CN112364178A - Method for identifying invisible real control people of enterprise based on enterprise associated knowledge graph - Google Patents
Method for identifying invisible real control people of enterprise based on enterprise associated knowledge graph Download PDFInfo
- Publication number
- CN112364178A CN112364178A CN202011234884.9A CN202011234884A CN112364178A CN 112364178 A CN112364178 A CN 112364178A CN 202011234884 A CN202011234884 A CN 202011234884A CN 112364178 A CN112364178 A CN 112364178A
- Authority
- CN
- China
- Prior art keywords
- graph
- enterprise
- path
- association
- knowledge graph
- 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
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 18
- 230000001502 supplementing effect Effects 0.000 claims description 7
- 238000003062 neural network model Methods 0.000 claims description 4
- 239000013598 vector Substances 0.000 abstract description 21
- 239000013589 supplement Substances 0.000 abstract description 6
- 238000013528 artificial neural network Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract 1
- 238000013527 convolutional neural network Methods 0.000 description 14
- 239000011159 matrix material Substances 0.000 description 13
- 238000010586 diagram Methods 0.000 description 8
- 230000008520 organization Effects 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000003795 chemical substances by application Substances 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses an enterprise invisible real control person identification method based on an enterprise associated knowledge graph, which comprises the following specific steps of: and the constructed enterprise association knowledge graph is stored in a graph database and a relational database, and all association path graphs between two target nodes are searched through a graph traversal algorithm. And traversing the associated path graph to form one or more associated path sequence lists. And performing feature supplement and vectorization representation on the nodes and edges in the sequence, and marking whether the target node is a hidden real control person of the target enterprise. And then constructing a deep learning neural network based on an attention model, inputting a plurality of sequence list vectors as parameters, obtaining a model after training, and finally judging whether a corresponding target node in a connected sequence formed by any association relation is a hidden real control person or not by using the trained model.
Description
Technical Field
The invention relates to machine learning and knowledge-graph correlation techniques.
Background
In practice, social public investors often know from the annual newspaper of a listed company who the stockholder of a certain listed company is. However, the actual control of a company may be difficult to discern in some situations. The actual controlling person may be the stockholder, or may be the stockholder of the stockholder, or even other natural person, legal person, or other organization in addition to the actual controlling person. In some anti-fraud scenarios, the entity of some companies is hidden more deeply. The actual controller can control one enterprise through means of relatives holding stocks, guaranties, stock holding agents and the like.
The traditional method for identifying the controller based on the investment relation and strong rules is difficult to satisfy.
The method for identifying whether the invisible control relationship exists between people and enterprises and between enterprises is provided by combining technical means such as knowledge graph and machine learning, and the effect of relationship identification can be improved.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a method for identifying the hidden real control people of the enterprise based on the enterprise associated knowledge graph and machine learning, which can effectively improve the identification accuracy of the hidden real control people of the enterprise.
In order to solve the technical problems, the invention is solved by the following technical scheme:
(1) and constructing an enterprise associated knowledge graph. The method comprises the steps of collecting and purchasing public data, combining the public data with existing data, aligning and associating the public data with the existing data to form an enterprise associated knowledge map, storing the enterprise associated knowledge map into a map database, and storing an association relation through directed edges, such as an investment relation, so that directivity exists.
(2) And marking and arranging training samples. The form of the sample is "[ predictive entity ] control [ target entity ]. Wherein the "forecasting entity" may be a person or a business or organization; the "target entity" is a business. The predicted entity may have no direct relationship with the target entity, but an indirect relationship.
(3) And searching and constructing a sample association path graph. Because a direct incidence relation and an indirect incidence relation exist between two entities in the labeled sample, in the step, the training sample is used for searching a directed incidence path between a 'prediction entity' and a 'target entity' from the enterprise association knowledge graph obtained in the step through graph calculation, such as a depth-first traversal algorithm and a breadth-first traversal algorithm. The directed association path refers to various points and edges passing from the prediction entity to the target entity, the points refer to various other entities, and the edges refer to the directed association relationship graph between the entities.
(4) And traversing the associated path graph to form an associated path sequence list, supplementing and constructing the characteristics of the nodes and the edges to form a characteristic sequence list. And calculating and looping the associated path diagrams of the samples obtained in the steps to obtain a plurality of entity relationship sequences, and vectorizing. The method specifically comprises the following steps: and traversing the associated path graph of each sample, finding all communication paths from the 'prediction entity' to the 'target entity' for the sample, and removing the circular paths to form one or more non-circular communication paths. And performing characteristic engineering on the points and edges on each path in sequence to form characteristic representation. The characteristic engineering comprises supplementing characteristic values from the database for each node on the path and each edge of the incidence relation, and performing various characteristic engineering to form characteristic representation. The feature representations are connected in sequence, thereby generating one or more sequences of features for each labeled sample.
(5) And constructing a neural network model of the identification depth of the correlation sequence. The method specifically comprises the following steps: constructing an input layer, and inputting the characteristic sequence into a network in an ordered characteristic vector mode through word2 vec; constructing a BERT pre-training neural network layer for improving the feature vector of each node and edge, finding out key feature information in the sequence and forming an optimized ordered feature vector; and (2) constructing a CNN convolutional neural network layer, wherein a pair of samples possibly have a plurality of link characteristic sequences, so that characteristic vectors obtained by passing the plurality of link sequences through a BERT layer are subjected to full connection, and then the fully connected characteristic vectors are input into the CNN convolutional neural network for further characteristic optimization to find key characteristics on different links. And finally, classifying vectors obtained by a full connection layer in the CNN convolutional neural network through a softmax layer to obtain a prediction result (whether the prediction result is a hidden real control person or not).
(6) Inputting samples and training a model. And inputting the prepared sample data into the constructed neural network model, and training to obtain a final prediction model.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
fig. 1 is a flowchart illustrating implementation steps of a method for identifying an enterprise invisible real control person based on an enterprise-related knowledge base according to an embodiment of the present invention.
Fig. 2 is a diagram illustrating an association relationship between a prediction entity and a target entity, which is obtained by querying a query statement through a Cypher based on Neo4j in the method disclosed in the embodiment of the present invention.
Fig. 3 is a schematic diagram of an exemplary association path in the method disclosed in the embodiment of the present invention.
Fig. 4 is a schematic diagram illustrating that information indicating that company D invests company B in the method disclosed by the embodiment of the invention will be lost.
Fig. 5 is a pseudo code for calculating the shortest path of the directed graph by using the Dijkstra algorithm implemented by the adjacency matrix in the method disclosed by the embodiment of the present invention.
Fig. 6 is a diagram illustrating an embodiment of a method for obtaining all edge artifacts communicated in a graph according to an adjacency matrix of the graph.
Fig. 7 is a graph traversal algorithm pseudo code based on the shortest path algorithm in the method disclosed in the embodiment of the present invention.
Fig. 8 is a diagram illustrating a graph traversal algorithm path generation based on the shortest path algorithm in the method disclosed in the embodiment of the present invention.
Fig. 9 is an example diagram of an association relationship sample for demonstrating the conversion of a specific diagnosis in the method disclosed in the embodiment of the present invention.
Fig. 10 is a characteristic matrix diagram formed after conversion of an incidence relation sample used for demonstrating conversion of a specific diagnosis in the method disclosed by the embodiment of the invention.
Detailed Description
In order to more specifically describe the present invention, the following detailed description is provided for the technical solution of the present invention with reference to the accompanying drawings and the specific embodiments.
(1) And constructing an enterprise associated knowledge graph. By collecting and purchasing public data, cleaning and structuring the data, combining the data with the existing structured data, aligning and associating the data to form an enterprise associated knowledge graph, and storing the enterprise associated knowledge graph into a graph database, such as Neo4j, the nodes stored in the graph database comprise: personnel, enterprises, institutions, etc.; the edges stored in the graph database include: various directive relationships such as an arbitrary relationship, an investment relationship, a guarantee relationship, an affiliation relationship, an upstream supplier relationship, a downstream supplier relationship, a debt relationship, a litigation relationship, a classmate relationship, a teacher-student relationship, and the like. The attribute data is then stored in a relational database, such as MySql, including the attributes of the nodes and the attributes of the edges. Node attributes, such as: registering capital, establishing time, operating range, industry classification, operating address, industry to which the operating address belongs, industry status, whether to appear on the market, whether to be a taxpayer, academic calendar, gender, age, city and the like; edge attributes, such as: guarantee amount, investment proportion, collateral value, bond amount, relationship type, litigation amount, etc. The graph database and the relational database jointly form an enterprise associated knowledge graph.
(2) And marking and arranging training samples. And collecting sample data of various actual control relations, and editing and processing the sample data into a form of ' forecasting entity ' for actual control ' and ' target entity '. Wherein the "forecasting entity" may be a person or a business or organization; the "target entity" is a business. And the predicted entity may have no direct relationship but an indirect relationship with the target entity. Such as: company a 100% holdings company B, company B100% holdings company C, we can label "company a" real control company C ".
(3) And searching and constructing a sample association path graph. Because a direct incidence relation and an indirect incidence relation exist between two entities in the labeled sample, in the step, the training sample is used for searching a directed incidence path graph between a prediction entity and a target entity from the enterprise incidence knowledge graph obtained in the step through graph calculation, such as a depth-first traversal algorithm and a breadth-first traversal algorithm. The directed association path graph refers to various points and edges passing from a prediction entity to a target entity, the points refer to various other entities on the graph, such as various people, enterprises and organizations on the association path, and the edges refer to direct association relations among the entities on the graph, such as investment relations, guarantee relations, debt-right relations, and relative relations. Here we implement the incidence relation graph between the query prediction entity and the target entity through Cypher query statement based on Neo4 j. See figure 2.
(4) And traversing the associated path graph to form an associated path sequence list, supplementing and constructing the characteristics of the nodes and the edges to form a characteristic sequence list. And traversing the associated path graph of the sample obtained in the step to find one or more shortest associated path sequence lists containing all association relations (edges) from the starting node to the target node, and vectorizing. The method specifically comprises the following steps: and designing and developing a graph traversal algorithm based on a shortest path algorithm, traversing the associated path graph of each sample, and finding one or more shortest associated path sequence lists containing all association relations (edges) from the starting node to the target node.
Referring to the associative path graph of fig. 3, we will need to traverse a list of path sequences containing all associations (edges) from their starting node to the target node:
sequence 1: p ═ a (1), B (2), C (3), D (4), E (5), F () }
Sequence 2: p ═ a (1), B (2), C (3), D (6), B (2), C (3), D (4), E (5), F () }
And (3) sequence: p ═ a (1), B (2), C (3), D (4), E (5), F () }
Because a ring path may exist in an original graph, a common depth-first traversal algorithm can cause the dead loop to enter; while the common loop-free depth-first traversal algorithm can cause part of associated information to be lost, which is referred to information of investment B company of company D in the figure 4; and the common shortest path algorithm can only find a single shortest path, so that the relation loss cannot meet the requirement.
Here, we designed and developed a graph traversal algorithm based on the shortest path algorithm for improving the processing of the circular paths, and form one or more shortest connected path sequence lists containing all the association relations (edges) from the starting node to the target node of the sample. The graph traversal algorithm based on the shortest path algorithm includes three basic algorithms.
The first algorithm is as follows: the Dijkstra algorithm is implemented using the adjacency matrix to compute the shortest path of the directed graph, see fig. 5.
And (3) algorithm II: all edges in the graph that have been connected are obtained from the adjacency matrix of the graph, see fig. 6.
And (3) algorithm III: graph traversal algorithm based on shortest path algorithm, see fig. 7.
The specific steps are shown in figure 8.
And after a plurality of associated paths are obtained, performing characteristic engineering on all points and edges on each path in sequence to form characteristic representation of a specific sequence. For different node types, the method is supplemented by inquiring characteristic values of relevant dimensions in a relational database, such as: people inquire and supplement different school calendars, sexes, ages and the like; the enterprise inquires and supplements industries, establishment duration, registration capital, employee number, the place of business and the like; the organization queries and supplements the organization type, the location of the organization, and the like. And for different edge types, supplementing by inquiring related dimension characteristic values in a re-relational database, such as: the relationship between relatives queries and supplements the type of relatives; the relationship between students can be used for inquiring and supplementing schools, years and professional names; the investment relation inquires and supplements the investment amount, the investment proportion and the like; the relationship between duties is inquired and supplemented for the type of duties and the number of years of duties.
And after information is supplemented for each node and edge on the path, starting to perform characteristic engineering for each node and edge to form characteristic representation. Firstly, an m + n-dimensional characteristic number array is constructed for each entity and each edge, wherein m represents the total number of the dimensions of the node, and n represents the total number of the dimensions of the edge. Then, performing characteristic engineering on all latitude characteristic values, and adopting different technologies according to different types of the characteristic values, such as: if the numerical value is of a type, continuous numerical value discrete words are converted into discrete enumeration values through a box separation and partition technology, or characteristic values are compressed through methods such as standardization, normalization and the like, the characteristics of the continuous numerical values are reserved, and large values are compressed; if the character string type is adopted, the characteristic values are cleaned and classified into discrete enumeration values; for the date type, the number of seconds from the current time is obtained by calculation, and then normalization, and the like are performed. After the m + n dimensional features are formed, connecting the entities and the edges according to a path sequence, and forming m + n sequences, wherein each sequence is a series of feature values generated by one dimension according to the path sequence.
See fig. 9 for an illustrative case. The feature matrix is formed after conversion, see fig. 10.
(5) And constructing a neural network model of the identification depth of the correlation sequence. The BERT model is generally used for natural language processing tasks, semantic vector representations of a target word and each word of a context are used as input in a decode stage by utilizing an Attention mechanism in a transform of the BERT model, Query vector representations of the target word, Key vector representations of each word of the context and original Value representations of the target word and each word of the context are obtained through linear transformation, similarity between the Query vector and each Key vector is calculated to serve as weight, and finally, the weight relation between each target word and the word of the context is formed, and the weight sum is 1. And performing weighted fusion on the Value vector of the target word and the Value vectors of the upper and lower characters through dot multiplication to serve as the output of the Attention. In the invention, an Attention mechanism of a BERT model is utilized to pre-train a path feature matrix, and different enumerated feature values are used as Token to be embedding. The method specifically comprises the following steps:
and constructing a path feature matrix input layer. Because the types of feature values of different dimensions are different, different embedding methods are adopted. For enumerated eigenvalues, a BERT pre-training model is constructed, an Embedding layer of the CNN is replaced by the BERT model, and then the output of the BERT is used as the input of the CNN. Inputting an enumerated value serving as Token into a Token embedding layer of a BERT model to realize vectorization representation, capturing effective transfer information through pre-training, and inputting the effective information into a network in an ordered characteristic vector mode; for the numerical eigenvalues, we use it directly, since normalization has been done in the previous step. And then transforming the matrix into the eigenvector in a full connection mode.
And (4) constructing a CNN convolutional neural network layer, inputting the feature vector obtained in the last step into the CNN convolutional neural network, and performing further feature optimization. Since there may be multiple paths for a pair of samples, we perform path feature matrix input layer and CNN convolution layer for multiple paths respectively, sum and normalize the output vectors obtained from the fully-connected layers of CNN. And finally, classifying the obtained vector through a softmax layer, and outputting a prediction result (whether the vector is a hidden real control person or not).
(6) Inputting samples and training a model. And initializing and assigning values to each vector of the path characteristic matrix input layer, sending the processed sample into a BERT pre-training model for fine adjustment, simultaneously training the path characteristic matrix input layer, and changing through a reverse propagation gradient to finally obtain a vector matrix of the characteristic input layer and a trained hidden real control person classification model.
Claims (1)
1. An enterprise invisible real control person identification method based on an enterprise associated knowledge graph is characterized by comprising the following steps:
s1, constructing an enterprise associated knowledge graph, wherein the graph takes entity objects such as people and enterprises as nodes, and the association relations between people and enterprises, between people and people, between enterprises as edges, collecting data and storing the entity objects and the association relations into the enterprise associated knowledge graph, and the graph can be stored by using a graph database and a relational database;
s2, marking and sorting the training samples, marking whether the relation between the prediction entity and the target entity hides the real control person, and forming a uniform training sample set;
s3, searching and constructing a sample association path graph, and searching a complete association path graph from the predicted entity to the target entity in the enterprise association knowledge graph according to the training sample;
and S4, traversing the associated path graph to form an associated path sequence list, supplementing and constructing the characteristics of nodes and edges to form a characteristic sequence list, obtaining one or more associated path sequence lists containing all association relations by utilizing a graph traversal algorithm, supplementing characteristic dimensions by utilizing the information in the existing enterprise associated knowledge graph, performing characteristic engineering, and forming the associated path characteristic sequence list of the sample.
S5, constructing an association sequence recognition deep neural network model for training the model and vectorizing according to different feature types;
and S6, inputting a sample, training a model and obtaining a final prediction model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011234884.9A CN112364178A (en) | 2020-11-08 | 2020-11-08 | Method for identifying invisible real control people of enterprise based on enterprise associated knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011234884.9A CN112364178A (en) | 2020-11-08 | 2020-11-08 | Method for identifying invisible real control people of enterprise based on enterprise associated knowledge graph |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112364178A true CN112364178A (en) | 2021-02-12 |
Family
ID=74509760
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011234884.9A Pending CN112364178A (en) | 2020-11-08 | 2020-11-08 | Method for identifying invisible real control people of enterprise based on enterprise associated knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112364178A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113326377A (en) * | 2021-06-02 | 2021-08-31 | 上海生腾数据科技有限公司 | Name disambiguation method and system based on enterprise incidence relation |
| CN113641878A (en) * | 2021-08-09 | 2021-11-12 | 平安科技（深圳）有限公司 | Real control person identification method, device, equipment and medium suitable for directed loop |
| CN114169458A (en) * | 2021-12-13 | 2022-03-11 | 平安科技（深圳）有限公司 | Method and device for identifying cheater, storage medium and computer equipment |
| CN114202223A (en) * | 2021-12-16 | 2022-03-18 | 深圳前海微众银行股份有限公司 | Enterprise credit risk scoring method, device, equipment and storage medium |
| CN114611515A (en) * | 2022-01-28 | 2022-06-10 | 江苏省联合征信有限公司 | Method and system for identifying actual control person of enterprise based on enterprise public opinion information |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110909986A (en) * | 2019-11-04 | 2020-03-24 | 苏宁金融科技(南京)有限公司 | Suspected actual controller risk identification method and system based on knowledge graph |
| CN110968701A (en) * | 2019-11-05 | 2020-04-07 | 量子数聚（北京）科技有限公司 | Relationship map establishing method, device and equipment for graph neural network |
| CN111400504A (en) * | 2020-03-12 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | Method and device for identifying enterprise key people |
| CN111553786A (en) * | 2020-04-24 | 2020-08-18 | 中金汇安(北京)科技有限公司 | Bank shareholder loan association transaction mining method and system based on graphic database |
-
2020
- 2020-11-08 CN CN202011234884.9A patent/CN112364178A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110909986A (en) * | 2019-11-04 | 2020-03-24 | 苏宁金融科技(南京)有限公司 | Suspected actual controller risk identification method and system based on knowledge graph |
| CN110968701A (en) * | 2019-11-05 | 2020-04-07 | 量子数聚（北京）科技有限公司 | Relationship map establishing method, device and equipment for graph neural network |
| CN111400504A (en) * | 2020-03-12 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | Method and device for identifying enterprise key people |
| CN111553786A (en) * | 2020-04-24 | 2020-08-18 | 中金汇安(北京)科技有限公司 | Bank shareholder loan association transaction mining method and system based on graphic database |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113326377A (en) * | 2021-06-02 | 2021-08-31 | 上海生腾数据科技有限公司 | Name disambiguation method and system based on enterprise incidence relation |
| CN113326377B (en) * | 2021-06-02 | 2023-10-13 | 上海生腾数据科技有限公司 | Name disambiguation method and system based on enterprise association relationship |
| CN113641878A (en) * | 2021-08-09 | 2021-11-12 | 平安科技（深圳）有限公司 | Real control person identification method, device, equipment and medium suitable for directed loop |
| CN114169458A (en) * | 2021-12-13 | 2022-03-11 | 平安科技（深圳）有限公司 | Method and device for identifying cheater, storage medium and computer equipment |
| CN114202223A (en) * | 2021-12-16 | 2022-03-18 | 深圳前海微众银行股份有限公司 | Enterprise credit risk scoring method, device, equipment and storage medium |
| CN114611515A (en) * | 2022-01-28 | 2022-06-10 | 江苏省联合征信有限公司 | Method and system for identifying actual control person of enterprise based on enterprise public opinion information |
| CN114611515B (en) * | 2022-01-28 | 2023-12-12 | 江苏省联合征信有限公司 | Method and system for identifying enterprise actual control person based on enterprise public opinion information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112364178A (en) | Method for identifying invisible real control people of enterprise based on enterprise associated knowledge graph | |
| CN111581395B (en) | Model fusion triplet representation learning system and method based on deep learning | |
| CN110245238B (en) | Graph embedding method and system based on rule reasoning and syntax mode | |
| CN112036842B (en) | Intelligent matching device for scientific and technological service | |
| US20150347489A1 (en) | Information retrieval system and method based on query and record metadata in combination with relevance between disparate items in classification systems | |
| CN112559734B (en) | Brief report generating method, brief report generating device, electronic equipment and computer readable storage medium | |
| CN112925908A (en) | Attention-based text classification method and system for graph Attention network | |
| CN113269239B (en) | Relation network node classification method based on multichannel convolutional neural network | |
| WO2023035330A1 (en) | Long text event extraction method and apparatus, and computer device and storage medium | |
| CN113010683B (en) | Entity relationship identification method and system based on improved graph attention network | |
| CN109697679A (en) | Intellectual property services guidance method and system | |
| CN112989761A (en) | Text classification method and device | |
| CN116821372A (en) | Knowledge graph-based data processing method and device, electronic equipment and medium | |
| Okhunov et al. | Tools to support the Development and Promotion of Innovative Projects | |
| CN115221387A (en) | Enterprise information integration method based on deep neural network | |
| CN115982338B (en) | Domain knowledge graph question-answering method and system based on query path sorting | |
| CN116842195A (en) | Knowledge graph and large model based automatic generation report generation method | |
| CN111353728A (en) | Risk analysis method and system | |
| Handayani et al. | Sentiment Analysis Of Electric Cars Using Recurrent Neural Network Method In Indonesian Tweets | |
| Gao et al. | Mining and Visualizing Cost and Schedule Risks from News Articles with NLP and Network Analysis | |
| CN111626331B (en) | Automatic industry classification device and working method thereof | |
| Thant et al. | Evaluation of student academic performance using Naïve Bayes classifier | |
| Sehrawat | Learning word embeddings from 10-K filings for financial NLP tasks | |
| CN112732889A (en) | Student retrieval method and device based on cooperative network | |
| Trivedi et al. | A review paper on a comparative study of supervised learning approaches |
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 | ||
| CB02 | Change of applicant information |
Address after: 310019 room 808, floor 8, building 1, No. 8 and 10, Jiuhe Road, Shangcheng District, Hangzhou, Zhejiang Applicant after: Zhejiang youshuzhi Technology Co.,Ltd. Address before: 310000 room 808, 8 / F, building 4, No. 9, Jiuhuan Road, Jianggan District, Hangzhou City, Zhejiang Province Applicant before: HANGZHOU YOUSHU FINANCE INFORMATION SERVICES CO.,LTD. |
|
| CB02 | Change of applicant information |