Enterprise risk judgment method and related equipment based on artificial intelligence

Abstract

The present application proposes an enterprise risk judgment method, device, electronic device and storage medium based on artificial intelligence. The enterprise risk judgment method based on artificial intelligence includes: querying the historical registration information of the enterprise from the business database, and obtaining the historical public information of the enterprise based on the open source information; comparing the historical registration information and historical public information of the enterprise to construct the risk feature vector of the enterprise; labeling the enterprise to obtain the label of each risk feature vector; calculating the preferred dimension factor based on the risk feature vector and the label, reducing the dimension of the risk feature vector to the preferred dimension factor to obtain the preferred risk feature vector; training the enterprise risk judgment network based on the preferred risk feature vector and the enterprise label; collecting the preferred risk feature vector of the enterprise to be judged and inputting it into the enterprise risk judgment network to obtain the risk judgment result. The present application can enhance the difference in feature vectors between enterprises of different risk types, thereby improving the accuracy of enterprise risk judgment.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an enterprise risk judgment method, device, electronic device and storage medium based on artificial intelligence. Background Art

[0002] Every year, companies make some changes to basic information, such as shareholder information, equity changes, administrative license qualifications, and corresponding administrative penalties. All these changes are recorded in the database of the industrial and commercial system. In theory, the annual report of the company or the immediate public disclosure document should contain all changes in basic information to ensure that the public and investors can obtain the company's real operating conditions in real time. However, some companies cannot make public announcements of changes in corporate information in real time, which affects the public and investors' judgment of the company's real operating conditions, thereby increasing the company's investment risks.

[0003] Usually, the open source information of an enterprise is obtained through crawler technology, and the keywords and the frequency of keywords in all open source information are obtained to obtain the risk judgment results of the enterprise. However, this method cannot screen a large amount of open source information, the calculation is large, and the accuracy of enterprise risk judgment is low. Summary of the invention

[0004] In view of the above content, it is necessary to propose an enterprise risk judgment method and related equipment based on artificial intelligence to solve the technical problem of how to improve the accuracy of enterprise risk judgment. The related equipment includes an enterprise risk judgment device based on artificial intelligence, electronic equipment and storage medium.

[0005] This application is based on an artificial intelligence-based enterprise risk assessment method, the method comprising:

[0006] Query the historical registration information of the enterprise from the business database according to the enterprise ID, and obtain the historical public information of the enterprise according to the open source information of the enterprise, wherein the historical registration information and the historical public information both include all basic information of the enterprise in each year during the historical period;

[0007] Comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises;

[0008] Labeling the enterprise to obtain a label of each risk feature vector;

[0009] Calculating a preferred dimension factor based on the risk feature vector and the label to obtain a preferred dimension reduction matrix, and performing dimension reduction processing on all risk feature vectors based on the preferred dimension reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector;

[0010] Training a classification network based on the preferred risk feature vector and the label of the preferred risk feature vector to obtain an enterprise risk judgment network;

[0011] The optimal risk feature vector of the enterprise to be judged is collected as the input of the enterprise risk judgment network to obtain the risk judgment result of the enterprise to be judged.

[0012] In some embodiments, querying the historical registration information of the enterprise from the business database based on the enterprise ID, and obtaining the historical public information of the enterprise based on the open source information of the enterprise, includes:

[0013] Storing the registration contents of basic information of all enterprises each year to build a business database, wherein the basic information includes all basic character segments related to the basic information;

[0014] Store the public disclosure of basic information of all enterprises each year to build an open source database;

[0015] Search the business database according to the enterprise ID to obtain the historical registration information of the enterprise, wherein the historical registration information includes the registration content of the basic information of the enterprise in each year during the historical period;

[0016] The open source database is searched according to the enterprise ID to obtain the historical public information of the enterprise, wherein the historical public information includes the public content of the basic information of the enterprise each year in the historical period.

[0017] In some embodiments, comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises includes:

[0018] Taking any character segment among all the basic character segments of the basic information as the target character segment;

[0019] Acquire the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise, wherein the target enterprise is any one of all enterprises;

[0020] Traversing all basic character segments of the basic information to obtain the attribute value of each basic character segment;

[0021] Arranging the attribute values ​​of all basic character segments in a fixed order to obtain the risk feature vector of the target enterprise;

[0022] Traverse the historical registration information and historical public information of all enterprises to obtain the risk feature vector of each enterprise.

[0023] In some embodiments, the acquiring the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise includes:

[0024] Search the historical registration information of the target enterprise based on the target character segment to obtain the registration content of the target character segment in each year;

[0025] Searching the historical public information of the target enterprise based on the target character segment to obtain the public information of the target character segment in each year;

[0026] Compare the registered content and the publicized content of the target character segment in the same year to obtain the comparison results for each year;

[0027] The comparison results of each year are counted to obtain the attribute value of the target character segment.

[0028] In some embodiments, the calculating of the preferred dimension factor based on the risk feature vector and the label to obtain a preferred dimension reduction matrix, and performing dimension reduction processing on all risk feature vectors based on the preferred dimension reduction matrix to obtain a preferred risk feature vector and the label of the preferred risk feature vector, includes:

[0029] Processing the risk feature vectors of all enterprises according to the PCA algorithm to obtain all principal component vectors and the eigenvalues ​​of the principal component vectors;

[0030] Arrange the principal component vectors from top to bottom along the column direction in the order of the eigenvalues ​​from large to small to obtain a principal component vector matrix;

[0031] Traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain the preferred dimension factor;

[0032] Selecting a certain number of principal component vectors from the principal component vector matrix from top to bottom to form a preferred dimension reduction matrix, wherein the certain number is the same as the preferred dimension factor;

[0033] The risk feature vector is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain a preferred risk feature vector, and the label of the risk feature vector is used as the label of the preferred risk feature vector.

[0034] In some embodiments, traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain the preferred dimension factor includes:

[0035] A1, selecting n principal component vectors from the principal component vector matrix from top to bottom to form an nth dimension reduction matrix;

[0036] A2, performing dimensionality reduction processing on the risk feature vector based on the nth dimensionality reduction matrix to obtain an n-dimensional risk feature vector;

[0037] A3, copying the label of the risk feature vector to the dimensioned n-dimensional risk feature vector to obtain a label of the n-dimensional risk feature vector, where the label is 0 or 1;

[0038] A4, calculating the complexity index of the n-dimensional risk feature vector according to the complexity index model;

[0039] A5, repeat steps A1 to A4 to obtain the complexity index of the risk feature vectors of all dimensions after dimensionality reduction, and take the dimension corresponding to the minimum value of the complexity index as the preferred dimensionality factor.

[0040] In some embodiments, the complexity index model satisfies the relationship:

[0041]

[0042] in, are the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; M is the number of all n-dimensional risk feature vectors; is the L2 distance between the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector; are the labels of the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; α n is the complexity index of the n-dimensional risk feature vector, is a user-defined function, which satisfies the relationship:

[0043]

[0044] The embodiment of the present application also provides an enterprise risk judgment device based on artificial intelligence, the device comprising:

[0045] An acquisition unit is used to query the historical registration information of the enterprise from the business database according to the enterprise ID, and to acquire the historical public information of the enterprise according to the open source information of the enterprise, wherein the historical registration information and the historical public information both include all basic information of the enterprise in each year in the historical period;

[0046] A comparison unit, used for comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises;

[0047] A labeling unit, used for labeling the enterprise to obtain a label of each risk feature vector;

[0048] A dimensionality reduction unit, configured to calculate a preferred dimensionality factor based on the risk feature vector and the label to obtain a preferred dimensionality reduction matrix, and perform dimensionality reduction processing on all risk feature vectors based on the preferred dimensionality reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector;

[0049] A training unit, used for training a classification network based on the preferred risk feature vector and a label of the preferred risk feature vector to obtain an enterprise risk judgment network;

[0050] The risk judgment unit collects the preferred risk feature vector of the enterprise to be judged as the input of the enterprise risk judgment network to obtain the risk judgment result of the enterprise to be judged.

[0051] The present application also provides an electronic device, the electronic device comprising:

[0052] A memory storing at least one instruction;

[0053] A processor executes instructions stored in the memory to implement the enterprise risk judgment method based on artificial intelligence.

[0054] An embodiment of the present application also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the enterprise risk judgment method based on artificial intelligence.

[0055] In summary, this application can collect the differences between the public information and the registered information of the enterprise to construct a feature vector, and reduce the feature vectors of all enterprises to the preferred dimensionality factor, thereby enhancing the difference in feature vectors between enterprises of different risk types. The feature vectors under the preferred dimensionality factor are used as input to train the classification network, which can improve the accuracy of enterprise risk judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of a preferred embodiment of the enterprise risk judgment method based on artificial intelligence involved in this application.

[0057] Figure 2 It is a schematic diagram of the structure of the historical registration information involved in this application.

[0058] Figure 3 It is a functional module diagram of a preferred embodiment of the enterprise risk judgment device based on artificial intelligence involved in this application.

[0059] Figure 4 It is a structural schematic diagram of an electronic device of a preferred embodiment of the enterprise risk judgment method based on artificial intelligence involved in this application. DETAILED DESCRIPTION

[0060] In order to more clearly understand the purpose, features and advantages of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. In the following description, many specific details are set forth to facilitate a full understanding of the present application, and the embodiments described are only a part of the embodiments of the present application, rather than all of the embodiments.

[0061] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used herein in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0063] The embodiment of the present application provides an enterprise risk judgment method based on artificial intelligence, which can be applied to one or more electronic devices. The electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0064] An electronic device can be any electronic product that can interact with a customer, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0065] The electronic device may also include a network device and / or a client device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud consisting of a large number of hosts or network servers based on cloud computing.

[0066] The network where the electronic device is located includes but is not limited to the Internet, wide area network, metropolitan area network, local area network, virtual private network (VPN), etc.

[0067] like Figure 1 As shown, it is a flow chart of a preferred embodiment of the enterprise risk judgment method based on artificial intelligence of the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted.

[0068] S10, querying the historical registration information of the enterprise from the business database according to the enterprise ID, and obtaining the historical public information of the enterprise according to the open source information of the enterprise, wherein the historical registration information and the historical public information both include all basic information of the enterprise every year in the historical period.

[0069] In an optional embodiment, querying the historical registration information of the enterprise from the business database according to the enterprise ID, and obtaining the historical public information of the enterprise according to the open source information of the enterprise, includes:

[0070] Storing the registration contents of basic information of all enterprises each year to build a business database, wherein the basic information includes all basic character segments related to the basic information;

[0071] Store the public disclosure of basic information of all enterprises each year to build an open source database;

[0072] Search the business database according to the enterprise ID to obtain the historical registration information of the enterprise, wherein the historical registration information includes the registration content of the basic information of the enterprise in each year during the historical period;

[0073] The open source database is searched according to the enterprise ID to obtain the historical public information of the enterprise, wherein the historical public information includes the public content of the basic information of the enterprise each year in the historical period.

[0074] In an optional embodiment, each enterprise will correspond to an exclusive ID information, and the ID information can be the unified social credit code of the enterprise. Every year, the basic information changes of the enterprise must be registered in the industrial and commercial system. The basic information of the enterprise includes the name of the enterprise shareholder, the type of administrative penalty, the share before the equity change, the share after the equity change, and other basic character segments related to the basic information of the enterprise. Each basic character corresponds to specific registration content every year. A data table is created by executing SQL statements, and the enterprise ID, all basic character segments, and the specific registration content of the basic character segments each year and the corresponding year are stored in the data table to form a business database.

[0075] In this optional embodiment, the historical registration information of the enterprise corresponding to the enterprise ID can be extracted by searching the enterprise ID in the business database through an intelligent search engine, and the historical registration information includes the registration content of all basic information of the enterprise in each year during the historical period, and the historical period is the past 3 years. Among them, the intelligent search is a new generation of search engine combined with artificial intelligence technology. In addition to providing traditional quick retrieval and relevance sorting functions, the engine should be noted that only users with specific access rights can obtain the historical registration information of the enterprise from the business database.

[0076] In this optional embodiment, every enterprise needs to publicize all basic information every year, and the public information of all enterprises is kept as an open source database. All users have the right to access the open source database. The type of information stored in the open source database is the same as the type of information in the business database. By searching the enterprise ID in the open source database through an intelligent search engine, the historical public information of the enterprise corresponding to the enterprise ID can be extracted. The historical public information includes the public content of all basic information of the enterprise each year within the historical time. In this optional embodiment, the historical time is the past 3 years.

[0077] In this optional embodiment, each enterprise corresponds to a historical registration information and a historical public information, and the historical registration information includes the registration content corresponding to all basic character segments of the enterprise in each of the past three years, and all the basic character segments constitute the basic information of the enterprise; the number of the basic character segments is recorded as N, and the structural diagram of the historical registration information of an enterprise ID is as follows: Figure 2 As shown; the historical public information contains the public information corresponding to all basic character segments of the enterprise in each of the past three years, and the structure of the historical public information is the same as the structure of the historical registration information.

[0078] In this way, the historical registration information and historical disclosure information of each enterprise are obtained, providing a data basis for the subsequent construction of enterprise risk feature vectors.

[0079] S11, comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises.

[0080] In an optional embodiment, comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises includes:

[0081] Taking any character segment among all the basic character segments of the basic information as the target character segment;

[0082] Acquire the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise, wherein the target enterprise is any one of all enterprises;

[0083] Traversing all basic character segments of the basic information to obtain the attribute value of each basic character segment;

[0084] Arranging the attribute values ​​of all basic character segments in a fixed order to obtain the risk feature vector of the target enterprise;

[0085] Traverse the historical registration information and historical public information of all enterprises to obtain the risk feature vector of each enterprise.

[0086] In this optional embodiment, the step of acquiring the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise includes:

[0087] Search the historical registration information of the target enterprise based on the target character segment to obtain the registration content of the target character segment in each year;

[0088] Searching the historical public information of the target enterprise based on the target character segment to obtain the public information of the target character segment in each year;

[0089] Compare the registered content and the publicized content of the target character segment in the same year to obtain the comparison results for each year;

[0090] The comparison results of each year are counted to obtain the attribute value of the target character segment.

[0091] In this optional embodiment, one basic character segment is randomly selected from all basic character segments as the target character segment, and after obtaining the historical registration information and historical public information corresponding to the target enterprise, "target character segment + year" is used as a keyword, and the keyword is searched for in the historical registration information through an intelligent search engine to obtain a first search result, and the first search result is the registration content of the target character segment corresponding to the year; at the same time, the keyword is searched for in the historical public information through an intelligent search engine to obtain a second search result, and the second search result is the public content of the target character segment corresponding to the year; the first search result and the second search result are used as a comparison group of the target character segment.

[0092] In this optional embodiment, all years in the historical registration information are traversed to update the keywords, and all comparison groups of the target character segment are obtained, wherein the comparison groups correspond to the years one by one; it is determined whether the contents of the two searches in each comparison group are the same, if they are the same, it means that the registration information and the public information of the target character segment are the same, and the comparison result of the comparison group is recorded as 0; if they are different, it means that the registration information and the public information of the target character segment are different, and the comparison result of the comparison group is recorded as 1; the sum of the comparison results of all comparison groups of the target character segment is calculated as the attribute value of the target character segment.

[0093] In this optional embodiment, all basic character segments are traversed, and the attribute value of each basic character segment is obtained according to the calculation method of the attribute value of the target character segment. All attribute values ​​are arranged into a vector of N rows and 1 column in a fixed order. The vector of N rows and 1 column is the risk feature vector of the target enterprise, wherein N is the number of the basic character segments. The fixed order is to ensure that the basic character segments represented by the same row in the risk feature vectors of different enterprises are the same.

[0094] In this optional embodiment, the risk feature vectors of all enterprises are obtained according to the above-mentioned method for obtaining the risk feature vector of the target enterprise.

[0095] In this way, the historical registration information and the historical public information of the enterprise can be used to obtain the risk feature vector corresponding to the enterprise. The risk feature vector can well reflect the characteristics of the enterprise and realize the vectorization of the characteristics.

[0096] S12, labeling the enterprise to obtain a label of each risk feature vector.

[0097] In an optional embodiment, the label of the risk feature vector is obtained according to the traditional comparison method. If the enterprise has multiple information that are not disclosed in accordance with regulations, and after manual review, it is believed that the enterprise is at risk, then the label of the risk feature vector of the enterprise is 1; if the enterprise basically discloses information in accordance with the actual situation, and after manual review, it is believed that the enterprise is not at risk, then the label of the risk feature vector of the enterprise is 0.

[0098] In this way, labels of all risk feature vectors are obtained through manual review for subsequent training of the enterprise risk judgment network.

[0099] S13, calculating a preferred dimension factor based on the risk feature vector and the label to obtain a preferred dimension reduction matrix, and performing dimension reduction processing on all risk feature vectors based on the preferred dimension reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector.

[0100] In an optional embodiment, the calculating the preferred dimension factor based on the risk feature vector and the label to obtain a preferred dimension reduction matrix, and performing dimension reduction processing on all risk feature vectors based on the preferred dimension reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector, includes:

[0101] Processing the matrix composed of risk characteristic vectors of all enterprises according to the PCA algorithm to obtain all principal component vectors and the characteristic values ​​of the principal component vectors;

[0102] Arrange the principal component vectors from top to bottom along the column direction in the order of the eigenvalues ​​from large to small to obtain a principal component vector matrix;

[0103] Traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain the preferred dimension factor;

[0104] Selecting a certain number of principal component vectors from the principal component vector matrix from top to bottom to form a preferred dimension reduction matrix, wherein the certain number is the same as the preferred dimension factor;

[0105] The risk feature vector is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain a preferred risk feature vector, and the label of the risk feature vector is used as the label of the preferred risk feature vector.

[0106] Among them, PCA (Principal Component Analysis) is the principal component analysis method and is the most widely used data dimensionality reduction algorithm.

[0107] In an optional embodiment, since the enterprise's risk feature vector contains information on all basic character segments, but information on some basic character segments does not participate in the process of enterprise risk judgment, the risk feature vector contains redundant information, which will affect the accuracy of the enterprise's risk judgment.

[0108] In this optional embodiment, in order to remove redundant information in the risk feature vector, it is necessary to perform dimensionality reduction processing on the risk feature vector to obtain a preferred risk feature vector, and at the same time assign the label of the risk feature vector to the preferred risk feature vector after dimensionality reduction, and the dimension of the preferred risk feature vector is the preferred dimensional factor. The preferred risk feature vector can meet two conditions: the first condition is that the similarity between the preferred risk feature vectors with the same label is high, and the second condition is that the similarity between the preferred risk feature vectors with different labels is low, thereby enhancing the difference between the preferred risk feature vectors with different labels and improving the accuracy of enterprise risk judgment.

[0109] In this optional embodiment, the number of risk feature vectors is recorded as M, and each risk feature vector is a vector with N rows and 1 column, so all risk feature vectors can form a data matrix with N rows and M columns, wherein the risk feature vectors correspond to enterprises one by one. The PCA algorithm is used to obtain all principal component vectors of the data matrix, wherein the principal component direction is the direction vector of the main feature distribution of the data, and the step of obtaining all principal component vectors of the data matrix by PCA includes:

[0110] Step 1: In a data matrix with N rows and M columns, calculate the mean of the M values ​​in each row, and subtract the mean of each row from the M values ​​in the row to obtain a centralized data matrix X, where the centralized data matrix X also has N rows and M columns;

[0111] Step 2: Calculate the covariance matrix C. The calculation formula of the covariance matrix C is as follows:

[0112]

[0113] Where X is the centralized data matrix, X T is the transposed matrix of the centralized data matrix, and the covariance matrix C is a square matrix with N rows and N columns;

[0114] Step 3: Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C, where the eigenvectors are the principal component vectors. A total of N principal component vectors can be obtained, and the principal component vectors are vectors of 1 row and N columns, wherein the eigenvalues ​​correspond one-to-one to the principal component vectors.

[0115] In this optional embodiment, a total of N principal component vectors are obtained, and each principal component vector corresponds to an eigenvalue. According to the idea of ​​the PCA algorithm, the larger the eigenvalue, the more data features are retained on the principal component vector corresponding to the eigenvalue. The principal component vectors are arranged from top to bottom along the column direction in descending order of the eigenvalue to obtain a principal component vector matrix, and the principal component vectors in the principal component vector matrix are further traversed according to the complexity index model to obtain the preferred dimension factor.

[0116] In this optional embodiment, traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain the preferred dimension factor includes:

[0117] A1, selecting the first n rows of principal component vectors from the principal component vector matrix to form an nth dimension reduction matrix;

[0118] A2, based on the nth dimensionality reduction matrix, the risk feature vector is subjected to dimensionality reduction processing to obtain an n-dimensional risk feature vector, and the calculation process of the dimensionality reduction processing is as follows:

[0119] T n =P n T N

[0120] Among them, P n is the nth dimension reduction matrix; T N is the data matrix of N rows and M columns, including M risk feature vectors, and the dimension of the risk feature vector is N; T n is a data matrix of n rows and M columns, including M n-dimensional risk feature vectors, and satisfies N≥n;

[0121] A3, copying the label of the risk feature vector to the n-dimensional risk feature vector after dimensionality reduction, to obtain a label of the n-dimensional risk feature vector, wherein the label is 0 or 1;

[0122] A4, calculating the complexity index of the n-dimensional risk feature vector according to a complexity index model, wherein the complexity index model satisfies the relationship:

[0123]

[0124] in, are the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; M is the number of all n-dimensional risk feature vectors; is the L2 distance between the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector; are the labels of the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; α nis the complexity index of the n-dimensional risk feature vector. The smaller the value of the complexity index is, the greater the distance between the n-dimensional risk feature vectors of different labels is. The smaller the distance between the n-dimensional risk feature vectors of the same label is, the more accurate the enterprise risk judgment result is. is a user-defined function, which satisfies the relationship:

[0125]

[0126] A5. Repeat steps A1 to A4 to obtain the complexity index of the risk feature vector of all dimensions after dimensionality reduction. The value range of all dimensions includes 1 dimension, N dimensions, and all dimensions between 1 dimension and N dimension. The dimension corresponding to the minimum value of the complexity index is taken as the preferred dimensionality factor.

[0127] In this optional embodiment, after obtaining the preferred dimensionality factor, a certain number of principal component vectors are selected from the principal component vector matrix from top to bottom to form a preferred dimensionality reduction matrix, and the certain number is the same as the preferred dimensionality factor. According to the method of step A2 and step A3, the risk feature vector is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain the preferred risk feature vector corresponding to the preferred dimensionality factor, and the label of the preferred risk feature vector is obtained at the same time.

[0128] In this way, redundant information in the risk feature vector can be removed to obtain an optimal risk feature vector, and the optimal risk feature vector can improve the accuracy of the enterprise risk judgment result.

[0129] S14, training a classification network based on the preferred risk feature vector and the label of the preferred risk feature vector to obtain an enterprise risk judgment network.

[0130] In an optional embodiment, the preferred risk feature vector and the corresponding label are used as a training data set to train a classification network. The classification network may adopt a ResNet network. Meanwhile, the classification network is a binary classification network.

[0131] In this optional embodiment, the loss function adopts a cross entropy function, and the calculation formula of the cross entropy function is as follows:

[0132]

[0133] Among them, N * is the number of training batches, that is, each input N during the training process * Calculate the loss function once for each optimal risk feature vector; is the label of the kth optimal risk feature vector, y kis the output result of the kth preferred risk feature vector, and the value range of the output result of the kth preferred risk feature vector is [0,1]. The closer to 1, the greater the risk of the kth preferred risk feature vector. The gradient descent method is continuously iterated to update the parameters in the classification network. When the value of the loss function no longer decreases, the training is stopped to obtain the enterprise risk judgment network.

[0134] In this way, the enterprise risk judgment network can be obtained, and the enterprise risk judgment results can be directly obtained based on the enterprise risk judgment network.

[0135] S15, collecting the optimal risk feature vector of the enterprise to be judged as the input of the enterprise risk judgment network, and obtaining the risk judgment result of the enterprise to be judged.

[0136] In an optional embodiment, historical registration information and historical public information of the enterprise to be judged are collected to construct a risk feature vector to be judged, and the risk feature vector to be judged is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain a preferred dimensionality reduction vector to be judged. The processing process of the dimensionality reduction processing is as follows:

[0137] t * =P * T *

[0138] Among them, t * is the optimal feature vector to be judged, T * is the risk feature vector to be judged, P * is the preferred dimension reduction matrix.

[0139] In this optional embodiment, the risk judgment result can be directly obtained by inputting the preferred feature vector to be judged into the enterprise risk judgment network. The value range of the risk judgment result is [0,1]. The closer the value is to 1, the greater the enterprise risk.

[0140] In this way, the enterprise risk judgment results can be directly obtained through the registration information and public information of the enterprises in the database with the help of the enterprise risk judgment network.

[0141] In summary, this application can collect the differences between the public information and the registered information of the enterprise to construct a feature vector, and reduce the feature vectors of all enterprises to the preferred dimensionality factor, thereby enhancing the difference in feature vectors between enterprises of different risk types. The feature vectors under the preferred dimensionality factor are used as input to train the classification network, which can improve the accuracy of enterprise risk judgment.

[0142] See also Figure 3 , Figure 31 is a functional module diagram of a preferred embodiment of the enterprise risk judgment device based on artificial intelligence of the present application. The enterprise risk judgment device based on artificial intelligence 11 includes an acquisition unit 110, a comparison unit 111, a labeling unit 112, a dimension reduction unit 113, a training unit 114, and a risk judgment unit 115. The module / unit referred to in the present application refers to a series of computer-readable instruction segments that can be executed by the processor 13 and can complete fixed functions, which are stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0143] In an optional embodiment, the acquisition unit 110 is used to query the historical registration information of the enterprise from the business database based on the enterprise ID, and to obtain the historical public information of the enterprise based on the open source information of the enterprise. The historical registration information and the historical public information both include all basic information of the enterprise every year in the historical period.

[0144] In an optional embodiment, querying the historical registration information of the enterprise from the business database according to the enterprise ID, and obtaining the historical public information of the enterprise according to the open source information of the enterprise, includes:

[0145] Storing the registration contents of basic information of all enterprises each year to build a business database, wherein the basic information includes all basic character segments related to the basic information;

[0146] Store the public disclosure of basic information of all enterprises each year to build an open source database;

[0147] Search the business database according to the enterprise ID to obtain the historical registration information of the enterprise, wherein the historical registration information includes the registration content of the basic information of the enterprise in each year during the historical period;

[0148] The open source database is searched according to the enterprise ID to obtain the historical public information of the enterprise, wherein the historical public information includes the public content of the basic information of the enterprise each year in the historical period.

[0149] In an optional embodiment, each enterprise will correspond to an exclusive ID information, and the ID information can be the unified social credit code of the enterprise. Every year, the basic information changes of the enterprise must be registered in the industrial and commercial system. The basic information of the enterprise includes the name of the enterprise shareholder, the type of administrative penalty, the share before the equity change, the share after the equity change, and other basic character segments related to the basic information of the enterprise. Each basic character corresponds to specific registration content every year. A data table is created by executing SQL statements, and the enterprise ID, all basic character segments, and the specific registration content of the basic character segments each year and the corresponding year are stored in the data table to form a business database.

[0150] In this optional embodiment, the historical registration information of the enterprise corresponding to the enterprise ID can be extracted by searching the enterprise ID in the business database through an intelligent search engine, and the historical registration information includes the registration content of all basic information of the enterprise in each year during the historical period, and the historical period is the past 3 years. Among them, the intelligent search is a new generation of search engine combined with artificial intelligence technology. In addition to providing traditional quick retrieval and relevance sorting functions, the engine should be noted that only users with specific access rights can obtain the historical registration information of the enterprise from the business database.

[0151] In this optional embodiment, every enterprise needs to publicize all basic information every year, and the public information of all enterprises is kept as an open source database. All users have the right to access the open source database. The type of information stored in the open source database is the same as the type of information in the business database. By searching the enterprise ID in the open source database through an intelligent search engine, the historical public information of the enterprise corresponding to the enterprise ID can be extracted. The historical public information includes the public content of all basic information of the enterprise each year within the historical time. In this optional embodiment, the historical time is the past 3 years.

[0152] In this optional embodiment, each enterprise corresponds to a historical registration information and a historical public information, and the historical registration information includes the registration content corresponding to all basic character segments of the enterprise in each of the past three years, and all the basic character segments constitute the basic information of the enterprise; the number of the basic character segments is recorded as N, and the structural diagram of the historical registration information of an enterprise ID is as follows: Figure 2 As shown; the historical public information contains the public information corresponding to all basic character segments of the enterprise in each of the past three years, and the structure of the historical public information is the same as the structure of the historical registration information.

[0153] In an optional embodiment, the comparison unit 111 is used to compare the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises.

[0154] In an optional embodiment, comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises includes:

[0155] Taking any character segment among all the basic character segments of the basic information as the target character segment;

[0156] Acquire the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise, wherein the target enterprise is any one of all enterprises;

[0157] Traversing all basic character segments of the basic information to obtain the attribute value of each basic character segment;

[0158] Arranging the attribute values ​​of all basic character segments in a fixed order to obtain the risk feature vector of the target enterprise;

[0159] Traverse the historical registration information and historical public information of all enterprises to obtain the risk feature vector of each enterprise.

[0160] In this optional embodiment, the step of acquiring the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise includes:

[0161] Search the historical registration information of the target enterprise based on the target character segment to obtain the registration content of the target character segment in each year;

[0162] Searching the historical public information of the target enterprise based on the target character segment to obtain the public information of the target character segment in each year;

[0163] Compare the registered content and the publicized content of the target character segment in the same year to obtain the comparison results for each year;

[0164] The comparison results of each year are counted to obtain the attribute value of the target character segment.

[0165] In this optional embodiment, one basic character segment is randomly selected from all basic character segments as the target character segment, and after obtaining the historical registration information and historical public information corresponding to the target enterprise, "target character segment + year" is used as a keyword, and the keyword is searched for in the historical registration information through an intelligent search engine to obtain a first search result, and the first search result is the registration content of the target character segment corresponding to the year; at the same time, the keyword is searched for in the historical public information through an intelligent search engine to obtain a second search result, and the second search result is the public content of the target character segment corresponding to the year; the first search result and the second search result are used as a comparison group of the target character segment.

[0166] In this optional embodiment, all years in the historical registration information are traversed to update the keywords, and all comparison groups of the target character segment are obtained, wherein the comparison groups correspond to the years one by one; it is determined whether the contents of the two searches in each comparison group are the same, if they are the same, it means that the registration information and the public information of the target character segment are the same, and the comparison result of the comparison group is recorded as 0; if they are different, it means that the registration information and the public information of the target character segment are different, and the comparison result of the comparison group is recorded as 1; the sum of the comparison results of all comparison groups of the target character segment is calculated as the attribute value of the target character segment.

[0167] In this optional embodiment, all basic character segments are traversed, and the attribute value of each basic character segment is obtained according to the calculation method of the attribute value of the target character segment. All attribute values ​​are arranged into a vector of N rows and 1 column in a fixed order. The vector of N rows and 1 column is the risk feature vector of the target enterprise, wherein N is the number of the basic character segments. The fixed order is to ensure that the basic character segments represented by the same row in the risk feature vectors of different enterprises are the same.

[0168] In this optional embodiment, the risk feature vectors of all enterprises are obtained according to the above-mentioned method for obtaining the risk feature vector of the target enterprise.

[0169] In an optional embodiment, the labeling unit 112 is used to label the enterprise to obtain a label of each risk feature vector.

[0170] In an optional embodiment, the label of the risk feature vector is obtained according to the traditional comparison method. If the enterprise has multiple information that are not disclosed in accordance with regulations, and after manual review, it is believed that the enterprise is at risk, then the label of the risk feature vector of the enterprise is 1; if the enterprise basically discloses information in accordance with the actual situation, and after manual review, it is believed that the enterprise is not at risk, then the label of the risk feature vector of the enterprise is 0.

[0171] In an optional embodiment, the dimensionality reduction unit 113 is used to calculate a preferred dimensionality factor based on the risk feature vector and the label to obtain a preferred dimensionality reduction matrix, and perform dimensionality reduction processing on all risk feature vectors based on the preferred dimensionality reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector.

[0172] In an optional embodiment, the calculating the preferred dimension factor based on the risk feature vector and the label to obtain a preferred dimension reduction matrix, and performing dimension reduction processing on all risk feature vectors based on the preferred dimension reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector, includes:

[0173] Processing the matrix composed of risk characteristic vectors of all enterprises according to the PCA algorithm to obtain all principal component vectors and the characteristic values ​​of the principal component vectors;

[0174] Arrange the principal component vectors from top to bottom along the column direction in the order of the eigenvalues ​​from large to small to obtain a principal component vector matrix;

[0175] Traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain the preferred dimension factor;

[0176] Selecting a certain number of principal component vectors from the principal component vector matrix from top to bottom to form a preferred dimension reduction matrix, wherein the certain number is the same as the preferred dimension factor;

[0177] The risk feature vector is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain a preferred risk feature vector, and the label of the risk feature vector is used as the label of the preferred risk feature vector.

[0178] Among them, PCA (Principal Component Analysis) is the principal component analysis method and is the most widely used data dimensionality reduction algorithm.

[0179] In an optional embodiment, since the enterprise's risk feature vector contains information on all basic character segments, but information on some basic character segments does not participate in the process of enterprise risk judgment, the risk feature vector contains redundant information, which will affect the accuracy of the enterprise's risk judgment.

[0180] In this optional embodiment, in order to remove redundant information in the risk feature vector, it is necessary to perform dimensionality reduction processing on the risk feature vector to obtain a preferred risk feature vector, and at the same time assign the label of the risk feature vector to the preferred risk feature vector after dimensionality reduction, and the dimension of the preferred risk feature vector is the preferred dimensional factor. The preferred risk feature vector can meet two conditions: the first condition is that the similarity between the preferred risk feature vectors with the same label is high, and the second condition is that the similarity between the preferred risk feature vectors with different labels is low, thereby enhancing the difference between the preferred risk feature vectors with different labels and improving the accuracy of enterprise risk judgment.

[0181] In this optional embodiment, the number of risk feature vectors is recorded as M, and each risk feature vector is a vector with N rows and 1 column, so all risk feature vectors can form a data matrix with N rows and M columns, wherein the risk feature vectors correspond to enterprises one by one. The PCA algorithm is used to obtain all principal component vectors of the data matrix, wherein the principal component direction is the direction vector of the main feature distribution of the data, and the step of obtaining all principal component vectors of the data matrix by PCA includes:

[0182] Step 1: In a data matrix with N rows and M columns, calculate the mean of the M values ​​in each row, and subtract the mean of each row from the M values ​​in the row to obtain a centralized data matrix X, where the centralized data matrix X also has N rows and M columns;

[0183] Step 2: Calculate the covariance matrix C. The calculation formula of the covariance matrix C is as follows:

[0184]

[0185] Where X is the centralized data matrix, X T is the transposed matrix of the centralized data matrix, and the covariance matrix C is a square matrix with N rows and N columns;

[0186] Step 3: Calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C, where the eigenvectors are the principal component vectors. A total of N principal component vectors can be obtained, and the principal component vectors are vectors of 1 row and N columns, wherein the eigenvalues ​​correspond one-to-one to the principal component vectors.

[0187] In this optional embodiment, a total of N principal component vectors are obtained, and each principal component vector corresponds to an eigenvalue. According to the idea of ​​the PCA algorithm, the larger the eigenvalue, the more data features are retained on the principal component vector corresponding to the eigenvalue. The principal component vectors are arranged from top to bottom along the column direction in descending order of the eigenvalue to obtain a principal component vector matrix, and the principal component vectors in the principal component vector matrix are further traversed according to the complexity index model to obtain the preferred dimension factor.

[0188] In this optional embodiment, traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain the preferred dimension factor includes:

[0189] A1, selecting the first n rows of principal component vectors from the principal component vector matrix to form an nth dimension reduction matrix;

[0190] A2, based on the nth dimensionality reduction matrix, the risk feature vector is subjected to dimensionality reduction processing to obtain an n-dimensional risk feature vector, and the calculation process of the dimensionality reduction processing is as follows:

[0191] T n =P n T N

[0192] Among them, P n is the nth dimension reduction matrix; T N is the data matrix of N rows and M columns, including M risk feature vectors, and the dimension of the risk feature vector is N; T n is a data matrix of n rows and M columns, including M n-dimensional risk feature vectors, and satisfies N≥n;

[0193] A3, copying the label of the risk feature vector to the n-dimensional risk feature vector after dimensionality reduction, to obtain a label of the n-dimensional risk feature vector, wherein the label is 0 or 1;

[0194] A4, calculating the complexity index of the n-dimensional risk feature vector according to a complexity index model, wherein the complexity index model satisfies the relationship:

[0195]

[0196] in, are the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; M is the number of all n-dimensional risk feature vectors; is the L2 distance between the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector; are the labels of the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; α n is the complexity index of the n-dimensional risk feature vector. The smaller the value of the complexity index is, the greater the distance between the n-dimensional risk feature vectors of different labels is. The smaller the distance between the n-dimensional risk feature vectors of the same label is, the more accurate the enterprise risk judgment result is. is a user-defined function, which satisfies the relationship:

[0197]

[0198] A5. Repeat steps A1 to A4 to obtain the complexity index of the risk feature vector of all dimensions after dimensionality reduction. The value range of all dimensions includes 1 dimension, N dimensions, and all dimensions between 1 dimension and N dimension. The dimension corresponding to the minimum value of the complexity index is taken as the preferred dimensionality factor.

[0199] In this optional embodiment, after obtaining the preferred dimensionality factor, a certain number of principal component vectors are selected from the principal component vector matrix from top to bottom to form a preferred dimensionality reduction matrix, and the certain number is the same as the preferred dimensionality factor. According to the method of step A2 and step A3, the risk feature vector is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain the preferred risk feature vector corresponding to the preferred dimensionality factor, and the label of the preferred risk feature vector is obtained at the same time.

[0200] In an optional embodiment, the training unit 114 is used to train a classification network based on the preferred risk feature vector and the label of the preferred risk feature vector to obtain an enterprise risk judgment network.

[0201] In an optional embodiment, the preferred risk feature vector and the corresponding label are used as a training data set to train a classification network. The classification network may adopt a fully connected network. Meanwhile, the classification network is a binary classification network.

[0202] In this optional embodiment, the loss function adopts a cross entropy function, and the calculation formula of the cross entropy function is as follows:

[0203]

[0204] Among them, N * is the number of training batches, that is, each input N during the training process* Calculate the loss function once for each optimal risk feature vector; is the label of the kth optimal risk feature vector, y k is the output result of the kth preferred risk feature vector, and the value range of the output result of the kth preferred risk feature vector is [0,1]. The closer to 1, the greater the risk of the kth preferred risk feature vector. The gradient descent method is continuously iterated to update the parameters in the classification network. When the value of the loss function no longer decreases, the training is stopped to obtain the enterprise risk judgment network.

[0205] In an optional embodiment, the risk judgment unit 115 collects the preferred risk feature vector of the enterprise to be judged as the input of the enterprise risk judgment network to obtain the risk judgment result of the enterprise to be judged.

[0206] In an optional embodiment, historical registration information and historical public information of the enterprise to be judged are collected to construct a risk feature vector to be judged, and the risk feature vector to be judged is subjected to dimensionality reduction processing based on the preferred dimensionality reduction matrix to obtain a preferred dimensionality reduction vector to be judged. The processing process of the dimensionality reduction processing is as follows:

[0207] t * =P * T *

[0208] Among them, t * is the optimal feature vector to be judged, T * is the risk feature vector to be judged, P * is the preferred dimension reduction matrix.

[0209] In this optional embodiment, the risk judgment result can be directly obtained by inputting the preferred feature vector to be judged into the enterprise risk judgment network. The value range of the risk judgment result is [0,1]. The closer the value is to 1, the greater the enterprise risk.

[0210] It can be seen from the above technical solutions that the present application can collect the differences between the public information and the registered information of an enterprise to construct a feature vector, and reduce the feature vectors of all enterprises to the preferred dimensionality factor, thereby enhancing the difference in feature vectors between enterprises of different risk types. The feature vectors under the preferred dimensionality factor are used as input to train the classification network, which can improve the accuracy of enterprise risk judgment.

[0211] See also Figure 4, is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device 1 includes a memory 12 and a processor 13. The memory 12 is used to store computer-readable instructions, and the processor 13 is used to execute the computer-readable instructions stored in the memory to implement the enterprise risk judgment method based on artificial intelligence described in any of the above embodiments.

[0212] In an optional embodiment, the electronic device 1 further includes a bus, a computer program stored in the memory 12 and executable on the processor 13, such as an enterprise risk determination program based on artificial intelligence.

[0213] Figure 4 Only the electronic device 1 having the memory 12 and the processor 13 is shown, and those skilled in the art can understand that Figure 4 The structure shown does not constitute a limitation on the electronic device 1 , and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0214] Combination Figure 1 The memory 12 in the electronic device 1 stores a plurality of computer-readable instructions to implement an enterprise risk determination method based on artificial intelligence, and the processor 13 can execute the plurality of instructions to implement:

[0215] Query the historical registration information of the enterprise from the business database according to the enterprise ID, and obtain the historical public information of the enterprise according to the open source information of the enterprise, wherein the historical registration information and the historical public information both include all basic information of the enterprise in each year during the historical period;

[0216] Comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises;

[0217] Labeling the enterprise to obtain a label of each risk feature vector;

[0218] Calculating a preferred dimension factor based on the risk feature vector and the label to obtain a preferred dimension reduction matrix, and performing dimension reduction processing on all risk feature vectors based on the preferred dimension reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector;

[0219] Training a classification network based on the preferred risk feature vector and the label of the preferred risk feature vector to obtain an enterprise risk judgment network;

[0220] The optimal risk feature vector of the enterprise to be judged is collected as the input of the enterprise risk judgment network to obtain the risk judgment result of the enterprise to be judged.

[0221] Specifically, the specific implementation method of the processor 13 for the above instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0222] Those skilled in the art will appreciate that the schematic diagram is merely an example of the electronic device 1 and does not constitute a limitation on the electronic device 1. The electronic device 1 may be a bus structure or a star structure. The electronic device 1 may also include more or less other hardware or software than shown in the diagram, or a different arrangement of components. For example, the electronic device 1 may also include input and output devices, network access devices, etc.

[0223] It should be noted that the electronic device 1 is only an example, and other existing or future electronic products that are suitable for the present application should also be included in the protection scope of the present application and included here by reference.

[0224] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium can be non-volatile or volatile. The readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 12 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. The memory 12 can not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the enterprise risk judgment program based on artificial intelligence, but also can be used to temporarily store data that has been output or is to be output.

[0225] In some embodiments, the processor 13 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 1, and uses various interfaces and lines to connect the various components of the entire electronic device 1, and executes or executes programs or modules stored in the memory 12 (for example, executing an enterprise risk judgment program based on artificial intelligence, etc.), and calls the data stored in the memory 12 to execute various functions of the electronic device 1 and process data.

[0226] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-mentioned embodiments of the enterprise risk judgment method based on artificial intelligence, for example Figure 1 Steps shown.

[0227] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules / units may be a series of computer-readable instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into an acquisition unit 110, a comparison unit 111, a labeling unit 112, a dimensionality reduction unit 113, a training unit 114, and a risk judgment unit 115.

[0228] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to execute a part of the enterprise risk judgment method based on artificial intelligence described in each embodiment of the present application.

[0229] If the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, and can also instruct the relevant hardware devices to complete through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented.

[0230] The computer program includes computer program code, which may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory and other memory, etc.

[0231] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0232] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0233] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. Figure 4 Only one arrow is used in the figure, but it does not mean that there is only one bus or one type of bus. The bus is configured to realize the connection and communication between the memory 12 and at least one processor 13, etc.

[0234] An embodiment of the present application also provides a computer-readable storage medium (not shown in the figure), in which computer-readable storage medium is stored computer-readable instructions, and the computer-readable instructions are executed by a processor in an electronic device to implement the enterprise risk judgment method based on artificial intelligence described in any of the above embodiments.

[0235] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0236] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0237] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0238] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the specification can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any specific order.

[0239] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application.

Claims

1. An enterprise risk judgment method based on artificial intelligence, characterized in that: The method comprises: Query the historical registration information of the enterprise from the business database according to the enterprise ID, and obtain the historical public information of the enterprise according to the open source information of the enterprise, wherein the historical registration information and the historical public information both include all basic information of the enterprise in each year during the historical period; Comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises; Labeling the enterprise to obtain a label of each risk feature vector; Based on the risk feature vector and the label, a preferred dimension factor is calculated to obtain a preferred dimension reduction matrix, and based on the preferred dimension reduction matrix, all risk feature vectors are subjected to dimension reduction processing to obtain a preferred risk feature vector and the label of the preferred risk feature vector, including: according to the PCA algorithm, a matrix composed of the risk feature vectors of all enterprises is processed to obtain all principal component vectors and the eigenvalues ​​of the principal component vectors; the principal component vectors are arranged from top to bottom along the column direction in the order of the eigenvalues ​​from large to small to obtain a principal component vector matrix; according to the complexity index model, the principal component vectors in the principal component vector matrix are traversed to obtain a preferred dimension factor; a certain number of principal component vectors are selected from the principal component vector matrix from top to bottom to form a preferred dimension reduction matrix, and the certain number is the same as the preferred dimension factor; based on the preferred dimension reduction matrix, the risk feature vector is subjected to dimension reduction processing to obtain Take a preferred risk feature vector, and use the label of the risk feature vector as the label of the preferred risk feature vector; wherein, traversing the principal component vectors in the principal component vector matrix according to the complexity index model to obtain a preferred dimension factor includes: A1, selecting n principal component vectors from the principal component vector matrix from top to bottom to form an n-th dimension reduction matrix; A2, performing dimension reduction processing on the risk feature vector based on the n-th dimension reduction matrix to obtain an n-dimensional risk feature vector; A3, copying the label of the risk feature vector to the n-dimensional risk feature vector after dimension reduction, and obtaining a label of the n-dimensional risk feature vector, wherein the label is 0 or 1; A4, calculating the complexity index of the n-dimensional risk feature vector according to the complexity index model; A5, repeating steps A1 to A4 to obtain the complexity index of all dimensional risk feature vectors after dimension reduction, and taking the dimension corresponding to the minimum value of the complexity index as the preferred dimension factor; Training a classification network based on the preferred risk feature vector and the label of the preferred risk feature vector to obtain an enterprise risk judgment network; The optimal risk feature vector of the enterprise to be judged is collected as the input of the enterprise risk judgment network to obtain the risk judgment result of the enterprise to be judged.

2. The enterprise risk judgment method based on artificial intelligence as claimed in claim 1, characterized in that: The querying of the historical registration information of the enterprise from the business database based on the enterprise ID, and obtaining the historical public information of the enterprise based on the open source information of the enterprise, includes: Storing the registration contents of basic information of all enterprises each year to build a business database, wherein the basic information includes all basic character segments related to the basic information; Store the public disclosure of basic information of all enterprises each year to build an open source database; Search the business database according to the enterprise ID to obtain the historical registration information of the enterprise, wherein the historical registration information includes the registration content of the basic information of the enterprise in each year during the historical period; The open source database is searched according to the enterprise ID to obtain the historical public information of the enterprise, wherein the historical public information includes the public content of the basic information of the enterprise each year in the historical period.

3. The enterprise risk judgment method based on artificial intelligence as claimed in claim 1, characterized in that: The comparing the historical registration information and the historical public information of all enterprises to construct the risk feature vectors of all enterprises includes: Taking any character segment among all the basic character segments of the basic information as the target character segment; Acquire the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise, wherein the target enterprise is any one of all enterprises; Traversing all basic character segments of the basic information to obtain the attribute value of each basic character segment; Arranging the attribute values ​​of all basic character segments in a fixed order to obtain the risk feature vector of the target enterprise; Traverse the historical registration information and historical public information of all enterprises to obtain the risk feature vector of each enterprise.

4. The enterprise risk judgment method based on artificial intelligence as claimed in claim 3 is characterized in that: The step of obtaining the attribute value of the target character segment based on the historical registration information and historical public information of the target enterprise includes: Search the historical registration information of the target enterprise based on the target character segment to obtain the registration content of the target character segment in each year; Searching the historical public information of the target enterprise based on the target character segment to obtain the public information of the target character segment in each year; Compare the registered content and the publicized content of the target character segment in the same year to obtain the comparison results for each year; The comparison results of each year are counted to obtain the attribute value of the target character segment.

5. The enterprise risk judgment method based on artificial intelligence as claimed in claim 1, characterized in that: The complexity index model satisfies the relationship: in, are the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; M is the number of all n-dimensional risk feature vectors; is the L2 distance between the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector; are the labels of the i-th n-dimensional risk feature vector and the j-th n-dimensional risk feature vector respectively; α n is the complexity index of the n-dimensional risk feature vector, is a user-defined function, which satisfies the relationship:

6. An enterprise risk judgment device based on artificial intelligence, characterized in that: The device is used to implement the enterprise risk judgment method based on artificial intelligence as described in any one of claims 1 to 5, and the device includes: An acquisition unit is used to query the historical registration information of the enterprise from the business database according to the enterprise ID, and to acquire the historical public information of the enterprise according to the open source information of the enterprise, wherein the historical registration information and the historical public information both include all basic information of the enterprise in each year in the historical period; A comparison unit, used for comparing the historical registration information and the historical public information of all enterprises to construct risk feature vectors of all enterprises; A labeling unit, used for labeling the enterprise to obtain a label of each risk feature vector; A dimensionality reduction unit, configured to calculate a preferred dimensionality factor based on the risk feature vector and the label to obtain a preferred dimensionality reduction matrix, and perform dimensionality reduction processing on all risk feature vectors based on the preferred dimensionality reduction matrix to obtain a preferred risk feature vector and a label of the preferred risk feature vector; A training unit, used for training a classification network based on the preferred risk feature vector and a label of the preferred risk feature vector to obtain an enterprise risk judgment network; The risk judgment unit collects the preferred risk feature vector of the enterprise to be judged as the input of the enterprise risk judgment network to obtain the risk judgment result of the enterprise to be judged.

7. An electronic device, characterized in that: The electronic device comprises: a memory storing computer-readable instructions; and A processor executes computer-readable instructions stored in the memory to implement the enterprise risk judgment method based on artificial intelligence as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the enterprise risk judgment method based on artificial intelligence as described in any one of claims 1 to 5.

Images