CN115689717A - Enterprise risk early warning method, device, electronic equipment, medium and program product - Google Patents
Enterprise risk early warning method, device, electronic equipment, medium and program product Download PDFInfo
- Publication number
- CN115689717A CN115689717A CN202210261606.5A CN202210261606A CN115689717A CN 115689717 A CN115689717 A CN 115689717A CN 202210261606 A CN202210261606 A CN 202210261606A CN 115689717 A CN115689717 A CN 115689717A
- Authority
- CN
- China
- Prior art keywords
- emotion
- value
- enterprise
- risk
- text information
- 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.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure provides an enterprise risk early warning method, an enterprise risk early warning device, electronic equipment, a medium and a computer program product. The method and the device can be used in the technical field of artificial intelligence. The enterprise risk early warning method comprises the following steps: acquiring public opinion data of an enterprise, wherein the public opinion data comprises text information; determining a customer name, a risk label and an emotion value of the public opinion data according to the text information; acquiring transaction data of a corresponding enterprise according to the customer name; calculating a corresponding enterprise risk value according to the emotion value and the transaction data; and returning an enterprise risk early warning result according to the enterprise risk value.
Description
Technical Field
The present disclosure relates to the field of artificial intelligence technologies, and in particular, to an enterprise risk early warning method, apparatus, electronic device, medium, and computer program product.
Background
The traditional credit risk assessment of the bank mostly depends on information such as an asset liability statement, a profit statement, a cash flow statement and a tax payment situation of an enterprise, but many small and medium-sized micro enterprises generally lack the information, and can only rely on assessing and monitoring operation flow of the enterprise before, during and after credit or obtain the information through third party access modes such as credit data and business information, and the information has time lag. Under the condition of serious data loss, a mode of manually searching negative information of a client on the Internet is usually adopted as a data collecting means, under the background of big data, small and medium-sized micro-enterprises leave more and more public opinion traces in the Internet, and public opinion text data on the Internet is analyzed by an artificial intelligence means, so that the client risk information can be acquired more conveniently and timely, and the credit risk monitoring efficiency is improved. Meanwhile, the accuracy of enterprise risk assessment can be improved by combining with enterprise operation information for assessment.
Disclosure of Invention
In view of the above, the present disclosure provides a timely and accurate enterprise risk early warning method, apparatus, electronic device, computer-readable storage medium, and computer program product.
One aspect of the present disclosure provides an enterprise risk early warning method, including: acquiring public opinion data of an enterprise, wherein the public opinion data comprises text information; determining a client name, a risk label and an emotion value of the public opinion data according to the text information; acquiring transaction data of a corresponding enterprise according to the customer name; calculating a corresponding enterprise risk value according to the emotion value and the transaction data; and returning an enterprise risk early warning result according to the enterprise risk value.
According to the enterprise risk early warning method disclosed by the embodiment of the disclosure, public sentiment data and transaction data of an enterprise are combined for computational analysis, so that client risk information can be acquired more conveniently and timely, and the credit risk monitoring efficiency and accuracy are improved. Therefore, the timeliness and the accuracy of enterprise risk assessment can be improved.
In some embodiments, the determining, according to the text information, a name of a customer of the public opinion data specifically includes: and matching the text information with a pre-constructed enterprise name library to obtain the client name of the public opinion data.
In some embodiments, the determining a risk label of the public opinion data according to the text information specifically includes: and matching the text information with a pre-constructed risk label library to obtain a risk label of the public opinion data.
In some embodiments, the determining an emotion value of the public opinion data according to the text information specifically includes: determining a keyword set in the text information; determining a relevant word set of the keyword set according to the degree of association in the text information; calculating emotion orientation probability according to the keyword set and the associated word set; and determining the emotion value according to the emotion orientation probability.
In some embodiments, the determining the keyword set in the text information specifically includes: extracting m keywords in the text information, wherein m is an integer greater than or equal to 1; and taking the m keywords as the keyword set.
In some embodiments, the determining, according to the degree of association, a set of associated words of the keyword set in the text information specifically includes: matching n similar words of each keyword in the keyword set in the text information, wherein n is an integer greater than or equal to 1; calculating the association degree of the keyword and each similar word in the n similar words; taking the similar word with the highest degree of association with the keyword in the n similar words as the associated word corresponding to the keyword; and taking the m associated words as the associated word set.
In some embodiments, the calculating the association degree between the keyword and each of the n similar words specifically includes: calculating the co-occurrence similarity of the keyword and each similar word in the n similar words according to the times of the keyword in the text information and each similar word in the n similar words appearing at the same time and the times of the keyword appearing in the text information; calculating semantic similarity between the keyword and each similar word in the n similar words according to the word vector of the keyword and the word vector of the similar word; and multiplying the co-occurrence similarity and the semantic similarity to obtain the association degree of the keyword and each similar word in the n similar words.
In some embodiments, the calculating an emotion orientation probability according to the keyword set and the associated word set specifically includes: determining an aspect level feature vector of the text information according to the relevant word set; determining a context semantic feature vector of the text information according to the keyword set; calculating emotion weight according to the aspect level feature vector and the context semantic feature vector; and calculating the emotion orientation probability according to the emotion weight.
In some embodiments, the determining the emotion value according to the emotion orientation probability includes: comparing the positive emotion probability value with the negative emotion probability value; and taking the maximum emotion probability value as the emotion value.
In some embodiments, the calculating a corresponding enterprise risk value according to the sentiment value and the transaction data specifically includes: calculating a transaction risk value according to the transaction data; and calculating an enterprise risk value according to the emotion value and the transaction risk value.
In some embodiments, returning an enterprise risk early warning result according to the enterprise risk value specifically includes: comparing the enterprise risk value with a preset risk threshold value; and when the enterprise risk value is larger than or equal to the risk threshold value, returning an enterprise risk early warning result.
Another aspect of the present disclosure provides an enterprise risk early warning device, including: the first acquisition module is used for executing acquisition of public opinion data of an enterprise, and the public opinion data comprises text information; the determining module is used for determining the client name, the risk label and the emotion value of the public opinion data according to the text information; the second acquisition module is used for acquiring the transaction data of the corresponding enterprise according to the customer name; the calculation module is used for calculating a corresponding enterprise risk value according to the emotion value and the transaction data; and the return module is used for returning the enterprise risk early warning result according to the enterprise risk value.
Another aspect of the present disclosure provides an electronic device comprising one or more processors and one or more memories, wherein the memories are used for storing executable instructions, which when executed by the processors, implement the method as described above.
Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions for implementing the method as described above when executed.
Another aspect of the disclosure provides a computer program product comprising a computer program comprising computer executable instructions for implementing the method as described above when executed.
Drawings
The above and other objects, features and advantages of the present disclosure will become more apparent from the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:
fig. 1 schematically illustrates an exemplary system architecture to which the methods, apparatus, and methods may be applied, in accordance with an embodiment of the present disclosure;
FIG. 2 schematically illustrates a flow chart of an enterprise risk early warning method according to an embodiment of the disclosure;
fig. 3 schematically shows a flowchart of determining a customer name of public opinion data according to text information according to an embodiment of the present disclosure;
fig. 4 schematically shows a flowchart of determining a risk label of public opinion data according to text information according to an embodiment of the present disclosure;
fig. 5 schematically shows a flowchart of determining an emotion value of public opinion data according to text information according to an embodiment of the present disclosure;
FIG. 6 schematically illustrates a flow chart for determining a set of keywords in a text message according to an embodiment of the disclosure;
fig. 7 schematically shows a flowchart of determining a set of related words of a set of keywords according to a degree of association in text information according to an embodiment of the present disclosure;
FIG. 8 schematically illustrates a flow chart for calculating a degree of association of a keyword with each of n similar words according to an embodiment of the present disclosure;
FIG. 9 schematically shows a flow chart for calculating the emotion orientation probability based on a keyword set and a related word set according to an embodiment of the present disclosure;
FIG. 10 schematically illustrates a flow chart for determining sentiment value from sentiment orientation probability according to an embodiment of the present disclosure;
FIG. 11 schematically illustrates a flow chart for calculating a corresponding business risk value based on an sentiment value and transactional data according to an embodiment of the disclosure;
FIG. 12 schematically illustrates a flow chart for returning enterprise risk early warning results based on enterprise risk values, according to an embodiment of the present disclosure;
FIG. 13 schematically illustrates a block diagram of an enterprise risk early warning device, according to an embodiment of the present disclosure;
FIG. 14 schematically illustrates a block diagram of an aspect level sentiment analysis model based on an attention mechanism according to an embodiment of the present disclosure;
FIG. 15 schematically shows a block diagram of an electronic device according to an embodiment of the disclosure.
Detailed Description
Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is illustrative only and is not intended to limit the scope of the present disclosure. In the following detailed description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the disclosure. It may be evident, however, that one or more embodiments may be practiced without these specific details. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present disclosure.
In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, necessary security measures are taken, and the commonness and the customs are not violated. In the technical scheme of the disclosure, the data acquisition, collection, storage, use, processing, transmission, provision, disclosure, application and other processing are all in accordance with the regulations of relevant laws and regulations, necessary security measures are taken, and the public order and good custom are not violated.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The terms "comprises," "comprising," and the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.
In those instances where a convention analogous to "at least one of A, B, or C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B, or C" would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and/or A, B, and C together, etc.). The terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", may explicitly or implicitly include one or more of the described features.
The traditional credit risk assessment of the bank mostly depends on information such as an asset liability statement, a profit statement, a cash flow statement and a tax payment situation of an enterprise, but many small and medium-sized micro enterprises generally lack the information, and can only rely on assessing and monitoring operation flow of the enterprise before, during and after credit or obtain the information through third party access modes such as credit data and business information, and the information has time lag. Under the condition of serious data loss, a mode of manually searching negative information of a client on the Internet is usually adopted as a data collecting means, under the background of big data, small and medium-sized micro-enterprises leave more and more public opinion traces in the Internet, and public opinion text data on the Internet is analyzed by an artificial intelligence means, so that the client risk information can be acquired more conveniently and timely, and the credit risk monitoring efficiency is improved. Meanwhile, the accuracy of enterprise risk assessment can be improved by combining with enterprise operation information for assessment.
Embodiments of the present disclosure provide an enterprise risk early warning method, apparatus, electronic device, computer-readable storage medium, and computer program product. The enterprise risk early warning method comprises the following steps: acquiring public opinion data of an enterprise, wherein the public opinion data comprises text information; determining a client name, a risk label and an emotion value of public sentiment data according to the text information; acquiring transaction data of a corresponding enterprise according to the name of the client; calculating a corresponding enterprise risk value according to the emotion value and the transaction data; and returning an enterprise risk early warning result according to the enterprise risk value.
It should be noted that the enterprise risk early warning method, apparatus, electronic device, computer-readable storage medium, and computer program product of the present disclosure may be used in the field of artificial intelligence, and may also be used in any field other than the field of artificial intelligence, for example, the field of finance, and the field of the present disclosure is not limited herein.
Fig. 1 schematically illustrates an exemplary system architecture 100 to which the enterprise risk early warning method, apparatus, electronic device, computer-readable storage medium and computer program product may be applied, according to embodiments of the disclosure. It should be noted that fig. 1 is only an example of a system architecture to which the embodiments of the present disclosure may be applied to help those skilled in the art understand the technical content of the present disclosure, and does not mean that the embodiments of the present disclosure may not be applied to other devices, systems, environments or scenarios.
As shown in fig. 1, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The user may use the terminal devices 101, 102, 103 to interact with the server 105 via the network 104 to receive or send messages or the like. The terminal devices 101, 102, 103 may have installed thereon various communication client applications, such as shopping-like applications, web browser applications, search-like applications, instant messaging tools, mailbox clients, social platform software, etc. (by way of example only).
The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.
The server 105 may be a server providing various services, such as a background management server (for example only) providing support for websites browsed by users using the terminal devices 101, 102, 103. The backend management server may analyze and process the received data such as the user request, and feed back a processing result (for example, a web page, information, or data obtained or generated according to the user request) to the terminal device.
It should be noted that the enterprise risk early warning method provided by the embodiment of the present disclosure may be generally executed by the server 105. Accordingly, the enterprise risk early warning apparatus provided by the embodiments of the present disclosure may be generally disposed in the server 105. The enterprise risk early warning method provided by the embodiment of the present disclosure may also be executed by a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105. Correspondingly, the enterprise risk early warning apparatus provided by the embodiment of the present disclosure may also be disposed in a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
The enterprise risk early warning method according to the embodiment of the disclosure will be described in detail below with reference to fig. 2 to 12 based on the scenario described in fig. 1.
Fig. 2 schematically shows a flowchart of an enterprise risk early warning method according to an embodiment of the present disclosure.
As shown in fig. 2, the enterprise risk early warning method of the embodiment includes operations S210 to S250.
In operation S210, public opinion data of a business is acquired, the public opinion data including text information. The public opinion data may be understood as news data about stock right change of an enterprise obtained from channels such as a news portal and a network media, but is not limited thereto, and may also be other news information about the enterprise obtained from other channels, and is not particularly limited herein.
In operation S220, a customer name, a risk label, and an emotion value of the public opinion data are determined according to the text information.
As a possible implementation manner, as shown in fig. 3, the operation S220 determines a name of a client of the public opinion data according to the text information, and specifically includes an operation S310.
In operation S310, the text information is matched with a pre-constructed company name library to obtain a customer name of the public opinion data. The enterprise names in the enterprise name library comprise enterprise full names and enterprise short names, when the text information is matched with the enterprise name library, the text information can be split into a plurality of participles, the participles are matched with the enterprise names in the enterprise name library, and if the enterprise names are matched, the enterprise names are client names corresponding to the text information. Determining a business corresponding to the public opinion data may be facilitated through operation S310.
As a possible implementation manner, as shown in fig. 4, operation S220 determines a risk label of public opinion data according to the text information, and specifically includes operation S410.
In operation S410, the text information is matched with a pre-constructed risk label library to obtain a risk label of public opinion data. The risk labels in the risk label library may be calculated according to the cosine similarity, and the risk labels may be, for example, an external guarantee, an overdue credit, and the like, but are not limited thereto.
When the text information is matched with the risk label library, the text information can be divided into a plurality of participles, the participles are matched with the risk labels in the risk label library, and if the risk labels are matched, the risk labels can be used as the risk labels corresponding to the text information. When the text information is matched with the risk label library, semantic information of the text information can be extracted, the semantic information is matched with the risk labels in the risk label library, and if the risk labels are matched, the risk labels can be used as risk labels corresponding to the text information.
It may be convenient to determine a risk label corresponding to the public opinion data through operation S410.
As a possible implementation manner, as shown in fig. 5, the operation S220 determines the sentiment value of the public sentiment data according to the text information, and specifically includes operations S510 to S540.
In operation S510, a set of keywords in the text information is determined.
In some specific examples, as shown in fig. 6, operation S510 determines a keyword set in the text message, specifically including operation S511 and operation S512.
In operation S511, m keywords in the text information are extracted, where m is an integer greater than or equal to 1.
In operation S512, m keywords are taken as a keyword set.
The determination of the keyword set in the text information may be facilitated through operations S511 and S512.
In operation S520, in the text information, a related word set of the keyword set is determined according to the degree of association.
In some specific examples, as shown in fig. 7, the operation S520 specifically includes operations S521 to S524, in which a set of related words of the keyword set is determined according to the degree of association in the text information.
In operation S521, n similar words of each keyword in the keyword set are matched in the text information, where n is an integer greater than or equal to 1. The n similar words matching the keywords can be understood as matching according to a matching rule, and the matching rule can be matched with words with similar semantics to the keywords; a matching rule may also be understood as matching words having a number of identical words to the keyword, a being an integer greater than or equal to 1.
In operation S522, a degree of association of the keyword with each of the n similar words is calculated.
In operation S523, the similar word having the highest degree of association with the keyword among the n similar words is set as the related word corresponding to the keyword.
In operation S524, the m related words are set as a set of related words.
Thus, the related-word set of the keyword set can be easily specified in the text information according to the degree of association in operations S521 to S524.
As a possible implementation manner, as shown in fig. 8, operation S522 calculates the association degree of the keyword with each similar word in the n similar words, specifically including operations S5221 to S5223.
In operation S5221, a co-occurrence similarity between the keyword and each of the n similar words is calculated according to the number of times that the keyword appears simultaneously with each of the n similar words in the text information and the number of times that the keyword appears in the text information. For example, the co-occurrence similarity may be represented by d co-occurrence And the co-occurrence similarity can be calculated by formula (1).
Wherein, w i Representing a keyword, w j Denotes the analogous word, N (w) i ，w j ) Representing a keyword w i And the like w j Number of simultaneous occurrences, N (w) i ) Representing a keyword w i The number of occurrences.
In operation S5222, semantic similarity between the keyword and each of the n similar words is calculated according to the word vector of the keyword and the word vectors of the similar words. For example, semantic similarity may be represented by d semantic (w i ，w j ) And (4) showing. And the semantic similarity can be calculated by formula (2).
Wherein the content of the first and second substances,
represents a keyword w i The vector of (a) is determined,
denotes the similar word w j The vector of (2).
In operation S5223, the co-occurrence similarity is multiplied by the semantic similarity to obtain the association degree between the keyword and each of the n similar words. For example, the degree of association may be weight (w) ij ) And the degree of association can be calculated by formula (3).
weight(w ij )＝d co-occurrence (w i ，w j )·d semantic (w i ，w j ) (3)
Therefore, n degrees of relevance corresponding to the keywords can be obtained, and the similar words with the highest degree of relevance to the keywords can be obtained by comparing the n degrees of relevance, so that the similar words can be used as the related words corresponding to the keywords. The calculation of the degree of association of the keyword with each of the n similar words may be facilitated through operations S5221 to S5223.
In operation S530, emotion orientation probabilities are calculated from the keyword set and the associated word set.
As a possible implementation manner, as shown in FIG. 9, the operation S530 calculates the emotion orientation probability according to the keyword set and the related word set, and specifically includes operations S531 to S534.
In operation S531, an aspect-level feature vector of the text information is determined from the set of related words. It should be noted that each related Word in the related Word set can be represented by a Word vector using a Word2Vec Word vector model, and given a Word embedding matrix, each related Word is represented by a multidimensional vector, so that a related Word vector set x 'and each related Word vector x' i Denotes, x '= { x' 1 ，x’ 2 ，......，x’ n Wherein n represents the number of associated word vectors in the associated word vector set, n is an integer of 1 or more, and i is 1 or more and 1 or lessn is an integer.
Further, the relevant word vector set can be used as the input of the LSTM neural network model, and the hidden state h of the LSTM neural network model at the time t in the two-way transmission process is set t The final output of the LSTM neural network model is a sequence h t ＝{h 1 ，h 2 ，......，h n H, the state of the last moment of the sequence h due to the characteristics of the LSTM neural network model n The state information of all relevant words in the front is basically contained, the aspect level characteristics are maximally reserved, and therefore the state is output to h n Aspect level feature vector h as text aspect 。
By analyzing the aspect level characteristics, the analysis granularity is more detailed. It should be noted that, according to the difference of granularity, the text emotion analysis can be divided into the following three types: chapter level, sentence level, aspect level. The chapter level mainly classifies the emotional polarity of the whole text, the sentence level mainly aims at the emotional tendency of each sentence in the text, and the aspect level needs to dig out the entity expressing the viewpoint of the viewpoint person in the sentence and the attribute thereof so as to classify the emotional tendency of one aspect of the entity. For example, in the comment of "the mobile phone screen is large, but the battery consumes fast", the emotion polarity of the comment is positive for the term of "the mobile phone screen", and negative emotion polarity for the term of "the battery", and when only sections or sentences are judged, the positive and negative polarities are offset, and fine-grained information in the public opinion text is ignored, so that fine-grained aspect-level emotion analysis is necessary for the text.
In operation S532, a context semantic feature vector of the text information is determined according to the keyword set. Similarly, word2Vec Word vector model can be used to represent each keyword in the keyword set by Word vector, a Word embedding matrix is given, each keyword is represented by a multidimensional vector, so that a keyword vector set x can be obtained, and each keyword vector is represented by x i It is shown that, x = { x 1 ，x 2 ，......，x n N represents the number of the keyword vectors in the keyword vector set, and n is more than or equal toI is an integer of 1 or more and n or less. Further, the keyword vector set can be used as the input of the LSTM neural network model, and the context semantic feature vector h can be obtained through the LSTM neural network model t semantic 。
In operation S533, emotion weights are calculated from the aspect-level feature vector and the context semantic feature vector. For example, the sentiment weight may be given by α t It is shown that the emotion weight can be calculated by formula (4) and formula (5).
Wherein u is t And representing the degree of association between the aspect level feature vector and the context semantic feature vector, W represents a weight matrix, and b represents an offset. Here, introduce
Measure u t The importance degree of the emotion is normalized to form probability distribution, and finally, the emotion weight alpha is obtained t 。
In operation S534, the emotion orientation probability is calculated according to the emotion weight. For example, the emotion orientation probability can be represented by y, and the emotion orientation probability can be obtained by formula (6) and formula (7).
y＝softmax(W s ·o+b s )＝(y 1 ，y 2 ) (7)
The intermediate quantity can be calculated through the emotion weight and the context semantic feature vector, the intermediate quantity is used as the input of the soffmax function, the emotion orientation probability can be obtained, wherein 1 > y 1 ＞0，1＞y 2 ＞0，y 1 +y 2 ＝1，W s Weight matrix for the classification layer of the softmax function, b s The offset of the layer is classified for the softmax function.
Therefore, the emotion orientation probability can be calculated from the keyword set and the related word set through operations S531 to S534.
In operation S540, an emotion value is determined according to the emotion orientation probability.
As one possible implementation, the emotion orientation probabilities include positive emotion probability values and negative emotion probability values, e.g., y can be expressed 1 Set as the forward emotion probability value, let y 2 As shown in fig. 10, operation S540 determines an emotion value according to the emotion orientation probability, which specifically includes operation S541 and operation S542.
In operation S541, the magnitudes of the positive emotion probability value and the negative emotion probability value are compared.
In operation S542, the maximum emotion probability value is used as an emotion value. For example, y may be compared 1 And y 2 When y is 1 Greater than y 2 When, get y 1 As an emotional value; when y is 1 Less than y 2 When, get y 2 As an emotional value.
The determination of the emotion value from the emotion orientation probability can be facilitated by operations S541 and S542.
In operation S230, transaction data of the corresponding business is acquired according to the customer name. Wherein the transaction data may include, but is not limited to, customer name, contract information, loan currency, current financing amount, affiliated group customers, etc. reflecting the customer's credit basis business data at the bank.
In operation S240, a corresponding enterprise risk value is calculated based on the sentiment value and the transaction data.
As one practical way, as shown in fig. 11, operation S240 calculates a corresponding enterprise risk value according to the sentiment value and the transaction data, and specifically includes operations S241 and S242.
In operation S241, a transaction risk value is calculated according to the transaction data. For example, the transaction data includes a customer financing balance, a group financing balance to which the customer belongs, an I-team customer total financing balance, and an I-team customer total financing balance. The transaction risk value is represented by B and can be calculated by formula (8).
Wherein, alpha is the balance weight of the client financing, beta is the balance weight of the group client financing, and alpha + beta =1.
In operation S242, a business risk value is calculated based on the sentiment value and the transaction risk value. For example, the enterprise risk value is represented by R, and can be calculated by formula (9).
R＝k*y*B (9)
K is a constant, the value of k is a real number, and preferably, the value of k is a positive integer; preferably, k has a value of 100, in order to show the result. Calculating a corresponding enterprise risk value based on the sentiment value and the transaction data may be facilitated through operations S241 and S242.
And returning an enterprise risk early warning result according to the enterprise risk value in operation S250.
As an implementable manner, as shown in fig. 12, operation S250 returns the enterprise risk early warning result according to the enterprise risk value, and specifically includes operation S251 and operation S252.
In operation S251, the enterprise risk value is compared with a preset risk threshold.
In operation S252, when the enterprise risk value is greater than or equal to the risk threshold, an enterprise risk early warning result is returned. The risk early warning result can be at least one of an enterprise name, an enterprise label, an enterprise risk value, a positive emotion probability value and a negative emotion probability value.
According to the enterprise risk early warning method disclosed by the embodiment of the disclosure, public sentiment data and transaction data of an enterprise are combined for computational analysis, so that client risk information can be acquired more conveniently and timely, and the credit risk monitoring efficiency and accuracy are improved. Therefore, timeliness and accuracy of enterprise risk assessment can be improved.
Based on the enterprise risk early warning method, the disclosure also provides an enterprise risk early warning device 10. The enterprise risk early warning device 10 will be described in detail with reference to fig. 13.
Fig. 13 schematically shows a block diagram of the structure of the enterprise risk early warning device 10 according to the embodiment of the present disclosure.
The enterprise risk early warning device 10 comprises a first obtaining module 1, a determining module 2, a second obtaining module 3, a calculating module 4 and a returning module 5.
A first obtaining module 1, where the first obtaining module 1 is configured to perform operation S210: public opinion data of an enterprise is obtained, and the public opinion data comprises text information.
A determining module 2, the determining module 2 being configured to perform operation S220: and determining the client name, the risk label and the emotion value of the public opinion data according to the text information.
A second obtaining module 3, where the second obtaining module 3 is configured to execute operation S230: and acquiring the transaction data of the corresponding enterprise according to the name of the client.
A calculating module 4, the calculating module 4 being configured to perform operation S240: and calculating a corresponding enterprise risk value according to the emotion value and the transaction data.
A return module 5, the return module 5 being configured to perform operation S250: and returning an enterprise risk early warning result according to the enterprise risk value.
According to this disclosed embodiment's risk early warning device of enterprise, through the public opinion data with the enterprise and trade data combination carry out computational analysis, can acquire customer's risk information more conveniently and in time, improve credit risk monitoring efficiency and rate of accuracy. Therefore, timeliness and accuracy of enterprise risk assessment can be improved.
In addition, according to the embodiment of the present disclosure, any multiple modules of the first obtaining module 1, the determining module 2, the second obtaining module 3, the calculating module 4, and the returning module 5 may be combined and implemented in one module, or any one module may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in one module.
According to an embodiment of the present disclosure, at least one of the first obtaining module 1, the determining module 2, the second obtaining module 3, the calculating module 4 and the returning module 5 may be at least partially implemented as a hardware circuit, such as a Field Programmable Gate Array (FPGA), a Programmable Logic Array (PLA), a system on a chip, a system on a substrate, a system on a package, an Application Specific Integrated Circuit (ASIC), or may be implemented by hardware or firmware in any other reasonable manner of integrating or packaging a circuit, or implemented by any one of three implementations of software, hardware and firmware, or implemented by a suitable combination of any several of them.
Alternatively, at least one of the first obtaining module 1, the determining module 2, the second obtaining module 3, the calculating module 4 and the returning module 5 may be at least partly implemented as a computer program module which, when executed, may perform a corresponding function.
A block diagram of an attention-based aspect-level sentiment analysis model according to an embodiment of the present disclosure is described in detail below with reference to fig. 14. It is to be understood that the following description is illustrative only and is not intended as a specific limitation of the disclosure.
The method and the device have the advantages that the aspect-level emotion analysis technology based on the attention mechanism is used for conducting fine-grained emotion analysis on client public opinion text data, and potential credit risks are judged by combining client transaction data, so that real-time monitoring and early warning of the credit risks are achieved.
1. The traditional emotion analysis method based on the emotion dictionary cannot accurately identify and analyze the content of spoken expressions, network expressions, acronyms and the like on the social media. Therefore, the emotion analysis model based on deep learning is adopted, and the emotion analysis accuracy is improved.
2. The emotion analysis method is in aspect level and can analyze fine-grained emotion tendencies. According to different granularities, text sentiment analysis can be divided into the following three types: chapter level, sentence level, aspect level. The chapter grade mainly classifies the emotional polarity of the whole text, the sentence grade mainly aims at the emotional tendency of each sentence in the text, and the aspect grade needs to dig out the entity expressing the viewpoint of the viewpoint person in the sentence and the attribute thereof so as to classify the emotional tendency of one aspect of the entity. For example, in the comment of "the mobile phone screen is large but the battery consumes fast", the emotion polarity is positive for the term of "the mobile phone screen", and negative emotion polarity for the entity of "the battery", only the piece of text or sentence is judged, the positive and negative polarities are offset, and fine-grained information in the public sentiment text is ignored, so that fine-grained aspect emotion analysis is necessary for the text.
3. The method and the device can effectively improve the effect of aspect-level emotion analysis by combining an attention mechanism on the basis of the aspect-level emotion analysis. The attention mechanism quantifies the importance degree of the words at each position to the objective function by carrying out weight calculation on the words at different positions in the text, thereby realizing the further extraction of the public sentiment text aspect information and improving the accuracy of the text aspect level sentiment analysis.
The "aspect-level-attention mechanism network" used in the present disclosure is different from the conventional attention mechanism network in that the conventional method mainly directly calculates the degree of association between the text context semantic features and the target aspect categories, and the "aspect-level-attention mechanism network" first learns the text aspect features based on the text context and then constructs attention weights between the text context semantic features and the aspect features. Calculating the correlation between text context semantic features and aspect features facilitates extracting potential context words relevant to text aspect category prediction.
The enterprise risk early warning device can comprise a data access module, a semantic emotion analysis module, a risk value evaluation module and a risk early warning module.
A data access module: the module is mainly responsible for accessing and storing two parts of data, namely credit transaction information data and public opinion data such as news portals, network media, debt transactions and the like. The transaction data comprises customer name, contract information, loan currency, current financing amount, affiliated group customers and the like, which reflect the credit basic business data of the customer in the bank. The public opinion data mainly comprises titles, texts, sources/authors, release time and the like of articles or information. The data access module provides the primarily processed structured data to the semantic emotion analysis module.
A semantic emotion analysis module: and performing semantic emotion analysis on the public sentiment data output by the data access module, outputting a client name, a risk label and an emotion value, and simultaneously using the 3 types of information as the input of the risk early warning module. (1) The client name is matched with the full name of the client and the short name of the client in an accurate matching mode, and the identification of the client to which the public opinion data belongs is realized. (2) The risk label adopts cosine similarity to calculate the similarity between the defined 51-class risk label (the label library still needs to be calculated) and the public opinion text theme, and classifies the risk label. (3) The emotion value is calculated through an aspect level emotion analysis model based on an attention mechanism, the model judges the positive direction and the negative direction of the article, and the emotion value is given and scored.
An aspect-level emotion analysis model based on an attention mechanism is shown in fig. 14, and comprises the following parts:
1) An input layer: and inputting public sentiment text of the emotion analysis model.
2) Embedding the layer: the Word co-occurrence feature of the text is obtained by using a text feature extraction algorithm based on the Word co-occurrence, a Word2Vec Word vector model is introduced to express words in the Word co-occurrence feature by using Word vectors, a Word embedding matrix is given, and each Word is expressed by using a 300-dimensional vector.
First, the co-occurrence similarity between two words, d, is calculated co-occurrence (w i ，w j ) Wherein N (w) i ，w j ) Representing simultaneous occurrences of a word w in a document i And w j Number of times of (N) (w) i ) Representing occurrences of word w in a document i The number of times of (c).
Secondly, calculating the semantic similarity d between every two words by using a cosine similarity formula semantic (w i ，w j ). The semantic similarity d between words semantic (w i ，w j ) Similarity to co-occurrence d co-occurrence (w i ，w j ) Multiplying as an evaluation function for measuring the degree of association between words and calculating the weight (w) of the degree of association ij )。
weight(w ij )＝d co-occurrence (w i ，w j )·d semantic (w i ，w j )
Then, for word w in the text i We select the word with the highest degree of association with the word, i.e., weight (w) ij ) Word w with highest weight j As the word w i The shadow word of (1) will [ w i ，w j ]Together as word co-occurrence features of the text.
Converting all words and ' shadow words ' into word vectors x and x ' to obtain a text word co-occurrence sequence x = { x = { (x) } 1 ，x 2 ，......，x n And x '= { x' 1 ，x’ 2 ，......，x’ n Where n represents the number of words in the text. The method can effectively extract the implicit semantic features and accurately describe the words with similar meanings but low co-occurrence frequency, so that the accuracy of subsequent sentiment analysis is improved.
3) LSTM layer: and obtaining the public opinion text aspect feature vector and the public opinion text context feature vector by utilizing the bidirectional LSTM.
The bidirectional LSTM layer is an extension of the unidirectional LSTM network by adding a second layer. The bidirectional LSTM network contains two sub-network structures representing forward and backward delivery, respectively.
In the formula (I), the compound is shown in the specification,
representing the output result of forward LSTM at time t,
shows the output result of backward LSTM at time t, and combines the two to obtain the output result h of bi-directional LSTM t 。
We generate facet-level feature representations on text using bi-directional LSTM, co-occurrence sequence of text words x '= { x' 1 ，x’ 2 ，......，x’ n As the input of the bidirectional LSTM, and sets the hidden state h at the time t in the bidirectional LSTM process t D, and finally the output of the bidirectional LSTM is the sequence h t ＝{h 1 ，h 2 ，......，h n Due to the self-character of LSTM, the state h of the last moment of the sequence n Basically contains the state information of all the words in the foregoing, maximally retains the aspect characteristics, so that this state is output as h n Aspect feature h as text aspect 。
After constructing the aspect features of the text, we use x = { x } in order to get semantic relations between words in different positions in the text 1 ，x 2 ，......，x n Using the two-way LSTM to construct a context semantic feature sequence h of the text for the word co-occurrence feature sequence of the text as an input t semantic The context semantic feature h t semantic The feature is different from the text word co-occurrence feature in that the semantic feature at any position of the text sequence comprises the semantic information of the context word, and the text word co-occurrence feature only reflects the semantic characteristics of the word.
4) Attention-driven layer: constructing an attention mechanism by using aspect features and context semantic feature sequences of the text to obtain an attention weight alpha t 。
Firstly h is firstly carried out semantic And h aspect Generating u through one layer of neural network learning t ，u t Representing the degree of association between context semantics and aspect features, wherein W and b are neural network bias, and a word-level context vector is introduced
Measure u t And normalizing to form probability distribution to finally obtain the attention weight alpha t 。
In the traditional attention mechanism network, the association degree between the text context semantic features and the target aspect categories is mainly directly calculated, and the difference of the aspect level-attention mechanism network lies in that the text aspect features are learned based on the text context and attention weights between the text context semantic features and the aspect features are constructed. Calculating the correlation between text context semantic features and aspect features facilitates extracting potential context words relevant to text aspect category prediction.
5) An output layer: and finally outputting a result y by the public opinion text aspect level feature vector.
Using attention weight alpha t And text context semantic feature sequence h semantic And constructing a text semantic output o with the attention weight of the aspect features. o is used as the input of the emotional tendency prediction function, and the emotional score is calculated by utilizing the softmax function, wherein W is less s And b s The weight matrix and offset of the softmax classification layer are sorted.
y＝softmax(W s ·o+b s )
1＞y i ＞0
And (3) outputting a calculation result y of the layer as an emotion score, for example, if the output result y is [0.31,0.69],0.31 represents a positive probability, and 0.69 represents a negative probability, so that the emotion analysis result is judged to be a negative emotion, taking a probability value y =0.69 as the emotion score, and if the score is higher, the emotion polarity is stronger. If the emotion is judged to be positive, a result without negative risk is directly output.
A risk value evaluation module: and evaluating the risk value of the client aiming at the transaction data output by the data access module, and outputting a stage risk value B aiming at each client. And (4) carrying out risk scoring according to the financing balance of the client in the bank and the financing balance of the group to which the client belongs, wherein the higher the financing balance of the client is, the greater the loss suffered by the bank after the client generates credit risk is, and therefore, the higher the risk value is. Alpha and beta are the customer financing balance and group customer financing balance weight.
α+β＝1
Risk early warning module: and scoring the comprehensive credit risk of the client according to the scores output by the semantic emotion analysis module and the risk value evaluation module, and amplifying the result by 100 times in the same proportion so as to be displayed conveniently, wherein the comprehensive score is R.
R＝100×y×B
And the public opinion early warning threshold is set, the public opinion early warning is hit when the threshold is exceeded, the system automatically pushes early warning information to the to-do category of a customer manager to prompt risks, and the original text of the public opinion information is checked in a hyperlink mode.
As exemplified below.
The data access module is used for preliminarily processing the public opinion data into the following form:
title: company A offers up to 4.5 billion guarantees to subsidiary B.
Source/author: finance and economics channel B
Release time: 2021, 6 months and 29 days
The text is as follows: 6.29A later, company A discloses that company full fund B needs to apply 8.9 million yuan of credit line Renminbi to Shenzhen division of a certain Bank stock Limited company due to project development, and the credit period is 5 years. Company A provides responsibility guarantee for the debt under the credit item according to the holding proportion, the guarantee principal amount does not exceed 4.539 billion RMB, and the guarantee period is three years from the effective date of the guarantee contract to the date when the term of each debt under the credit item expires.
By the date of the announcement, the balance of the external guarantee (excluding sales mortgage guarantee provided by subsidiaries to customers) of company a and the equity controlling subsidiary company B is 459 billion, accounting for 45.28% of the equity of capital east that was approved by the subsidiary company in the recent past.
The semantic emotion analysis module analyzes that a client corresponding to the public opinion information is company A; the risk label is "guarantee to the outside world"; the emotion score y =0.597 is calculated by an aspect level emotion analysis model based on an attention mechanism.
And the risk value evaluation module and the risk early warning module perform comprehensive credit risk calculation to obtain an R value (a fictitious numerical value). Supposing that the weight alpha and beta of the preset client financing balance and the group client financing balance are 0.7 and 0.3 respectively, the client financing balance is 11.27 million yuan, the total client financing balance of the department of the nation is 111027.33 million yuan, the group financing balance of the client is 1064.44 million yuan, and the total client financing balance of the group of the department of the nation is 55234.98 million yuan. If the public sentiment early warning threshold value is 0.2, 0.349 is larger than 0.2, and the system automatically triggers risk early warning prompt.
Fig. 15 schematically shows a block diagram of an electronic device adapted to implement the above method according to an embodiment of the present disclosure.
As shown in fig. 15, an electronic apparatus 900 according to an embodiment of the present disclosure includes a processor 901 which can perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM) 902 or a program loaded from a storage portion 908 into a Random Access Memory (RAM) 903. Processor 901 can include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor and/or related chipset(s) and/or a special purpose microprocessor (e.g., an Application Specific Integrated Circuit (ASIC)), and/or the like. The processor 901 may also include on-board memory for caching purposes. The processor 901 may comprise a single processing unit or a plurality of processing units for performing the different actions of the method flows according to embodiments of the present disclosure.
In the RAM 903, various programs and data necessary for the operation of the electronic apparatus 900 are stored. The processor 901, the ROM 902, and the RAM 903 are connected to each other through a bus 904. The processor 901 performs various operations of the method flows according to the embodiments of the present disclosure by executing programs in the ROM 902 and/or the RAM 903. Note that the programs may also be stored in one or more memories other than the ROM 902 and the RAM 903. The processor 901 may also perform various operations of the method flows according to embodiments of the present disclosure by executing programs stored in the one or more memories.
The present disclosure also provides a computer-readable storage medium, which may be contained in the apparatus/device/system described in the above embodiments; or may exist separately and not be assembled into the device/apparatus/system. The computer-readable storage medium carries one or more programs which, when executed, implement a method according to an embodiment of the disclosure.
According to embodiments of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, which may include, for example but is not limited to: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to embodiments of the present disclosure, a computer-readable storage medium may include the ROM 902 and/or RAM 903 described above and/or one or more memories other than the ROM 902 and RAM 903.
Embodiments of the present disclosure also include a computer program product comprising a computer program containing program code for performing the method illustrated in the flow chart. The program code is for causing a computer system to perform the methods of the embodiments of the disclosure when the computer program product is run on the computer system.
The computer program performs the above-described functions defined in the system/apparatus of the embodiments of the present disclosure when executed by the processor 901. The systems, apparatuses, modules, units, etc. described above may be implemented by computer program modules according to embodiments of the present disclosure.
In one embodiment, the computer program may be hosted on a tangible storage medium such as an optical storage device, a magnetic storage device, or the like. In another embodiment, the computer program may also be transmitted in the form of a signal over a network medium, distributed, and downloaded and installed via the communication section 909 and/or installed from the removable medium 911. The computer program containing program code may be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.
In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 909, and/or installed from the removable medium 911. The computer program, when executed by the processor 901, performs the above-described functions defined in the system of the embodiment of the present disclosure. The above described systems, devices, apparatuses, modules, units, etc. may be implemented by computer program modules according to embodiments of the present disclosure.
In accordance with embodiments of the present disclosure, program code for executing computer programs provided by embodiments of the present disclosure may be written in any combination of one or more programming languages, and in particular, these computer programs may be implemented using high level procedural and/or object oriented programming languages, and/or assembly/machine languages. The programming language includes, but is not limited to, programming languages such as Java, C + +, python, the "C" language, or the like. The program code may execute entirely on the user computing device, partly on the user device, partly on a remote computing device, or entirely on the remote computing device or server. In situations involving remote computing devices, the remote computing devices may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to external computing devices (e.g., through the internet using an internet service provider).
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
Those skilled in the art will appreciate that various combinations and/or combinations of features recited in the various embodiments and/or claims of the present disclosure can be made, even if such combinations or combinations are not expressly recited in the present disclosure. In particular, various combinations and/or combinations of the features recited in the various embodiments of the present disclosure and/or the claims may be made without departing from the spirit and teachings of the present disclosure. All such combinations and/or associations are within the scope of the present disclosure.
The embodiments of the present disclosure are described above. However, these examples are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments are described separately above, this does not mean that the measures in the embodiments cannot be used advantageously in combination. The scope of the disclosure is defined by the appended claims and equivalents thereof. Various alternatives and modifications can be devised by those skilled in the art without departing from the scope of the disclosure, and these alternatives and modifications are intended to fall within the scope of the disclosure.
Claims (15)
1. An enterprise risk early warning method is characterized by comprising the following steps:
acquiring public opinion data of an enterprise, wherein the public opinion data comprises text information;
determining a client name, a risk label and an emotion value of the public opinion data according to the text information;
acquiring transaction data of a corresponding enterprise according to the customer name;
calculating a corresponding enterprise risk value according to the emotion value and the transaction data; and
and returning an enterprise risk early warning result according to the enterprise risk value.
2. The method as claimed in claim 1, wherein the determining a customer name of the public opinion data according to the text information specifically comprises:
and matching the text information with a pre-constructed enterprise name library to obtain the client name of the public opinion data.
3. The method of claim 1, wherein determining the risk label of the public opinion data according to the text information specifically comprises:
and matching the text information with a pre-constructed risk label library to obtain the risk label of the public opinion data.
4. The method as claimed in claim 1, wherein the determining the emotion value of the public opinion data according to the text information specifically includes:
determining a keyword set in the text information;
determining a relevant word set of the keyword set according to the degree of association in the text information;
calculating emotion orientation probability according to the keyword set and the associated word set; and
and determining the emotion value according to the emotion orientation probability.
5. The method according to claim 4, wherein the determining the set of keywords in the text message specifically comprises:
extracting m keywords in the text information, wherein m is an integer greater than or equal to 1; and
and taking the m keywords as the keyword set.
6. The method according to claim 4, wherein the determining, in the text information, a related word set of the keyword set according to a degree of association specifically includes:
matching n similar words of each keyword in the keyword set in the text information, wherein n is an integer greater than or equal to 1;
calculating the association degree of the keyword and each similar word in the n similar words;
taking the similar word with the highest degree of association with the keyword in the n similar words as the associated word corresponding to the keyword; and
and taking the m associated words as the associated word set.
7. The method according to claim 6, wherein the calculating the degree of association between the keyword and each of the n similar words specifically comprises:
calculating the co-occurrence similarity of the keyword and each similar word in the n similar words according to the times of the keyword and each similar word in the n similar words in the text information and the times of the keyword in the text information;
calculating semantic similarity between the keyword and each similar word in the n similar words according to the word vector of the keyword and the word vector of the similar word; and
and multiplying the co-occurrence similarity and the semantic similarity to obtain the association degree of the keyword and each similar word in the n similar words.
8. The method according to claim 4, wherein the calculating of the emotion orientation probability according to the keyword set and the associated word set specifically comprises:
determining an aspect level feature vector of the text information according to the relevant word set;
determining a context semantic feature vector of the text information according to the keyword set;
calculating emotion weight according to the aspect level feature vector and the context semantic feature vector; and
and calculating the emotion orientation probability according to the emotion weight.
9. The method of claim 4, wherein the emotion orientation probabilities comprise a positive emotion probability value and a negative emotion probability value, and wherein determining the emotion value from the emotion orientation probability comprises:
comparing the positive emotion probability value with the negative emotion probability value; and
and taking the maximum emotion probability value as the emotion value.
10. The method according to claim 1, wherein calculating a corresponding business risk value based on the sentiment value and the transaction data specifically comprises:
calculating a transaction risk value according to the transaction data; and
and calculating an enterprise risk value according to the emotion value and the transaction risk value.
11. The method according to any one of claims 1 to 10, wherein returning an enterprise risk early warning result according to the enterprise risk value specifically comprises:
comparing the enterprise risk value with a preset risk threshold value; and
and when the enterprise risk value is greater than or equal to the risk threshold value, returning an enterprise risk early warning result.
12. An enterprise risk early warning device, characterized in that includes:
the first acquisition module is used for executing acquisition of public opinion data of an enterprise, and the public opinion data comprises text information;
the determining module is used for determining the client name, the risk label and the emotion value of the public opinion data according to the text information;
the second acquisition module is used for acquiring the transaction data of the corresponding enterprise according to the client name;
the calculation module is used for calculating a corresponding enterprise risk value according to the emotion value and the transaction data; and
and the returning module is used for returning the enterprise risk early warning result according to the enterprise risk value.
13. An electronic device, comprising:
one or more processors;
one or more memories for storing executable instructions that, when executed by the processor, implement the method of any one of claims 1-11.
14. A computer-readable storage medium, characterized in that the storage medium has stored thereon executable instructions which, when executed by a processor, implement the method according to any one of claims 1 to 11.
15. A computer program product comprising a computer program comprising one or more executable instructions which, when executed by a processor, implement the method of any one of claims 1 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210261606.5A CN115689717A (en) | 2022-03-16 | 2022-03-16 | Enterprise risk early warning method, device, electronic equipment, medium and program product |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210261606.5A CN115689717A (en) | 2022-03-16 | 2022-03-16 | Enterprise risk early warning method, device, electronic equipment, medium and program product |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115689717A true CN115689717A (en) | 2023-02-03 |
Family
ID=85060326
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210261606.5A Pending CN115689717A (en) | 2022-03-16 | 2022-03-16 | Enterprise risk early warning method, device, electronic equipment, medium and program product |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115689717A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116483822A (en) * | 2023-06-21 | 2023-07-25 | 建信金融科技有限责任公司 | Service data early warning method, device, computer equipment and storage medium |
| CN116596562A (en) * | 2023-07-18 | 2023-08-15 | 山东四季车网络科技有限公司 | Method and system for predicting false information risk of second-hand vehicle |
-
2022
- 2022-03-16 CN CN202210261606.5A patent/CN115689717A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116483822A (en) * | 2023-06-21 | 2023-07-25 | 建信金融科技有限责任公司 | Service data early warning method, device, computer equipment and storage medium |
| CN116483822B (en) * | 2023-06-21 | 2023-09-26 | 建信金融科技有限责任公司 | Service data early warning method, device, computer equipment and storage medium |
| CN116596562A (en) * | 2023-07-18 | 2023-08-15 | 山东四季车网络科技有限公司 | Method and system for predicting false information risk of second-hand vehicle |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Albalawi et al. | Using topic modeling methods for short-text data: A comparative analysis | |
| CN107908740B (en) | Information output method and device | |
| CN115689717A (en) | Enterprise risk early warning method, device, electronic equipment, medium and program product | |
| CN111783039B (en) | Risk determination method, risk determination device, computer system and storage medium | |
| CN111651552B (en) | Structured information determining method and device and electronic equipment | |
| CN112927082A (en) | Credit risk prediction method, apparatus, device, medium, and program product | |
| CN114119136A (en) | Product recommendation method and device, electronic equipment and medium | |
| CN116821372A (en) | Knowledge graph-based data processing method and device, electronic equipment and medium | |
| CN113392920B (en) | Method, apparatus, device, medium, and program product for generating cheating prediction model | |
| CN113220999A (en) | User feature generation method and device, electronic equipment and storage medium | |
| CN111126073B (en) | Semantic retrieval method and device | |
| Huang et al. | Application of informetrics on financial network text mining based on affective computing | |
| Duman | Social media analytical CRM: a case study in a bank | |
| Prakash et al. | Lexicon Based Sentiment Analysis (LBSA) to Improve the Accuracy of Acronyms, Emoticons, and Contextual Words | |
| CN113051396B (en) | Classification recognition method and device for documents and electronic equipment | |
| CN113095078A (en) | Associated asset determination method and device and electronic equipment | |
| CN111368036B (en) | Method and device for searching information | |
| CN114138976A (en) | Data processing and model training method and device, electronic equipment and storage medium | |
| CN114741501A (en) | Public opinion early warning method and device, readable storage medium and electronic equipment | |
| CN112133308A (en) | Method and device for multi-label classification of voice recognition text | |
| CN110599230A (en) | Method for constructing pricing model of second-hand vehicle, pricing method and device | |
| Xue et al. | The principle and implementation of sentiment analysis system | |
| KR102625347B1 (en) | A method for extracting food menu nouns using parts of speech such as verbs and adjectives, a method for updating a food dictionary using the same, and a system for the same | |
| CN114169316A (en) | Financial market income prediction model construction method and device and electronic equipment | |
| Derouiche et al. | Study of Tweets’ Sentiment Impact on Stock Prices during Class Actions: An Application to Sports Companies |
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 |