data document;
[0141] Generate a user data document set according to the user data documents of all target cleaning information in the cleaned information set.
[0142] In one embodiment, when the feature extraction module 102 performs a time series embedding operation on the user data document set to obtain a user document feature set, it is used for:
[0143] Select each user data document in the user data document set as a target data document one by one, perform time series splitting on the target data document to obtain a time series document sequence;
[0144] Perform document word segmentation on the time series document sequence to obtain a time series document phrase sequence;
[0145] Perform word vector conversion on the time series document phrase sequence to obtain a time series word vector group sequence;
[0146] Perform position encoding on the time series word vector group sequence to obtain a time series word feature group sequence;
[0147] Perform feature splicing according to the time series word feature group sequence to obtain a time series document feature sequence;
[0148] Perform time series feature convolution on the time series document feature sequence to obtain user document features;
[0149] Aggregate the user document features of all target data documents in the user data document set into a user document feature set.
[0150] In one embodiment, when the data archiving module 103 performs data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library, it is used for:
[0151] Perform user splitting on the pre-acquired historical approval data to obtain a user approval data set;
[0152] Extract a historical user data set and a user approval result set corresponding to the historical user data set from the user approval data set respectively;
[0153] Perform data stratification and document conversion on the historical user data set to obtain a historical user document set;
[0154] Perform time series embedding operation on the historical user document set to obtain a historical document feature set;
[0155] Perform database archiving on the historical document feature set and the user approval result set to obtain an approval case library.
[0156] In one embodiment, when the matching decision module 104 performs multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set, it is used for:
[0157] Select the user document features in the user document feature set one by one as the target user document features, and select the historical document features in the historical document feature set of the approval case library one by one as the target historical document features;
[0158] Perform multi-dimensional similarity calculation on the target user document features and the target historical document features to obtain a multi-dimensional matching degree;
[0159] Aggregate the multi-dimensional matching degrees of all target historical document features in the approval case library into a multi-dimensional matching degree set;
[0160] Perform extreme value screening on the multi-dimensional matching degree set to obtain an extreme value matching degree, and use the target historical document feature corresponding to the extreme value matching degree as the matching historical document feature;
[0161] Use the user approval result corresponding to the matching historical document feature in the approval case library as the matching decision result;
[0162] Aggregate the matching decision results corresponding to all target user document features in the user document feature set into a matching decision result set.
[0163] In one embodiment, when the matching decision module 104 performs multi-dimensional similarity calculation on the target user document features and the target historical document features to obtain a multi-dimensional matching degree, it is used for:
[0164] Perform feature level splitting on the target user document features to obtain user core features, user behavior features, and user derivative features;
[0165] Calculate a matching weight combination according to the user core features, the user behavior features, and the user derivative features;
[0166] Perform feature level splitting on the target historical document features to obtain historical core features, historical behavior features, and historical derivative features;
[0167] Calculate a core feature matching degree according to the user core features and the historical core features;
[0168] Calculate a behavior feature matching degree according to the user behavior features and the historical behavior features;
[0169] Calculate a derivative feature matching degree according to the user derivative features and the historical derivative features;
[0170] Perform weighted summation on the core feature matching degree, the behavior feature matching degree, and the derivative feature matching degree according to the matching weight combination to obtain a multi-dimensional matching degree.
[0171] In one embodiment, when performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis and decision result set, the analysis and decision module 105 is used for:
[0172] Performing approval analysis on the historical document feature set in the approval case library by using a preset autoregressive model to obtain an analysis and approval result set;
[0173] Calculating a result loss value between the user approval result set in the approval case library and the analysis and approval result set;
[0174] Performing iterative model training on the autoregressive model according to the result loss value to obtain an approval case model;
[0175] Performing approval analysis on the user document feature set by using the approval case model to obtain an analysis and decision result set.
[0176] The present invention provides a case-based risk control device based on artificial intelligence. First, by performing data stratification and document conversion on a pre-acquired user information set, a user data document set can be obtained to complete primary data classification and sorting, and be archived according to the importance of the data, realizing formatted data storage, facilitating subsequent data analysis, and improving the efficiency of risk control approval. By performing time series embedding operation on the user data document set to obtain a user document feature set, the user data document set can be converted into a feature form convenient for calculation, reducing the calculation dimension, improving the efficiency of risk control approval, and also facilitating subsequent feature matching. By performing data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library, the historical data can be archived and stored, and subsequent similarity fast retrieval and similarity calculation can be facilitated, improving the flexibility of risk control approval.
[0177] By performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set, the similarity matching can be performed by combining the features of three dimensions: core data, behavior characteristics, and derivative data, thereby improving the accuracy of approval result matching, and the flexible approval result matching can be performed according to the historical record data, improving the flexibility of risk control approval; by performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis and decision result set, the risk control approval of the user document feature set can be realized by using the method of deep learning, thereby improving the flexibility of risk control approval. By generating a risk control decision result set according to the matching decision result set and the analysis and decision result set, the risk control approval can be performed according to feature matching and the autoregressive model, expanding the approval scope of risk control approval, reducing the consumption of human resources, and improving the flexibility of risk control approval.
[0178] For the specific limitations of the case-based risk control device based on artificial intelligence, reference can be made to the limitations of the intelligent Q&A prescribing method in the above text, which will not be elaborated here. Each module in the above case-based risk control device based on artificial intelligence can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to the above modules.
[0179] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as shown. The computer device includes a processor, a memory, a network interface, and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile and / or volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external client through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the server side of a case-based risk control method based on artificial intelligence.
[0180] In one embodiment, a computer device is provided. The computer device can be a client, and its internal structure diagram can be as shown. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external server through a network connection. When the computer program is executed by the processor, it realizes the functions or steps on the client side of a case-based risk control method based on artificial intelligence.
[0181] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented:
[0182] Perform data stratification and document conversion on the pre-acquired user information set to obtain a user data document set;
[0183] Perform a temporal embedding operation on the user data document set to obtain a user document feature set;
[0184] Perform data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library;
[0185] Perform multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set;
[0186] Perform autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set;
[0187] Generate a risk control decision result set according to the matching decision result set and the analysis decision result set.
[0188] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:
[0189] Perform data stratification and document conversion on the pre-acquired user information set to obtain a user data document set;
[0190] Perform time series embedding operation on the user data document set to obtain a user document feature set;
[0191] Perform data standardization and database archiving on the pre-acquired historical approval data to obtain an approval case library;
[0192] Perform multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set;
[0193] Perform autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set;
[0194] Generate a risk control decision result set according to the matching decision result set and the analysis decision result set.
[0195] It should be noted that for the functions or steps that can be implemented by the above computer-readable storage medium or computer device, reference can be made to the relevant descriptions on the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described in detail here.
[0196] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.
[0197] Those skilled in the art can clearly understand that for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0198] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention and should all be included in the protection scope of the present invention. It should be noted that if non-company software tools or components appear in the embodiments of the present application, they are only used for illustrative introduction and do not represent actual use.
Claims
1. A case-based risk control method based on artificial intelligence, characterized in that Including: Performing data stratification and document conversion on a pre-acquired user information set to obtain a user data document set; Performing a time series embedding operation on the user data document set to obtain a user document feature set; Performing data standardization and database archiving on pre-acquired historical approval data to obtain an approval case library; Performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set; Performing autoregressive approval analysis on the user document feature set according to the approval case library to obtain an analysis decision result set; Generating a risk control decision result set according to the matching decision result set and the analysis decision result set.
2. The case-based risk control method based on artificial intelligence according to claim 1, wherein The performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set includes: Selecting the user document features in the user document feature set one by one as target user document features, and selecting the historical document features in the historical document feature set in the approval case library one by one as target historical document features; Performing multi-dimensional similarity calculation on the target user document features and the target historical document features to obtain a multi-dimensional matching degree; Aggregating the multi-dimensional matching degrees of all target historical document features in the approval case library into a multi-dimensional matching degree set; Performing extreme value screening on the multi-dimensional matching degree set to obtain an extreme value matching degree, and using the target historical document feature corresponding to the extreme value matching degree as the matching historical document feature; Using the user approval result corresponding to the matching historical document feature in the approval case library as the matching decision result; Aggregating the matching decision results corresponding to all target user document features in the user document feature set into a matching decision result set.
3. The case-based risk control method based on artificial intelligence according to claim 2, wherein The performing multi-dimensional similarity calculation on the target user document features and the target historical document features to obtain a multi-dimensional matching degree includes: Performing feature level splitting on the target user document features to obtain user core features, user behavior features, and user derivative features; Calculating a matching weight combination according to the user core features, the user behavior features, and the user derivative features; Performing feature level splitting on the target historical document features to obtain historical core features, historical behavior features, and historical derivative features; Calculating a core feature matching degree according to the user core features and the historical core features; Calculating a behavior feature matching degree according to the user behavior features and the historical behavior features; Calculating a derivative feature matching degree according to the user derivative features and the historical derivative features; Performing weighted summation on the core feature matching degree, the behavior feature matching degree, and the derivative feature matching degree according to the matching weight combination to obtain a multi-dimensional matching degree.
4. The case-based risk control method based on artificial intelligence according to claim 1, wherein The performing data stratification and document conversion on a pre-acquired user information set to obtain a user data document set includes: Performing data cleaning on a pre-acquired user information set to obtain a cleaned information set; Selecting the cleaned information in the cleaned information set one by one as target cleaned information, and performing core data matching on the target cleaned information to obtain target core data; Performing behavior feature extraction on the target cleaned information to obtain target behavior features; Extract risk indicators from the target cleaning information to obtain target derivative data; Perform document conversion on the target core data, the target behavior characteristics, and the target derivative data to obtain a user data document; Generate a user data document set according to the user data documents of all target cleaning information in the cleaning information set.
5. The case-based risk control method based on artificial intelligence according to claim 1, wherein The time series embedding operation on the user data document set to obtain a user document feature set, including: Select the user data documents in the user data document set one by one as target data documents, and perform time series splitting on the target data documents to obtain a time series document sequence; Perform document word segmentation on the time series document sequence to obtain a time series document phrase sequence; Perform word vector conversion on the time series document phrase sequence to obtain a time series word vector group sequence; Perform position encoding on the time series word vector group sequence to obtain a time series word feature group sequence; Perform feature splicing according to the time series word feature group sequence to obtain a time series document feature sequence; Perform time series feature convolution on the time series document feature sequence to obtain user document features; Aggregate the user document features of all target data documents in the user data document set into a user document feature set.
6. The case-based risk control method based on artificial intelligence according to claim 1, characterized in that, The data standardization and database archiving of the pre-obtained historical approval data to obtain an approval case library, including: Perform user splitting on the pre-obtained historical approval data to obtain a user approval data set; Extract the historical user data set and the user approval result set corresponding to the historical user data set from the user approval data set respectively; Perform data stratification and document conversion on the historical user data set to obtain a historical user document set; Perform time series embedding operation on the historical user document set to obtain a historical document feature set; Perform database archiving on the historical document feature set and the user approval result set to obtain an approval case library.
7. The case-based risk control method based on artificial intelligence according to claim 1, characterized in that The autoregressive approval analysis of the user document feature set according to the approval case library to obtain an analysis decision result set, including: Use a preset autoregressive model to perform approval analysis on the historical document feature set in the approval case library to obtain an analysis approval result set; Calculate the result loss value between the user approval result set in the approval case library and the analysis approval result set; Perform iterative model training on the autoregressive model according to the result loss value to obtain an approval case model; Use the approval case model to perform approval analysis on the user document feature set to obtain an analysis decision result set.
8. An artificial intelligence-based case-based risk control device, characterized in that, Including: A document conversion module for performing data stratification and document conversion on the pre-obtained user information set to obtain a user data document set; A feature extraction module for performing time series embedding operation on the user data document set to obtain a user document feature set; A data archiving module for performing data standardization and database archiving on the pre-obtained historical approval data to obtain an approval case library; A matching decision module for performing multi-dimensional similarity matching on the user document feature set according to the approval case library to obtain a matching decision result set; An analysis and decision-making module, configured to perform autoregressive approval analysis on the user document feature set according to the approval case library, so as to obtain an analysis and decision-making result set; A risk control decision-making module, configured to generate a risk control decision-making result set according to the matching decision-making result set and the analysis and decision-making result set.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the artificial intelligence-based case-style risk control method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the artificial intelligence-based case-style risk control method according to any one of claims 1 to 7 is implemented.