CN118569983B - Asset data processing system and method for enterprise credit risk management - Google Patents
Asset data processing system and method for enterprise credit risk management Download PDFInfo
- Publication number
- CN118569983B CN118569983B CN202411035356.9A CN202411035356A CN118569983B CN 118569983 B CN118569983 B CN 118569983B CN 202411035356 A CN202411035356 A CN 202411035356A CN 118569983 B CN118569983 B CN 118569983B
- Authority
- CN
- China
- Prior art keywords
- asset data
- credit
- enterprise asset
- semantic
- good
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The application relates to the field of intelligent data analysis, and provides an asset data processing system and method for enterprise credit risk management, which are used for extracting a set of enterprise asset data marked as good credit from a database and acquiring enterprise asset data to be evaluated, and carrying out semantic coding and optimization on the good credit enterprise asset data and the good credit enterprise asset data by adopting a data processing and analyzing technology based on deep learning, so that whether the credit of the good credit enterprise asset data is a good processing result or not is automatically obtained according to the semantic interactive matching characteristic between the good credit enterprise asset data and the good credit enterprise asset data, the automatic processing and analysis of the good credit of the enterprise asset data to be evaluated are realized, the manual intervention and subjective judgment are reduced, and the potential association and characteristic between the data can be captured, thereby improving the evaluation accuracy of the credit status of the enterprise.
Description
Technical Field
The present application relates to the field of intelligent data analysis, and more particularly, to an asset data processing system and method for enterprise credit risk management.
Background
In a strong market competition, businesses need to maintain their own market status and financial health through effective credit risk management. Enterprise credit risk management refers to the credit risk faced by an enterprise in an business activity, i.e., the potential risk that other enterprises or individuals fail to fulfill contractual obligations on time or to repay liabilities. Thus, credit assessment can help identify and quantify the credit risk faced by an enterprise, enabling the enterprise to take precautionary measures, reducing potential losses.
However, conventional evaluation methods may be too dependent on personal experience and judgment of the evaluation personnel, and cannot effectively recognize and utilize complex relationships and potential semantic associations between data, resulting in poor accuracy of the evaluation results.
Accordingly, there is a need for an asset data processing scheme for enterprise credit risk management that addresses the above-described technical problems.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides an asset data processing system and method for enterprise credit risk management, which are used for extracting a set of enterprise asset data marked as good credit from a database and acquiring enterprise asset data to be evaluated, and carrying out semantic coding and optimization on the good credit enterprise asset data and the good credit enterprise asset data by adopting a data processing and analyzing technology based on deep learning, so that whether the credit of the good credit enterprise asset data is a good processing result or not is automatically obtained according to the semantic interactive matching characteristic between the good credit enterprise asset data and the good credit enterprise asset data, the automatic processing and analysis of the good credit enterprise asset data are realized, the manual intervention and subjective judgment are reduced, and the potential association and characteristic between the data can be captured, thereby improving the evaluation accuracy of the credit status of the enterprise.
According to one aspect of the present application there is provided an asset data processing system for enterprise credit risk management, comprising:
a good enterprise asset data extraction module for extracting from the database a collection of enterprise asset data labeled as good credit;
the credit-good enterprise asset data semantic coding module is used for semantically coding each enterprise asset data marked as credit-good in the set of enterprise asset data marked as credit-good to obtain a set of credit-good enterprise asset data semantic coding feature vectors;
The significance global optimization module is used for inputting the set of the semantic coding feature vectors of the credit-good enterprise asset data into the context semantic-based significance global optimization module to obtain the set of the semantic coding feature vectors of the credit-good enterprise asset data;
The enterprise asset data acquisition module to be evaluated is used for acquiring enterprise asset data to be evaluated;
The enterprise asset data semantic coding module to be evaluated is used for performing semantic coding on the enterprise asset data to be evaluated to obtain semantic coding feature vectors of the enterprise asset data to be evaluated;
The unidirectional interaction matching module is used for taking the semantic coding feature vector of the enterprise asset data to be evaluated as a query feature vector, inputting the query feature vector and the set of semantic coding feature vectors of the enterprise asset data with good optimization credit into the unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as unidirectional matching result representation features;
And the processing result generation module is used for obtaining a processing result based on the unidirectional matching result representation characteristic, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not.
According to another aspect of the present application, there is provided an asset data processing method for enterprise credit risk management, comprising:
Extracting from the database a collection of enterprise asset data labeled as good credit;
Performing semantic coding on each enterprise asset data marked as good credit in the set of enterprise asset data marked as good credit to obtain a set of semantic coding feature vectors of the enterprise asset data marked as good credit;
Inputting the set of the semantic coding feature vectors of the credit-good enterprise asset data into a context semantic-based significance global optimization module to obtain a set of the semantic coding feature vectors of the optimization credit-good enterprise asset data;
Acquiring enterprise asset data to be evaluated;
performing semantic coding on the enterprise asset data to be evaluated to obtain semantic coding feature vectors of the enterprise asset data to be evaluated;
Taking the semantic coding feature vector of the enterprise asset data to be evaluated as a query feature vector, inputting the query feature vector and the set of semantic coding feature vectors of the enterprise asset data with good optimization credit into a unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as a unidirectional matching result representation feature;
and obtaining a processing result based on the unidirectional matching result representation feature, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not.
The application has at least the following technical effects: compared with the prior art, the asset data processing system and method for enterprise credit risk management provided by the application have the advantages that the collection of enterprise asset data marked as good credit is extracted from the database, the enterprise asset data to be evaluated is obtained, the deep learning-based data processing and analysis technology is adopted to carry out semantic coding and optimization on the enterprise asset data with good credit and the enterprise asset data to be evaluated, so that whether the credit of the enterprise asset data to be evaluated is a good processing result is automatically obtained according to the semantic interaction matching characteristic between the enterprise asset data to be evaluated and the enterprise asset data with good credit, the automatic processing and analysis of the enterprise asset data to be evaluated are realized, the manual intervention and subjective judgment are reduced, and the potential association and characteristic between the data can be captured, so that the evaluation accuracy of the enterprise credit condition is improved.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing embodiments of the present application in more detail with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of embodiments of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, do not limit the application. In the drawings, like reference numerals generally refer to like parts or steps.
FIG. 1 is a system block diagram of an asset data processing system for enterprise credit risk management in accordance with an embodiment of the application.
FIG. 2 is a schematic architecture diagram of an asset data processing system for enterprise credit risk management according to an embodiment of the application.
FIG. 3 is a block diagram of a saliency global optimization module in an asset data processing system for enterprise credit risk management, according to an embodiment of the application.
FIG. 4 is a block diagram of a one-way interaction matching module in an asset data processing system for enterprise credit risk management in accordance with an embodiment of the application.
FIG. 5 is a flow chart of an asset data processing method for enterprise credit risk management according to an embodiment of the application.
Detailed Description
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
In a competitive market environment, enterprises need to perform effective credit risk management to protect their market status and financial robustness. Enterprise credit risk management involves dealing with the credit risk faced by an enterprise in business operations, i.e., the potential risk that other companies or individuals may fail to fulfill contractual obligations or repay liabilities on time. Thus, by performing credit assessment, an enterprise can identify and measure the credit risk it is faced with, thereby taking precautions to reduce potential losses.
However, conventional evaluation methods may rely too much on subjective experience and judgment of the evaluator, and at the same time cannot effectively identify and utilize complex relationships and potential semantic links between data, resulting in poor accuracy of the evaluation result.
Therefore, in order to solve the technical problems, the technical concept of the application is to extract the set of enterprise asset data marked as good credit from a database and acquire the enterprise asset data to be evaluated, and to perform semantic coding and optimization on the good credit enterprise asset data and the enterprise asset data to be evaluated by adopting a data processing and analyzing technology based on deep learning, so as to automatically obtain whether the credit of the enterprise asset data to be evaluated is a good processing result according to the semantic interactive matching characteristics between the good credit enterprise asset data and the good credit enterprise asset data, thereby realizing the automatic processing and analysis of the enterprise asset data to be evaluated, reducing manual intervention and subjective judgment, and simultaneously capturing potential association and characteristics between the data, thereby improving the evaluation accuracy of the credit status of the enterprise.
FIG. 1 is a system block diagram of an asset data processing system for enterprise credit risk management in accordance with an embodiment of the application. FIG. 2 is a schematic architecture diagram of an asset data processing system for enterprise credit risk management according to an embodiment of the application. As shown in fig. 1 and 2, in an asset data processing system 100 for enterprise credit risk management, there is included: a good enterprise asset data extraction module 110 for extracting from the database a collection of enterprise asset data labeled as good credit; a credit-good enterprise asset data semantic coding module 120, configured to semantically code each of the set of credit-good enterprise asset data labeled as credit-good enterprise asset data to obtain a set of credit-good enterprise asset data semantic coding feature vectors; a saliency global optimization module 130, configured to input the set of semantic encoding feature vectors of the credit-good enterprise asset data into a context semantic-based saliency global optimization module to obtain a set of semantic encoding feature vectors of the optimization credit-good enterprise asset data; an enterprise asset data to be evaluated acquisition module 140 for acquiring enterprise asset data to be evaluated; the enterprise asset data semantic coding module to be evaluated 150 is configured to perform semantic coding on the enterprise asset data to be evaluated to obtain a semantic coding feature vector of the enterprise asset data to be evaluated; the unidirectional interaction matching module 160 is configured to take the semantic coding feature vector of the enterprise asset data to be evaluated as a query feature vector, input the query feature vector and the set of semantic coding feature vectors of the enterprise asset data with good optimization credit into a unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as a unidirectional matching result representation feature; and the processing result generating module 170 is configured to obtain a processing result based on the unidirectional matching result representing feature, where the processing result is used to represent whether the credit of the enterprise asset data to be evaluated is good.
In an embodiment of the present application, the good enterprise asset data extraction module 110 is configured to extract a collection of enterprise asset data labeled as good credit from a database. Specifically, enterprise asset data sets labeled as well-credited are positive examples of model learning. By analyzing this data, features and patterns associated with the good credit of the enterprise asset data, including financial indicators, transaction history, payment behavior, etc., can be extracted, which are key factors in assessing the risk of the enterprise asset credit.
In the embodiment of the present application, the semantic encoding module 120 is configured to semantically encode each of the set of good-credit enterprise asset data marked as good-credit enterprise asset data to obtain a set of semantic encoding feature vectors of the good-credit enterprise asset data. Accordingly, consider that each of the set of business asset data labeled as good-credit contains semantic information about different types of business credit asset data, and that there is a pattern of association and information between contexts between each business credit asset data. Based on this, in the technical solution of the present application, semantic encoding is performed on each enterprise asset data marked as good credit in the set of enterprise asset data marked as good credit to capture and other semantic relationships between contexts included in each enterprise asset data marked as good credit, so as to obtain a set of semantic encoding feature vectors of the good credit enterprise asset data. In particular, in one embodiment of the present application, after word segmentation processing is performed on each of the set of enterprise asset data marked as good credit, the set of semantic encoding feature vectors of the enterprise asset data marked as good credit is obtained by inputting the context semantic encoder based on the converter, so that each of the set of enterprise asset data marked as good credit can be divided into a plurality of words, and semantic information of the enterprise asset data marked as good credit is included in each word, so as to better understand the semantic information between each word in each of the enterprise asset data marked as good credit.
In the embodiment of the present application, the saliency global optimization module 130 is configured to input the set of semantic encoding feature vectors of the credit-good enterprise asset data into a context semantic-based saliency global optimization module to obtain a set of semantic encoding feature vectors of the optimization credit-good enterprise asset data. Accordingly, consider that the set of semantically encoded feature vectors of the credit-good enterprise asset data presents significant contextual semantic information and global semantic similarity sums based on the entire feature set between different types of enterprise asset data. Based on the above, in order to better comprehensively utilize context information among semantic coding features of different credit good enterprise asset data, including semantic association, semantic global similarity, significance difference and the like among the credit good enterprise asset data, semantic optimization is performed on a set of semantic coding feature vectors of the credit good enterprise asset data.
In detail, the context semantic-based saliency global optimization module extracts the most salient semantic feature information and global information based on the whole enterprise asset data from the set of semantic encoding feature vectors of the credit good enterprise asset data, so as to better capture the most representative distinguishing feature in the semantic encoding features of the credit good enterprise asset data, and comprehensively understand the feature distribution and structure of the whole credit good enterprise asset data set, thereby distinguishing the feature differences among different types of enterprise asset data more accurately and improving the comprehensive performance of credit evaluation. The resulting salient and global features are then fused and feature processed to enrich the characterizability of the credit-good enterprise asset data. And then, the feature vectors obtained after feature processing are used as weights to weight the set of the semantic coding feature vectors of the credit-good enterprise asset data, so that semantic association and features among the credit-good enterprise asset data can be better reflected, the precision and the efficiency of credit evaluation are improved, and more reliable risk management suggestions are provided for decision makers.
In particular, FIG. 3 is a block diagram of a saliency global optimization module in an asset data processing system for enterprise credit risk management according to an embodiment of the application. As shown in fig. 3, the saliency global optimization module 130 includes: a salient feature calculating unit 131, configured to extract a maximum value of each semantic encoding feature vector of the semantic encoding feature vectors of the credit-good enterprise asset data to obtain a semantic encoding feature vector of the credit-good enterprise asset data; a global feature calculation unit 132, configured to extract an average value of all the semantic encoding feature vectors of the credit-good enterprise asset data in the set of semantic encoding feature vectors of the credit-good enterprise asset data to obtain a semantic encoding feature vector of the global credit-good enterprise asset data; a convolution activating unit 133, configured to perform convolution encoding and feature activation processing on the feature vector for semantic encoding of the outstanding credit good enterprise asset data and the feature vector for semantic encoding of the global credit good enterprise asset data to obtain an activation vector for semantic encoding of the outstanding feature credit good enterprise asset data and an activation vector for semantic encoding of the global feature credit good enterprise asset data; a salient global fusion unit 134, configured to fuse the salient feature credit good enterprise asset data semantic coding activation vector and the global feature credit good enterprise asset data semantic coding activation vector to obtain a global-salient feature credit good enterprise asset data semantic coding feature vector; a nonlinear activating unit 135, configured to perform nonlinear activation on the global-salient feature credit good enterprise asset data semantic coding feature vector to obtain a global-salient feature credit good enterprise asset data semantic coding weight feature vector; and the feature weighted fusion unit 136 is configured to take the global-salient feature credit good enterprise asset data semantic coding weight feature vector as a weight, multiply the set of credit good enterprise asset data semantic coding feature vectors by location points and add the set of credit good enterprise asset data semantic coding feature vectors to obtain the set of optimized credit good enterprise asset data semantic coding feature vectors.
More specifically, in an embodiment of the present application, the convolution activation unit is configured to: inputting the semantic convolution encoding vector of the good enterprise asset data with outstanding credit, which is obtained after the semantic encoding feature vector of the good enterprise asset data with outstanding credit is subjected to one-dimensional convolution encoding, into a ReLU function for feature activation processing so as to obtain the semantic activation encoding feature vector of the good enterprise asset data with outstanding credit; performing matrix multiplication on the salient point convolution feature vector obtained after the salient point convolution encoding feature vector is subjected to point convolution encoding on the salient feature credit good enterprise asset data semantic activation encoding feature vector and a first weight matrix to obtain the salient feature credit good enterprise asset data semantic encoding activation vector; inputting the global credit good enterprise asset data semantic convolution coding vector obtained after the global credit good enterprise asset data semantic coding feature vector is subjected to one-dimensional convolution coding into a ReLU function to perform feature activation processing so as to obtain the global credit good enterprise asset data semantic activation coding feature vector; and performing matrix multiplication on the global point convolution characteristic vector obtained after the point convolution encoding of the global credit good enterprise asset data semantic activation encoding characteristic vector and a second weight matrix to obtain the global feature credit good enterprise asset data semantic encoding activation vector.
More specifically, in an embodiment of the present application, the nonlinear activation unit is configured to: inputting the global-salient feature credit good enterprise asset data semantic coding feature vector into a tanh function to obtain a first activated global-salient feature credit good enterprise asset data semantic coding feature vector; inputting the global-salient feature credit good enterprise asset data semantic coding feature vector into a Sigmoid function to obtain a second activated global-salient feature credit good enterprise asset data semantic coding feature vector; and multiplying the first activated global-salient feature credit good enterprise asset data semantic coding feature vector and the second activated global-salient feature credit good enterprise asset data semantic coding feature vector by location points to obtain the global-salient feature credit good enterprise asset data semantic coding weight feature vector.
In an embodiment of the present application, specifically, the significance global optimization module is configured to: processing the set of credit-good enterprise asset data semantic coding feature vectors with the following semantic optimization formula by using the context semantic-based significance global optimization module to obtain the set of optimization credit-good enterprise asset data semantic coding feature vectors; wherein, the semantic optimization formula is:
Wherein, A set of feature vectors is semantically encoded for the credit-good enterprise asset data,AndThe maximum value of each feature vector in the set of feature vectors and the average value of each feature vector,In the case of one-dimensional convolutional encoding,Is thatThe function of the function is that,In the case of a point convolution code,AndThe first weight matrix and the second weight matrix,AndThe salient feature credit good enterprise asset data semantic code activation vector and the global feature credit good enterprise asset data semantic code activation vector,Is saidThe function of the function is that,Is thatThe function of the function is that,Representing the multiplication by the position point,Is a collection of semantically encoded feature vectors of the optimization credit good enterprise asset data.
In an embodiment of the present application, the to-be-evaluated enterprise asset data obtaining module 140 is configured to obtain to-be-evaluated enterprise asset data. Specifically, the enterprise asset data to be evaluated may include information such as financial statements, transaction records, credit records, and the like. By processing this information, the characteristics of the enterprise asset to be evaluated, including key asset information, can be obtained. In the technical scheme of the application, the characteristics of the enterprise asset to be evaluated are required to be compared and matched with the characteristics of the enterprise asset data marked as good in credit, so as to judge whether the credit of the enterprise to be evaluated is good.
In the embodiment of the present application, the semantic encoding module 150 is configured to perform semantic encoding on the enterprise asset data to be evaluated to obtain a semantic encoding feature vector of the enterprise asset data to be evaluated. Accordingly, considering that the enterprise asset data to be evaluated contains key semantic information and characteristics about the enterprise asset data to be evaluated, the key semantic information has important roles and influences on credit evaluation of subsequent enterprises to be evaluated. Therefore, in the technical scheme of the application, the enterprise asset data to be evaluated is subjected to semantic coding so as to capture and extract important semantic information in the enterprise asset data to be evaluated, thereby obtaining semantic coding feature vectors of the enterprise asset data to be evaluated. In a specific embodiment of the present application, the semantic encoding process is performed on the enterprise asset data to be evaluated by using a context semantic encoder based on a converter to obtain semantic encoding feature vectors of the enterprise asset data to be evaluated.
In the embodiment of the present application, the unidirectional interaction matching module 160 is configured to take the semantic encoding feature vector of the enterprise asset data to be evaluated as a query feature vector, and input the query feature vector and the set of semantic encoding feature vectors of the enterprise asset data with good optimization credit into a unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as a unidirectional matching result representation feature. Specifically, in order to perform more accurate similarity calculation and correlation degree matching on the enterprise asset data semantic coding feature vectors to be evaluated based on the set of the optimization credit good enterprise asset data semantic coding feature vectors as a reference, in the technical scheme of the application, the enterprise asset data semantic coding feature vectors to be evaluated are used as query feature vectors, and the query feature vectors and the set of the optimization credit good enterprise asset data semantic coding feature vectors are input into a unidirectional attention granularity-by-granularity scanning interaction matching module to obtain unidirectional matching result representation vectors. That is, the unidirectional attention granularity-by-granularity scanning interaction matching module can help capture semantic interaction information between the query feature vector and the optimization credit good enterprise asset data feature vector set, so as to improve accuracy and effect of matching between data. In particular, the one-way matching semantic relevance and similarity between the query feature and the optimization credit good enterprise asset data is first determined by calculating a semantic metric coefficient between the query feature vector and the semantic encoding feature vector of each of the optimization credit good enterprise asset data. And then, taking the value obtained after normalization processing of each semantic measurement coefficient as a weight value to weight and fuse semantic coding feature vectors of the enterprise asset data with good optimization credit to highlight semantic coding features which are most important and similar to the query feature vectors. And finally, carrying out differential calculation on the unidirectional matching interaction remodelling characteristics obtained after weighted fusion and the query characteristics so as to more accurately understand the matching degree between the data to be evaluated and the credit good data, thereby realizing more accurate matching and evaluation processes.
In particular, FIG. 4 is a block diagram of a one-way interaction matching module in an asset data processing system for enterprise credit risk management in accordance with an embodiment of the application. As shown in fig. 4, the unidirectional interaction matching module 160 includes: a semantic similarity calculating unit 161, configured to calculate semantic similarity between the query feature vector and each semantic encoding feature vector of the optimization credit good enterprise asset data in the set of semantic encoding feature vectors of the optimization credit good enterprise asset data to obtain a sequence of one-way matching semantic metric coefficients of the enterprise asset data; a normalization processing unit 162, configured to normalize the sequence of the one-way matching semantic metric coefficients of the enterprise asset data to obtain a sequence of one-way matching semantic metric weight values of the enterprise asset data; a feature weighting unit 163, configured to calculate a weighted sum of the set of semantic encoding feature vectors of the optimization credit good enterprise asset data by using the sequence of the semantic metric weight values of the one-way matching of the enterprise asset data as a weight to obtain a one-way matching interaction remodelling feature vector of the enterprise asset data; a differential feature calculating unit 164, configured to calculate, as the unidirectional matching result representation vector, a differential feature vector between the unidirectional matching interaction remodelling feature vector of the enterprise asset data and the query feature vector.
More specifically, in an embodiment of the present application, the semantic similarity calculation unit is configured to: respectively cascading each optimization credit good enterprise asset data semantic coding feature vector in the query feature vector and the set of optimization credit good enterprise asset data semantic coding feature vectors to obtain a set of query enterprise asset data semantic coding joint feature vectors; after the matrix product of each query enterprise asset data semantic coding joint feature vector and the weight matrix in the set of query enterprise asset data semantic coding joint feature vectors is calculated respectively, the obtained feature vectors and the bias vectors are added according to positions to obtain a sequence of the query enterprise asset data semantic coding joint bias feature vectors; inputting the sequence of the query enterprise asset data semantic coding joint bias feature vector into a sigmoid activation function to obtain the sequence of the enterprise asset data unidirectional matching semantic measurement coefficients.
More specifically, in an embodiment of the present application, the normalization processing unit is configured to: taking each enterprise asset data one-way matching semantic measurement coefficient in the sequence of enterprise asset data one-way matching semantic measurement coefficients as an index of a natural constant to calculate an index function value based on the natural constant so as to obtain a sequence of enterprise asset data one-way matching semantic measurement coefficient logarithmic values; calculating the sum of the logarithmic values of the unidirectional matching semantic measurement coefficients of the enterprise asset data in the sequence of the logarithmic values of the unidirectional matching semantic measurement coefficients of the enterprise asset data to obtain the total sum of the unidirectional matching semantic measurement coefficients of the enterprise asset data; dividing each enterprise asset data one-way matching semantic measurement coefficient pair value in the enterprise asset data one-way matching semantic measurement coefficient pair value sequence with the enterprise asset data one-way matching semantic measurement total sum value to obtain the enterprise asset data one-way matching semantic measurement weight value sequence.
More specifically, in an embodiment of the present application, the feature weighting unit is configured to: multiplying each enterprise asset data one-way matching semantic measurement weight value in the sequence of enterprise asset data one-way matching semantic measurement weight values by a corresponding optimization credit good enterprise asset data semantic coding feature vector in the set of optimization credit good enterprise asset data semantic coding feature vectors according to positions to obtain a set of optimization credit good enterprise asset data semantic coding weight feature vectors; and calculating the position-wise summation of the set of the semantic coding weight feature vectors of the optimization credit good enterprise asset data to obtain the one-way matching interaction remodelling feature vector of the enterprise asset data.
In the embodiment of the application, specifically, the unidirectional interaction matching module is used for: inputting the query feature vector and the set of the semantic coding feature vectors of the enterprise asset data with good optimization credit into a unidirectional attention granularity-by-granularity scanning interactive matching module, and processing the query feature vector and the set of the semantic coding feature vectors of the enterprise asset data with good optimization credit by using the following interactive matching formula to obtain a unidirectional matching result expression vector; wherein, the interactive matching formula is:
Wherein, AndThe first set of semantically encoded feature vectors for the query feature vector and the optimization credit good enterprise asset data, respectivelyThe optimization credit good enterprise asset data semantically encodes feature vectors,Is a cascade of processes which are carried out,Is a matrix of the weights of the said weights,Is the offset vector of the said one,Is a sigmoid function of the number of bits,In sequence of unidirectionally matching semantic metric coefficients for the enterprise asset dataIndividual enterprise asset data matches semantic metric coefficients unidirectionally,Expressed in natural constantAs a function of the base of the exponentiation,Is the number of feature vectors in the set of optimization credit good enterprise asset data semantically encoded feature vectors,Is the first in the sequence of the one-way matching semantic metric weight values of the enterprise asset dataIndividual enterprise asset data one-way matches semantic metric weight values,Is the enterprise asset data one-way matching interactive remodelling feature vector,Indicating that the subtraction is performed by position,Is the one-way match result representation vector.
In this embodiment of the present application, the processing result generating module 170 is configured to obtain a processing result based on the unidirectional matching result indicating a feature, where the processing result is used to indicate whether the credit of the enterprise asset data to be evaluated is good. Specifically, in the embodiment of the present application, the processing result generating module is configured to: and inputting the unidirectional matching result representation vector into an asset data processing result generator based on the classifier to obtain a processing result, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not. That is, the unidirectional matching result representing characteristics obtained by single interactive matching between the semantic coding feature vector of the enterprise asset data to be evaluated and the set of semantic coding feature vectors of the enterprise asset data with good credit are utilized to carry out classification processing, so as to automatically obtain whether the credit of the enterprise asset data to be evaluated is a good processing result. By the method, automatic processing and analysis of enterprise asset data to be evaluated are realized, manual intervention and subjective judgment are reduced, and potential association and characteristics between the data can be captured, so that evaluation accuracy of enterprise credit conditions is improved.
Specifically, the set of the semantic encoding feature vector of the enterprise asset data to be evaluated and the set of the semantic encoding feature vector of the enterprise asset data to be optimized express the encoding text semantic feature of each enterprise asset data marked as credit good in the set of the enterprise asset data to be evaluated and the set of the enterprise asset data marked as credit good respectively, but the set of the semantic encoding feature vector of the enterprise asset data to be optimized carries out global context significance enhancement on the basis of the encoding semantic feature representation of source semantics, so that after the set of the semantic encoding feature vector of the enterprise asset data to be evaluated and the set of the semantic encoding feature vector of the enterprise asset data to be optimized are input into a one-way attention-level-by-granularity scanning interactive matching network, the obtained one-way matching result representation vector also has prior-posterior probability causal correlation relative to source text semantics due to the difference of the one-way semantic attention-interactive matching weight under the different feature distribution representations of the semantic encoding feature vector of the enterprise asset data to be evaluated and the semantic encoding feature vector of the enterprise asset data to be optimized, and the accuracy of classification result is affected.
Based on this, in a preferred embodiment, inputting the one-way matching result representation vector into a classifier-based asset data processing result generator to obtain a processing result includes: carrying out probability-based activation functions, such as sigmoid functions and softmax functions, on the characteristic values of the unidirectional matching result representation vectors to obtain a probabilistic unidirectional matching result representation vector; obtaining the unidirectional matching result representation vector, inputting the good probability value representing that the credit of the enterprise asset data to be evaluated is good, which is obtained by an asset data processing result generator based on a classifier; determining class-aware symbol values based on a comparison of each eigenvalue of the probabilistic unidirectional matching result representation vector with the good probability value, wherein the class-aware symbol values are equal to one, zero and negative one in response to the probabilistic unidirectional matching result representation vector eigenvalue being greater than, equal to and less than the good probability value, respectively; calculating the average value of all the characteristic values of the probabilistic unidirectional matching result representation vector to obtain an integral phase shift value; multiplying each characteristic value of the probabilistic unidirectional matching result representing vector by the class cognitive symbol value and the class integral phase shift value respectively, then carrying out weighted difference calculation, and taking an absolute value to obtain an optimized characteristic value of the probabilistic unidirectional matching result representing vector; inputting the optimized unidirectional matching result representation vector composed of the optimized eigenvalues into the classifier-based asset data processing result generator to obtain the processing result.
Specifically, in the preferred embodiment, the probabilistic unidirectional matching result representation vector is optimized to obtain an optimized unidirectional matching result representation vector, and the process formula is as follows:
Wherein, Is the feature value of each position in the probabilistic unidirectional matching result representation vector,Is the value of said good probability that,Is a function of the sign of the symbol,AndIs the weight of the parameter to be exceeded,Is the phase shift value of the class as a whole,Is the eigenvalue of each position in the vector represented by the optimized unidirectional matching result.
Therefore, in the optimization process, the class cognitive phase transformation response of the unidirectional matching result representing vector is obtained by comparing the probability amplitude of the characteristic value of the unidirectional matching result representing vector with the class probability, the characteristic distribution sequence invariance transformation based on the class differential distribution expansion is carried out on the phase shift response of the characteristic value of the unidirectional matching result representing vector relative to the class probability representation of the whole characteristic vector, so that the causal constraint of the posterior class probability of the unidirectional matching result representing vector on the prior characteristic distribution representation of the posterior class probability is realized, the accuracy of the unidirectional matching result representing vector input to the processing result obtained by the asset data processing result generator based on the classifier is improved, the automatic processing and analysis of the asset data of an enterprise to be evaluated are realized, the manual intervention and the subjective judgment are reduced, and the potential association and characteristics between the data can be captured, thereby the evaluation accuracy of the credit condition of the enterprise is improved.
In summary, the asset data processing system 100 for enterprise credit risk management according to the embodiments of the present application is illustrated by extracting a set of enterprise asset data marked as good credit from a database and acquiring enterprise asset data to be evaluated, and semantically encoding and optimizing the good credit enterprise asset data and the good credit enterprise asset data by using a data processing and analysis technology based on deep learning, so as to automatically obtain whether the credit of the good credit processing result of the good credit enterprise asset data is a good processing result according to the semantically interactive matching feature between the good credit enterprise asset data and the good credit enterprise asset data. By the method, automatic processing and analysis of enterprise asset data to be evaluated are realized, manual intervention and subjective judgment are reduced, and potential association and characteristics between the data can be captured, so that evaluation accuracy of enterprise credit conditions is improved.
As described above, the asset data processing system 100 for enterprise credit risk management according to an embodiment of the present application may be implemented in various terminal devices, such as a server or the like for asset data processing for enterprise credit risk management. In one example, asset data processing system 100 for enterprise credit risk management according to embodiments of the application may be integrated into a terminal device as a software module and/or hardware module. For example, the asset data processing system 100 for enterprise credit risk management may be a software module in the operating system of the terminal device or may be an application developed for the terminal device; of course, the asset data processing system 100 for enterprise credit risk management may likewise be one of the numerous hardware modules of the terminal device.
Alternatively, in another example, the asset data processing system 100 for enterprise credit risk management and the terminal device may be separate devices, and the asset data processing system 100 for enterprise credit risk management may be connected to the terminal device through a wired and/or wireless network and transmit the interaction information in a agreed data format.
In other embodiments of the present application, an asset data processing system for enterprise credit risk management is provided, which includes a data management module for implementing functions of importing, maintaining, data complement, data quality analysis, total score checksum and table management of bank data; a risk assessment module for covering assessment of credit risk, market risk, operational risk and liquidity risk; a risk control module for risk weighted asset metering, stress testing, capital management, risk early warning, and risk regulation; the risk slow-release management module is used for identifying, distributing and calculating the risk slow-release tools and ensuring that risk slow-release measures can optimally reduce risk exposure; an operational risk management module for capital metering and management for operational risks, including collection of loss data, calculation of internal loss multipliers, and operational risk capital requirement determination; a credit risk management module for assessment of credit risk, classification and metering of risk exposure, and calculation of RWA; a risk pricing and performance management module for assessing the value of the risk asset through risk pricing and monitoring and adjusting the relationship between risk bearing and revenue through performance management; report management module for showing and downloading all supervisory report and internal management report, such as structure and trend analysis of sufficient rate of capital, distribution of RWA, analysis of occupied capital, etc.; and the system management module is used for user management, role management, authority management, log management and the like, and ensures the security and compliance of the system.
FIG. 5 is a flow chart of an asset data processing method for enterprise credit risk management according to an embodiment of the application. As shown in fig. 5, in an asset data processing method for enterprise credit risk management, it includes: s110, extracting a set of enterprise asset data marked as good credit from a database; s120, performing semantic coding on each enterprise asset data marked as good credit in the set of enterprise asset data marked as good credit to obtain a set of semantic coding feature vectors of the enterprise asset data marked as good credit; s130, inputting the set of semantic coding feature vectors of the credit-good enterprise asset data into a context semantic-based significance global optimization module to obtain a set of semantic coding feature vectors of the optimization credit-good enterprise asset data; s140, acquiring enterprise asset data to be evaluated; s150, carrying out semantic coding on the enterprise asset data to be evaluated to obtain semantic coding feature vectors of the enterprise asset data to be evaluated; s160, taking the semantic coding feature vector of the enterprise asset data to be evaluated as a query feature vector, and inputting the query feature vector and the set of semantic coding feature vectors of the enterprise asset data with good optimization credit into a unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as a unidirectional matching result representation feature; and S170, based on the unidirectional matching result representing characteristics, obtaining a processing result, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not.
The specific operation of the respective steps in the above-described asset data processing method for enterprise credit risk management has been described in detail in the above description of the asset data processing system for enterprise credit risk management with reference to fig. 1 to 4, and thus, repetitive descriptions thereof will be omitted.
In summary, the asset data processing method for enterprise credit risk management according to the embodiment of the application is explained, which extracts the set of enterprise asset data marked as good credit from the database and acquires the enterprise asset data to be evaluated, and performs semantic coding and optimization on the good credit enterprise asset data and the enterprise asset data to be evaluated by adopting a data processing and analyzing technology based on deep learning, so as to automatically obtain whether the credit of the enterprise asset data to be evaluated is a good processing result according to the semantic interactive matching characteristic between the good credit enterprise asset data and the good credit enterprise asset data, realize the automatic processing and analysis of the enterprise asset data to be evaluated, reduce manual intervention and subjective judgment, and simultaneously capture the potential association and characteristics between the data, thereby improving the evaluation accuracy of the credit status of the enterprise.
The basic principles of the present application have been described above in connection with specific embodiments, but it should be noted that the advantages, benefits, effects, etc. mentioned in the present application are merely examples and not intended to be limiting, and these advantages, benefits, effects, etc. are not to be construed as necessarily possessed by the various embodiments of the application. Furthermore, the specific details of the foregoing application are for purposes of illustration and understanding only, and are not intended to limit the application to the particular details described above.
Claims (6)
1. An asset data processing system for enterprise credit risk management, comprising:
a good enterprise asset data extraction module for extracting from the database a collection of enterprise asset data labeled as good credit;
the credit-good enterprise asset data semantic coding module is used for semantically coding each enterprise asset data marked as credit-good in the set of enterprise asset data marked as credit-good to obtain a set of credit-good enterprise asset data semantic coding feature vectors;
The significance global optimization module is used for inputting the set of the semantic coding feature vectors of the credit-good enterprise asset data into the context semantic-based significance global optimization module to obtain the set of the semantic coding feature vectors of the credit-good enterprise asset data;
The enterprise asset data acquisition module to be evaluated is used for acquiring enterprise asset data to be evaluated;
The enterprise asset data semantic coding module to be evaluated is used for performing semantic coding on the enterprise asset data to be evaluated to obtain semantic coding feature vectors of the enterprise asset data to be evaluated;
The unidirectional interaction matching module is used for taking the semantic coding feature vector of the enterprise asset data to be evaluated as a query feature vector, inputting the query feature vector and the set of semantic coding feature vectors of the enterprise asset data with good optimization credit into the unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as unidirectional matching result representation features;
The processing result generation module is used for obtaining a processing result based on the unidirectional matching result representing characteristics, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not;
Wherein, the saliency global optimization module comprises:
the salient feature calculating unit is used for extracting the maximum value of each credit-good enterprise asset data semantic coding feature vector in the collection of the credit-good enterprise asset data semantic coding feature vectors to obtain the salient credit-good enterprise asset data semantic coding feature vector;
the global feature calculation unit is used for extracting the average value of all the semantic encoding feature vectors of the credit good enterprise asset data in the collection of the semantic encoding feature vectors of the credit good enterprise asset data so as to obtain the semantic encoding feature vector of the global credit good enterprise asset data;
The convolution activating unit is used for carrying out convolution encoding and feature activating processing on the characteristic vector of the semantic encoding of the outstanding credit good enterprise asset data and the characteristic vector of the semantic encoding of the global credit good enterprise asset data so as to obtain an activating vector of the semantic encoding of the outstanding feature credit good enterprise asset data and an activating vector of the semantic encoding of the global feature credit good enterprise asset data;
The salient global fusion unit is used for fusing the salient feature credit good enterprise asset data semantic coding activation vector and the global feature credit good enterprise asset data semantic coding activation vector to obtain a global-salient feature credit good enterprise asset data semantic coding feature vector;
The nonlinear activation unit is used for carrying out nonlinear activation on the global-salient feature credit good enterprise asset data semantic coding feature vector so as to obtain a global-salient feature credit good enterprise asset data semantic coding weight feature vector;
The feature weighting fusion unit is used for taking the global-salient feature credit good enterprise asset data semantic coding weight feature vector as a weight, multiplying the set of credit good enterprise asset data semantic coding feature vectors by position points and adding the set of credit good enterprise asset data semantic coding feature vectors to obtain the set of optimized credit good enterprise asset data semantic coding feature vectors;
wherein, the convolution activation unit is used for:
Inputting the semantic convolution encoding vector of the good enterprise asset data with outstanding credit, which is obtained after the semantic encoding feature vector of the good enterprise asset data with outstanding credit is subjected to one-dimensional convolution encoding, into a ReLU function for feature activation processing so as to obtain the semantic activation encoding feature vector of the good enterprise asset data with outstanding credit;
Performing matrix multiplication on the salient point convolution feature vector obtained after the salient point convolution encoding feature vector is subjected to point convolution encoding on the salient feature credit good enterprise asset data semantic activation encoding feature vector and a first weight matrix to obtain the salient feature credit good enterprise asset data semantic encoding activation vector;
Inputting the global credit good enterprise asset data semantic convolution coding vector obtained after the global credit good enterprise asset data semantic coding feature vector is subjected to one-dimensional convolution coding into a ReLU function to perform feature activation processing so as to obtain the global credit good enterprise asset data semantic activation coding feature vector;
Performing matrix multiplication on the global point convolution characteristic vector obtained after the point convolution encoding of the global credit good enterprise asset data semantic activation encoding characteristic vector and a second weight matrix to obtain the global feature credit good enterprise asset data semantic encoding activation vector;
wherein, the one-way interaction matching module includes:
The semantic similarity calculation unit is used for calculating semantic similarity between each optimization credit good enterprise asset data semantic coding feature vector in the query feature vector and the set of optimization credit good enterprise asset data semantic coding feature vectors so as to obtain a sequence of enterprise asset data unidirectional matching semantic measurement coefficients;
The normalization processing unit is used for performing normalization processing on the sequence of the one-way matching semantic measurement coefficient of the enterprise asset data to obtain a sequence of one-way matching semantic measurement weight values of the enterprise asset data;
The feature weighting unit is used for calculating the weighted sum of the set of the semantic coding feature vectors of the optimization credit good enterprise asset data by taking the sequence of the one-way matching semantic measurement weight values of the enterprise asset data as weight so as to obtain one-way matching interaction remodelling feature vectors of the enterprise asset data;
The differential feature calculation unit is used for calculating a differential feature vector between the enterprise asset data unidirectional matching interaction remodelling feature vector and the query feature vector as the unidirectional matching result representation vector;
the semantic similarity calculation unit is used for:
Respectively cascading each optimization credit good enterprise asset data semantic coding feature vector in the query feature vector and the set of optimization credit good enterprise asset data semantic coding feature vectors to obtain a set of query enterprise asset data semantic coding joint feature vectors;
After the matrix product of each query enterprise asset data semantic coding joint feature vector and the weight matrix in the set of query enterprise asset data semantic coding joint feature vectors is calculated respectively, the obtained feature vectors and the bias vectors are added according to positions to obtain a sequence of the query enterprise asset data semantic coding joint bias feature vectors;
inputting the sequence of the query enterprise asset data semantic coding joint bias feature vector into a sigmoid activation function to obtain the sequence of the enterprise asset data unidirectional matching semantic measurement coefficients.
2. The asset data processing system for enterprise credit risk management of claim 1, wherein the nonlinear activation unit is configured to:
Inputting the global-salient feature credit good enterprise asset data semantic coding feature vector into a tanh function to obtain a first activated global-salient feature credit good enterprise asset data semantic coding feature vector;
Inputting the global-salient feature credit good enterprise asset data semantic coding feature vector into a Sigmoid function to obtain a second activated global-salient feature credit good enterprise asset data semantic coding feature vector;
And multiplying the first activated global-salient feature credit good enterprise asset data semantic coding feature vector and the second activated global-salient feature credit good enterprise asset data semantic coding feature vector by location points to obtain the global-salient feature credit good enterprise asset data semantic coding weight feature vector.
3. The asset data processing system for enterprise credit risk management of claim 2, wherein the normalization processing unit is configured to:
taking each enterprise asset data one-way matching semantic measurement coefficient in the sequence of enterprise asset data one-way matching semantic measurement coefficients as an index of a natural constant to calculate an index function value based on the natural constant so as to obtain a sequence of enterprise asset data one-way matching semantic measurement coefficient logarithmic values;
calculating the sum of the logarithmic values of the unidirectional matching semantic measurement coefficients of the enterprise asset data in the sequence of the logarithmic values of the unidirectional matching semantic measurement coefficients of the enterprise asset data to obtain the total sum of the unidirectional matching semantic measurement coefficients of the enterprise asset data;
Dividing each enterprise asset data one-way matching semantic measurement coefficient pair value in the enterprise asset data one-way matching semantic measurement coefficient pair value sequence with the enterprise asset data one-way matching semantic measurement total sum value to obtain the enterprise asset data one-way matching semantic measurement weight value sequence.
4. An asset data processing system for enterprise credit risk management according to claim 3, wherein the feature weighting unit is configured to:
Multiplying each enterprise asset data one-way matching semantic measurement weight value in the sequence of enterprise asset data one-way matching semantic measurement weight values by a corresponding optimization credit good enterprise asset data semantic coding feature vector in the set of optimization credit good enterprise asset data semantic coding feature vectors according to positions to obtain a set of optimization credit good enterprise asset data semantic coding weight feature vectors;
And calculating the position-wise summation of the set of the semantic coding weight feature vectors of the optimization credit good enterprise asset data to obtain the one-way matching interaction remodelling feature vector of the enterprise asset data.
5. The asset data processing system for enterprise credit risk management of claim 4, wherein the processing result generation module is configured to: and inputting the unidirectional matching result representation vector into an asset data processing result generator based on the classifier to obtain a processing result, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not.
6. An asset data processing method for enterprise credit risk management, applied to an asset data processing system for enterprise credit risk management as claimed in claim 5, comprising:
Extracting from the database a collection of enterprise asset data labeled as good credit;
Performing semantic coding on each enterprise asset data marked as good credit in the set of enterprise asset data marked as good credit to obtain a set of semantic coding feature vectors of the enterprise asset data marked as good credit;
Inputting the set of the semantic coding feature vectors of the credit-good enterprise asset data into a context semantic-based significance global optimization module to obtain a set of the semantic coding feature vectors of the optimization credit-good enterprise asset data;
Acquiring enterprise asset data to be evaluated;
performing semantic coding on the enterprise asset data to be evaluated to obtain semantic coding feature vectors of the enterprise asset data to be evaluated;
Taking the semantic coding feature vector of the enterprise asset data to be evaluated as a query feature vector, inputting the query feature vector and the set of semantic coding feature vectors of the enterprise asset data with good optimization credit into a unidirectional attention granularity-by-granularity scanning interaction matching module to obtain a unidirectional matching result representation vector as a unidirectional matching result representation feature;
and obtaining a processing result based on the unidirectional matching result representation feature, wherein the processing result is used for representing whether the credit of the enterprise asset data to be evaluated is good or not.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411035356.9A CN118569983B (en) | 2024-07-31 | 2024-07-31 | Asset data processing system and method for enterprise credit risk management |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411035356.9A CN118569983B (en) | 2024-07-31 | 2024-07-31 | Asset data processing system and method for enterprise credit risk management |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118569983A CN118569983A (en) | 2024-08-30 |
| CN118569983B true CN118569983B (en) | 2024-09-20 |
Family
ID=92478606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411035356.9A Active CN118569983B (en) | 2024-07-31 | 2024-07-31 | Asset data processing system and method for enterprise credit risk management |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118569983B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119336744A (en) * | 2024-12-13 | 2025-01-21 | 数字丝路新疆建设运营有限公司 | Data quality monitoring system and method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112750029A (en) * | 2020-12-30 | 2021-05-04 | 北京知因智慧科技有限公司 | Credit risk prediction method, device, electronic equipment and storage medium |
| CN116777608A (en) * | 2023-08-25 | 2023-09-19 | 北京佳格天地科技有限公司 | Agricultural financial risk supervision system, method and storage medium based on big data |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040215551A1 (en) * | 2001-11-28 | 2004-10-28 | Eder Jeff S. | Value and risk management system for multi-enterprise organization |
| CN117271981B (en) * | 2023-11-21 | 2024-02-09 | 湖南嘉创信息科技发展有限公司 | Artificial intelligence management system based on cross-platform data interaction |
| CN117764724A (en) * | 2024-01-02 | 2024-03-26 | 中证鹏元资信评估股份有限公司 | Intelligent credit rating report construction method and system |
-
2024
- 2024-07-31 CN CN202411035356.9A patent/CN118569983B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112750029A (en) * | 2020-12-30 | 2021-05-04 | 北京知因智慧科技有限公司 | Credit risk prediction method, device, electronic equipment and storage medium |
| CN116777608A (en) * | 2023-08-25 | 2023-09-19 | 北京佳格天地科技有限公司 | Agricultural financial risk supervision system, method and storage medium based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118569983A (en) | 2024-08-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111401914A (en) | Risk assessment model training and risk assessment method and device | |
| CN118569983B (en) | Asset data processing system and method for enterprise credit risk management | |
| US20190080352A1 (en) | Segment Extension Based on Lookalike Selection | |
| CN116911929B (en) | Advertising service terminal and method based on big data | |
| CN118410416B (en) | Marketing electric charge recycling early warning method and system based on multi-mode double-channel attention | |
| CN117455059A (en) | Industry trend evaluation system based on data acquisition | |
| CN119049682A (en) | Medical asset management big data analysis method and system | |
| CN120235678A (en) | An intelligent enterprise management platform based on large language model | |
| CN111143533A (en) | Customer service method and system based on user behavior data | |
| CN116187524A (en) | A method and device for comparing supply chain analysis models based on machine learning | |
| CN119249221A (en) | Intelligent decision making system based on user prior information guidance | |
| CN118761842A (en) | A method and device for managing information security risk assessment of online transactions | |
| CN118537144A (en) | Agricultural insurance security risk prediction method and system | |
| CN116843345A (en) | Intelligent wind control system and method for trading clients based on artificial intelligence technology | |
| CN117132288A (en) | Analysis method, device, equipment and storage medium for easy-to-lose clients | |
| CN116956289A (en) | Method for dynamically adjusting potential blacklist and blacklist | |
| CN114581209A (en) | Method, device and equipment for training financial analysis model and storage medium | |
| CN116484230B (en) | Method for identifying abnormal business data and training method of AI digital person | |
| CN117745459B (en) | Flexible accounting method based on entry configuration | |
| CN119963064B (en) | Practical training base employment tracking investigation information processing method and system | |
| Ifriza et al. | Prediction of Peer-To-Peer Lending Default Risk Using Stacking Ensemble Learning Model and Random Forest Feature Selection | |
| CN119886539A (en) | Method and system for selecting groups of portrayal groups of enterprises and users | |
| CN120634279A (en) | Enterprise client wind control management method and system based on AI intelligence | |
| CN119513802A (en) | Data fraud detection method, device, computer equipment and storage medium | |
| Surekha et al. | Stock Price Forecasting by Time Series Analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |