CN116611895A - Transaction abnormality identification method, apparatus, computer device, and storage medium - Google Patents

Transaction abnormality identification method, apparatus, computer device, and storage medium Download PDF

Info

Publication number
CN116611895A
CN116611895A CN202310648422.9A CN202310648422A CN116611895A CN 116611895 A CN116611895 A CN 116611895A CN 202310648422 A CN202310648422 A CN 202310648422A CN 116611895 A CN116611895 A CN 116611895A
Authority
CN
China
Prior art keywords
transaction
merchant
client
reputation
transaction information
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310648422.9A
Other languages
Chinese (zh)
Inventor
黄琼
暨光耀
张晓娜
林嘉婷
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Industrial and Commercial Bank of China Ltd ICBC
Original Assignee
Industrial and Commercial Bank of China Ltd ICBC
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Industrial and Commercial Bank of China Ltd ICBC filed Critical Industrial and Commercial Bank of China Ltd ICBC
Priority to CN202310648422.9A priority Critical patent/CN116611895A/en
Publication of CN116611895A publication Critical patent/CN116611895A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0609—Buyer or seller confidence or verification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Business, Economics & Management (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Mathematical Physics (AREA)
  • Finance (AREA)
  • Computational Linguistics (AREA)
  • Accounting & Taxation (AREA)
  • General Business, Economics & Management (AREA)
  • Medical Informatics (AREA)
  • Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application relates to a transaction anomaly identification method, a transaction anomaly identification device, computer equipment, a storage medium and a computer program product, which can be used in the technical field of artificial intelligence. The method comprises the following steps: acquiring transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier; acquiring merchant characteristics corresponding to the merchant identifications and customer characteristics corresponding to the customer identifications; identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier; under the condition that the merchant reputation parameter and the client reputation parameter meet the preset reputation condition, inputting the transaction information to be identified into a pre-trained transaction anomaly identification model to obtain a transaction anomaly identification result aiming at the transaction information to be identified. By adopting the method, the transaction abnormality identification accuracy can be improved.

Description

Transaction abnormality identification method, apparatus, computer device, and storage medium
Technical Field
The present application relates to the field of artificial intelligence technology, and in particular, to a transaction anomaly identification method, apparatus, computer device, storage medium, and computer program product.
Background
With the development of internet technology, people live in a plurality of ways through online transaction, such as credit card payment. Therefore, in order to secure transaction security, it is necessary to perform abnormality recognition on the transaction.
In the prior art, when abnormal transaction identification is carried out, the transaction information is analyzed mainly in a manual auditing mode, so as to judge whether the transaction information is abnormal transaction or not; however, the transaction information is more, and the erroneous judgment and the missed judgment are easy to exist in a manual auditing mode, so that the abnormal recognition accuracy of the transaction is lower.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a transaction anomaly identification method, apparatus, computer device, computer-readable storage medium, and computer program product that can improve the accuracy of transaction anomaly identification.
In a first aspect, the present application provides a transaction anomaly identification method. The method comprises the following steps:
acquiring transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
acquiring merchant characteristics corresponding to the merchant identifications and customer characteristics corresponding to the customer identifications;
Identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier;
and under the condition that the merchant credit parameter and the client credit parameter meet the preset credit condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
In one embodiment, the identifying the merchant reputation parameter corresponding to the merchant identifier according to the merchant feature corresponding to the merchant identifier includes:
inputting the merchant characteristics corresponding to the merchant identification into a pre-trained reputation prediction model, and extracting merchant transaction characteristics and merchant basic characteristics from the merchant characteristics through the pre-trained reputation prediction model;
acquiring attention weights corresponding to the transaction characteristics of all the merchants, and carrying out fusion processing on the transaction characteristics of all the merchants according to the attention weights corresponding to the transaction characteristics of all the merchants to obtain target transaction characteristics corresponding to the merchant identifications;
Carrying out fusion processing on the target transaction characteristics corresponding to the merchant identifications and the merchant basic characteristics to obtain fusion characteristics corresponding to the merchant identifications;
and performing reputation prediction processing according to the fusion characteristics corresponding to the merchant identifications to obtain merchant reputation parameters corresponding to the merchant identifications.
In one embodiment, the identifying the client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier includes:
inputting the client characteristics corresponding to the client identifications into a pre-trained reputation prediction model, and extracting client transaction characteristics and client basic characteristics from the client characteristics through the pre-trained reputation prediction model;
acquiring attention weights corresponding to the client transaction characteristics, and carrying out fusion processing on the client transaction characteristics according to the attention weights corresponding to the client transaction characteristics to obtain target transaction characteristics corresponding to the client identifications;
carrying out fusion processing on the target transaction characteristic corresponding to the client identifier and the client basic characteristic to obtain a fusion characteristic corresponding to the client identifier;
And performing reputation prediction processing according to the fusion characteristics corresponding to the client identifications to obtain client reputation parameters corresponding to the client identifications.
In one embodiment, the pre-trained reputation prediction model is trained by:
acquiring sample transaction information and an actual transaction identification result corresponding to the sample transaction information; sample trade information carries sample merchant identification and sample customer identification;
acquiring sample merchant characteristics corresponding to the sample merchant identifications and sample customer characteristics corresponding to the sample customer identifications;
respectively inputting the sample merchant characteristics and the sample client characteristics into a reputation prediction model to be trained to obtain sample merchant reputation parameters corresponding to the sample merchant identifications and sample client reputation parameters corresponding to the sample client identifications;
performing full connection processing on the sample merchant reputation parameter and the sample customer reputation parameter to obtain a predicted transaction identification result corresponding to the sample transaction information;
and training the reputation prediction model to be trained according to the difference between the predicted transaction recognition result and the actual transaction recognition result to obtain a trained reputation prediction model, and taking the trained reputation prediction model as the pre-trained reputation prediction model.
In one embodiment, the pre-trained transaction anomaly identification model is trained by:
sample transaction information is obtained, and the sample transaction information is divided to obtain a training data set and a test data set;
deleting abnormal data in the training data set by using an isolated forest model to obtain a target training data set;
training the transaction anomaly recognition model to be trained according to the target training data set to obtain a trained transaction anomaly recognition model;
and retraining the trained transaction anomaly recognition model according to the test data set to obtain a trained transaction anomaly recognition model serving as the pre-trained transaction anomaly recognition model.
In one embodiment, after inputting the transaction information to be identified into a pre-trained transaction anomaly identification model to obtain a transaction anomaly identification result for the transaction information to be identified, the method further includes:
under the condition that the transaction abnormality identification result is abnormal transaction, determining a merchant transaction information set corresponding to the merchant identification and a customer transaction information set corresponding to the customer identification according to the transaction information to be identified;
Identifying a first association rule matched with the merchant transaction information set and a second association rule matched with the customer transaction information set from a plurality of association rules stored in an abnormal transaction association rule library;
screening a first target association rule with the largest corresponding association degree from the first association rules, and taking the association degree corresponding to the first target association rule as the first association degree between the transaction information to be identified and the merchant of which the merchant identification belongs;
screening a second target association rule with the largest corresponding association degree from the second association rules, and taking the association degree corresponding to the second target association rule as a second association degree between the transaction information to be identified and the client to which the client identifier belongs;
and updating the transaction anomaly identification result according to the first association degree and the second association degree to obtain a target transaction anomaly identification result aiming at the transaction information to be identified.
In one embodiment, the updating the transaction anomaly identification result according to the first association degree and the second association degree to obtain a target transaction anomaly identification result for the transaction information to be identified includes:
Comparing the first association degree with the second association degree with a preset association degree to obtain a comparison result; the comparison result is used for representing whether the transaction information to be identified is abnormal transaction or not;
and updating the transaction abnormality identification result according to the comparison result to obtain a target transaction abnormality identification result aiming at the transaction information to be identified.
In a second aspect, the application also provides a transaction abnormality identification device. The device comprises:
the information acquisition module is used for acquiring transaction information to be identified, which is broadcasted in the blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
the feature acquisition module is used for acquiring the merchant features corresponding to the merchant identifications and the customer features corresponding to the customer identifications;
the credit recognition module is used for recognizing merchant credit parameters corresponding to the merchant identifications according to the merchant characteristics corresponding to the merchant identifications and recognizing client credit parameters corresponding to the client identifications according to the client characteristics corresponding to the client identifications;
and the result determining module is used for inputting the transaction information to be identified into a pre-trained transaction abnormality identification model under the condition that the merchant reputation parameter and the client reputation parameter meet the preset reputation condition to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor which when executing the computer program performs the steps of:
acquiring transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
acquiring merchant characteristics corresponding to the merchant identifications and customer characteristics corresponding to the customer identifications;
identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier;
and under the condition that the merchant credit parameter and the client credit parameter meet the preset credit condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:
Acquiring transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
acquiring merchant characteristics corresponding to the merchant identifications and customer characteristics corresponding to the customer identifications;
identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier;
and under the condition that the merchant credit parameter and the client credit parameter meet the preset credit condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
In a fifth aspect, the present application also provides a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the steps of:
acquiring transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
acquiring merchant characteristics corresponding to the merchant identifications and customer characteristics corresponding to the customer identifications;
Identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier;
and under the condition that the merchant credit parameter and the client credit parameter meet the preset credit condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
The transaction anomaly identification method, the device, the computer equipment, the storage medium and the computer program product firstly acquire transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier; then, acquiring the merchant characteristics corresponding to the merchant identifications and the customer characteristics corresponding to the customer identifications; then, according to the merchant characteristics corresponding to the merchant identifications, the merchant reputation parameters corresponding to the merchant identifications are identified, and according to the client characteristics corresponding to the client identifications, the client reputation parameters corresponding to the client identifications are identified; and finally, under the condition that the merchant reputation parameter and the customer reputation parameter meet the preset reputation condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified. In this way, when the transaction anomaly identification is carried out, firstly, according to the merchant characteristics corresponding to the merchant identification carried by the transaction information to be identified and the client characteristics corresponding to the client identification, the merchant reputation parameter and the client reputation parameter are identified, and under the condition that the merchant reputation parameter and the client reputation parameter meet the preset reputation condition, the transaction anomaly identification model trained in advance is utilized to output the transaction anomaly identification result aiming at the transaction information to be identified; the merchant reputation parameter, the customer reputation parameter and the transaction information to be identified are comprehensively considered, and the transaction anomaly identification model trained in advance is utilized for identification, so that manual verification is not needed, the determination accuracy of the transaction anomaly identification result is improved, and the transaction anomaly identification accuracy is improved.
Drawings
FIG. 1 is a flow chart of a transaction anomaly identification method in one embodiment;
FIG. 2 is a block chain network architecture diagram in one embodiment;
FIG. 3 is a flowchart of a transaction anomaly identification method according to another embodiment;
FIG. 4 is a flowchart illustrating steps for calculating a merchant reputation parameter in one embodiment;
FIG. 5 is a flowchart illustrating steps for calculating a merchant reputation parameter in another embodiment;
FIG. 6 is a flowchart illustrating steps for calculating a client reputation parameter in one embodiment;
FIG. 7 is a flowchart illustrating training steps of a transaction anomaly identification model in one embodiment;
FIG. 8 is a flowchart illustrating steps for updating transaction anomaly recognition results in one embodiment;
FIG. 9 is a flow chart of a transaction anomaly identification method in yet another embodiment;
FIG. 10 is a flowchart illustrating steps for transmitting encrypted information to a blockchain network for broadcasting in one embodiment;
FIG. 11 is a block diagram showing a construction of a transaction abnormality recognition device in one embodiment;
fig. 12 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
It should be noted that, the transaction anomaly identification method, the device, the computer equipment, the storage medium and the computer program product provided by the application can be used in the field of financial science and technology, such as performing a series of identifications on the transaction information to be identified, which is broadcasted in the blockchain network, to obtain the transaction anomaly identification result aiming at the transaction information to be identified, without manual auditing, thereby improving the transaction anomaly identification accuracy; the method can also be used in other related fields, such as the artificial intelligence technical field, for identifying the transaction information to be identified by utilizing a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.
In one embodiment, as shown in fig. 1, a transaction anomaly identification method is provided, and the method is applied to a server for illustration in the embodiment; it will be appreciated that the method may also be applied to a terminal, and may also be applied to a system comprising a terminal and a server, and implemented by interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and the like; the server refers to a bank server, and can be implemented by a stand-alone server or a server cluster formed by a plurality of servers. In this embodiment, the method includes the steps of:
Step S101, obtaining transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier.
The blockchain network related to the application refers to a transaction blockchain network shown in fig. 2, and specifically comprises a consumption channel node, a merchant node and a banking node; the consumption channel node refers to a consumption channel terminal or a consumption channel server, the merchant node refers to a merchant terminal or a merchant server, and the bank node refers to a bank server. It should be noted that the present application is mainly exemplified by using a bank server as an execution subject.
The transaction information to be identified is transaction information related to credit card transaction, and specifically comprises a consumption channel identifier which is unique in the whole network and one or more credit card transaction information recorded by a consumption channel to which the consumption channel identifier belongs; the credit card transaction information comprises a unique transaction time stamp, commodity information of the transaction, transaction amount of the transaction, customer information, bank information corresponding to the customer and merchant information. It should be noted that, after interacting with the customer, the consumption channel node obtains the current transaction information as the transaction information to be identified, encrypts the transaction information to be identified and broadcasts the encrypted transaction information to the blockchain network, and both the merchant node and the bank node can obtain the encrypted transaction information to be identified from the blockchain network, decrypt the encrypted transaction information to be identified to obtain the original transaction information to be identified, and then perform validity verification on the transaction information to be identified.
The merchant identification refers to identification information corresponding to a merchant related to the transaction information to be identified, such as a merchant name, a merchant number and the like. The client identifier refers to identification information corresponding to a client related to the transaction information to be identified, such as a client name, a client number and the like.
Specifically, the server responds to a transaction abnormality identification request, acquires encrypted transaction information to be identified, which is broadcasted in a blockchain network, and decrypts the encrypted transaction information to be identified to obtain original transaction information to be identified; and analyzing the transaction information to be identified to obtain the merchant identification and the customer identification carried by the transaction information to be identified.
Step S102, acquiring the merchant characteristics corresponding to the merchant identification and the customer characteristics corresponding to the customer identification.
The merchant feature is used for characterizing feature information of the merchant, and specifically comprises a credit card consumption number ratio, a credit card consumption amount ratio, a historical month consumption amount mean value, a business range, a merchant address and the like.
The characteristic information of the client is characterized by specifically comprising a current month credit card consumption amount ratio, a last year credit card consumption amount ratio, a credit card amount exceeding 1 and the like.
Specifically, the server acquires merchant data corresponding to the merchant identifier from the blockchain network, and performs preprocessing on the merchant data, such as removing invalid data, redundant data and the like, so as to obtain preprocessed merchant data; performing feature extraction processing on the preprocessed merchant data to obtain merchant features; for example, the merchant data is input into a feature extraction model (such as a convolutional neural network model) to perform feature extraction processing, so as to obtain merchant features.
Similarly, the server acquires client data corresponding to the client identifier from the blockchain network, and preprocesses the client data, such as removing invalid data, redundant data and the like, so as to obtain preprocessed client data; performing feature extraction processing on the preprocessed customer data to obtain customer features; for example, the customer data is input into a feature extraction model (such as a convolutional neural network model) to perform feature extraction processing, so as to obtain the merchant features.
Step S103, identifying the merchant credit parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying the client credit parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier.
The merchant reputation parameter is a merchant reputation value, and is used for measuring the reputation of the merchant, for example, the higher the merchant reputation value is, the lower the possibility that the current transaction is an abnormal transaction is.
The client reputation parameter is a client reputation value, and is used for measuring the reputation of the client, for example, the higher the client reputation value is, the lower the possibility that the current transaction is an abnormal transaction is.
Specifically, the server inputs the merchant features corresponding to the merchant identifications into an attention mechanism model, attention weights corresponding to the merchant features are obtained through the attention mechanism model, and fusion processing is carried out on the merchant features by utilizing the attention weights corresponding to the merchant features to obtain fusion merchant features; performing reputation prediction processing according to the converged merchant characteristics to obtain merchant reputation parameters corresponding to merchant identifications; for example, the reputation parameter corresponding to the converged merchant feature is obtained as the merchant reputation parameter.
Similarly, the server inputs the client features corresponding to the client identifications into an attention mechanism model, attention weights corresponding to the client features are obtained through the attention mechanism model, and fusion processing is carried out on the client features by utilizing the attention weights corresponding to the client features to obtain fusion client features; performing reputation prediction processing according to the fused client characteristics to obtain client reputation parameters corresponding to the client identifications; for example, reputation parameters corresponding to the converged client characteristics are obtained as client reputation parameters.
Step S104, under the condition that the merchant reputation parameter and the customer reputation parameter meet the preset reputation condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
The merchant reputation parameter and the client reputation parameter meet the preset reputation condition, namely the merchant reputation parameter and the client reputation parameter are smaller than the preset reputation parameter, and particularly the merchant reputation value and the client reputation value are smaller than the preset reputation value.
The transaction anomaly recognition model trained in advance is a model for judging whether the transaction information is an anomaly transaction, such as a neural network model, a machine learning model, a single-class support vector machine classification model, a decision model and the like.
The abnormal transaction recognition result of the transaction information to be recognized refers to that the transaction information to be recognized is normal transaction or abnormal transaction.
It should be noted that, if both the merchant reputation parameter and the customer reputation parameter are smaller than the preset reputation parameter, the possibility that the transaction information to be identified is abnormal transaction is high, and at this time, the transaction information to be identified is identified again by using the pre-trained transaction abnormal identification model, so that the accuracy of identifying the transaction abnormal can be further improved, and erroneous judgment is avoided.
Specifically, the server compares the merchant reputation parameter and the client reputation parameter with preset reputation parameters, if the merchant reputation parameter and the client reputation parameter are smaller than the preset reputation parameter, the merchant reputation parameter and the client reputation parameter are indicated to meet the preset reputation condition, a pre-trained transaction anomaly identification model is obtained, transaction information to be identified is input into the pre-trained transaction anomaly identification model, feature extraction processing is conducted on the transaction information to be identified through the pre-trained transaction anomaly identification model, to-be-identified transaction characteristics are obtained, full connection processing is conducted on the to-be-identified transaction characteristics, and the probability that the transaction information to be identified belongs to anomaly transaction is obtained; if the probability is larger than the preset probability, the transaction information to be identified belongs to abnormal transaction, and if the probability is smaller than or equal to the preset probability, the transaction information to be identified belongs to normal transaction, so that a transaction abnormal identification result aiming at the transaction information to be identified is obtained.
For example, referring to FIG. 3, the server calculates a merchant reputation value and a customer reputation value using merchant features corresponding to merchants involved in the current transaction information and customer features corresponding to customers involved in the current transaction information; under the condition that the credit value of the merchant and the credit value of the client are both larger than the preset credit value, judging that the current transaction information is normal transaction, and storing the current transaction information into a normal transaction information base; under the condition that the merchant credit value and the client credit value are smaller than the preset credit value, judging that the current transaction information is suspicious transaction, inputting the current transaction information into a transaction abnormality recognition model trained based on historical data, and recognizing the current transaction information through the transaction abnormality recognition model to judge whether the current transaction information is abnormal transaction.
Further, under the condition that the abnormal transaction identification result is normal transaction, the node where the server is located returns success to the corresponding merchant node and pays corresponding amount to the merchant account; and under the condition that the transaction abnormality identification result is abnormal transaction, the node where the server is located refuses the transaction.
The transaction anomaly identification method provided by the embodiment of the invention firstly acquires the transaction information to be identified, which is broadcasted in the blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier; then, acquiring the merchant characteristics corresponding to the merchant identifications and the customer characteristics corresponding to the customer identifications; then, according to the merchant characteristics corresponding to the merchant identifications, the merchant reputation parameters corresponding to the merchant identifications are identified, and according to the client characteristics corresponding to the client identifications, the client reputation parameters corresponding to the client identifications are identified; and finally, under the condition that the merchant reputation parameter and the customer reputation parameter meet the preset reputation condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified. In this way, when the transaction anomaly identification is carried out, firstly, according to the merchant characteristics corresponding to the merchant identification carried by the transaction information to be identified and the client characteristics corresponding to the client identification, the merchant reputation parameter and the client reputation parameter are identified, and under the condition that the merchant reputation parameter and the client reputation parameter meet the preset reputation condition, the transaction anomaly identification model trained in advance is utilized to output the transaction anomaly identification result aiming at the transaction information to be identified; the merchant reputation parameter, the customer reputation parameter and the transaction information to be identified are comprehensively considered, and the transaction anomaly identification model trained in advance is utilized for identification, so that manual verification is not needed, the determination accuracy of the transaction anomaly identification result is improved, and the transaction anomaly identification accuracy is improved.
In one embodiment, as shown in fig. 4, the step S103 identifies a merchant reputation parameter corresponding to a merchant identifier according to a merchant feature corresponding to the merchant identifier, and specifically includes the following steps:
step S401, inputting the merchant characteristics corresponding to the merchant identifications into a pre-trained reputation prediction model, and extracting merchant transaction characteristics and merchant basic characteristics from the merchant characteristics through the pre-trained reputation prediction model.
Step S402, attention weights corresponding to all the trade features of the merchants are obtained, and fusion processing is carried out on all the trade features of the merchants according to the attention weights corresponding to the trade features of the merchants to obtain target trade features corresponding to the merchant identifications.
Step S403, the target transaction feature corresponding to the merchant identifier and the merchant basic feature are subjected to fusion processing, and fusion features corresponding to the merchant identifier are obtained.
And step S404, performing reputation prediction processing according to the fusion characteristics corresponding to the merchant identifications to obtain merchant reputation parameters corresponding to the merchant identifications.
The pre-trained reputation prediction model refers to a network model for predicting reputation parameters, such as a deep learning model, an attention mechanism model and the like.
The merchant transaction feature refers to a merchant business feature and is used for representing feature information related to merchant business, and specifically comprises a credit card consumption count ratio, a credit card consumption amount ratio, a historical month consumption amount average value, a historical abnormal transaction amount ratio, a historical abnormal transaction count ratio and the like. The merchant basic feature refers to a merchant fixed feature, and is used for representing feature information related to merchant attributes, and specifically comprises business scope, merchant address and the like.
The attention weight corresponding to the trade feature of the merchant is used for measuring the importance corresponding to the trade feature of the merchant. The target transaction characteristics of the merchant are characteristics obtained by fusing the transaction characteristics of the merchant.
The fusion characteristic corresponding to the merchant identifier refers to a splicing characteristic obtained after fusion processing is performed on the target transaction characteristic of the merchant and the basic characteristic of the merchant. Different fusion features correspond to different merchant reputation parameters.
Specifically, the server inputs the merchant characteristics corresponding to the merchant identifications into a pre-trained reputation prediction model, performs characteristic extraction processing on the merchant characteristics through the pre-trained reputation prediction model to obtain merchant transaction characteristics and merchant basic characteristics, acquires attention distribution values of the transaction characteristics of each merchant, and calculates attention weights of the transaction characteristics of each merchant according to the attention distribution values of the transaction characteristics of each merchant; according to the attention weight corresponding to each merchant transaction feature, carrying out pooling summation processing on each merchant transaction feature to obtain a total merchant transaction feature serving as a target transaction feature corresponding to the merchant identification; performing splicing processing on the target transaction characteristics corresponding to the merchant identifications and the merchant basic characteristics to obtain spliced characteristics serving as fusion characteristics corresponding to the merchant identifications; and determining the reputation parameter matched with the fusion feature by using the self-adaptive activation function as a merchant reputation parameter corresponding to the merchant identification.
For example, referring to fig. 5, the server first obtains the merchant information, and performs preprocessing (such as removing redundant data) on the merchant information to obtain preprocessed merchant information; feature extraction processing is carried out on the preprocessed merchant information to obtain merchant business features and merchant fixed features; the attention weight of each business feature of the commercial tenant is obtained through an attention mechanism algorithm; and carrying out pooling summation treatment on each business feature of the commercial tenant according to the attention weight of each business feature of the commercial tenant to obtain a total business feature of the commercial tenant, splicing and smoothing the total business feature and the commercial tenant fixed feature to obtain a spliced feature of the commercial tenant, and processing the spliced feature of the commercial tenant by utilizing a self-adaptive activation function to obtain a credit value of the commercial tenant. Specifically, the reputation value of the merchant involved in the transaction is calculated by the following steps:
step 5.1: acquiring concrete information of merchants on the blockchain, and calculating a credit value of the merchant according to the concrete information of the merchants on the blockchain; the merchant features are divided into merchant fixed features and merchant business features, wherein the merchant fixed features comprise business ranges, merchant addresses and the like, and the merchant business features comprise credit card consumption count proportion, credit card consumption amount proportion, historical month consumption amount average value, historical abnormal transaction amount proportion, historical abnormal transaction count proportion and the like. And carrying out normalization processing on the data, wherein the historical month consumption amount average value is normalized according to the lowest and highest month consumption amounts.
Step 5.2: in order to ensure the robustness and generalization capability of the model, the credit value of the merchant is calculated based on the historical business characteristics and the current business characteristics of the merchant through an attention mechanism algorithm. The data is first vector transformed.
Step 5.2.1: the local activating unit is utilized to extract historical business characteristics of the merchant, and then the historical business characteristics are weighted and pooled, so that the credit value of the merchant related to the current transaction can be calculated adaptively, and the credit value is expressed by the following formula:
wherein { e } 1 ,e 2 ,...,e H Feature vector list representing historical business features of merchant of length H, v A A vector representing current transaction information. a is a feed forward network, and its output is the activation weight.
Step 5.2.2: the business characteristics of a merchant are selected as the attention signal.
Step 5.2.3: the input related sequence is encoded, and the weight w (x) corresponding to the attention signal is obtained by comparing the similarity function i T), namely calculating to obtain weights according to a scaling dot product model of scale transformation in the attention distribution model; the attention profile is calculated specifically by the following form:
wherein x is i The business feature of the ith merchant is input, t is the fixed feature vector of the ith merchant, W, U, v is a learnable network parameter, and d is the dimension of the input information. Scaling dot product models that can employ scaling herein Weight w (x i Normalization processing is carried out on t) to obtain the attention distribution f i ：
Step 5.2.4: it is converted into an attention weight according to the similarity. The obtained attention distribution value and the corresponding area are weighted and summed to obtain the attention value of the attention signalAs attentionAnd (5) weighting.
Where Vi refers to merchant i, which corresponds to the transaction characteristics of the merchant's historical transactions.
Step 5.2.5: and weighting the input related sequence through the weight to obtain a vector which is used as an input aggregation characteristic.
Wherein the weight refers to the attention weight of step 5.2.4; the input correlation sequence is the most initial feature vector of the merchant.
Step 5.3: and calculating to obtain the credit value of the merchant by using the self-adaptive activation function.
Based on a plurality of attention values obtained by a merchant, the aggregated feature is obtained through step 5.2.5, and is spliced and fused with other features (such as merchant fixed features) of the merchant, and then the reputation value of the merchant can be obtained through step 5.3.
Wherein the adaptive activation function is represented by the following formula:
i.e., f (w) =p (w) ·w+ (1-p (w))·αw; where w represents an input of an activation function f (·) such as a feature of the original merchant, a feature processed by an attention mechanism, p (w) =i (w > 0) for controlling f (w) to switch between f (w) =w and f (w) =αw.
Wherein E is w And V w Representing the mean and variance of each batch input, respectively, epsilon is a constant. P (w) is between 0 and 1; p (w) is a variable function introduced for controlling f (w) to switch between f (w) =w and f (w) =αw, and the conventional activation function α=0 cannot be adapted.
In this embodiment, the merchant feature corresponding to the merchant identifier is input into the pre-trained reputation prediction model, and the pre-trained reputation prediction model is utilized to identify the merchant feature corresponding to the merchant identifier, so as to obtain the merchant reputation parameter, which is beneficial to introducing the merchant reputation parameter when the transaction anomaly identification is performed subsequently, so that the transaction anomaly identification accuracy is improved.
In one embodiment, the step S103 identifies, according to the client characteristics corresponding to the client identifier, the client reputation parameter corresponding to the client identifier, which specifically includes the following contents: inputting the client characteristics corresponding to the client identifications into a pre-trained reputation prediction model, and extracting client transaction characteristics and client basic characteristics from the client characteristics through the pre-trained reputation prediction model; acquiring attention weights corresponding to the transaction characteristics of all clients, and carrying out fusion processing on the transaction characteristics of all clients according to the attention weights corresponding to the transaction characteristics of all clients to obtain target transaction characteristics corresponding to the client identifications; carrying out fusion processing on the target transaction characteristic corresponding to the client identifier and the client basic characteristic to obtain a fusion characteristic corresponding to the client identifier; and performing reputation prediction processing according to the fusion characteristics corresponding to the client identifications to obtain client reputation parameters corresponding to the client identifications.
The customer transaction characteristic is a customer consumption characteristic and is used for representing characteristic information related to customer consumption, and specifically comprises a current month credit card consumption count ratio, a current month credit card consumption amount ratio, a current year credit card consumption amount ratio, a credit card amount exceeding 1, a current year abnormal transaction amount ratio, an overdue repayment amount and the like. The client basic feature refers to a client fixed feature for characterizing feature information related to client attributes, and specifically includes client interests, client ages, and the like.
The attention weight corresponding to the customer transaction characteristic is used for measuring the importance corresponding to the customer transaction characteristic. The target transaction characteristics of the clients refer to characteristics obtained by fusion processing of the transaction characteristics of the clients.
The fusion characteristic corresponding to the client identifier refers to a splicing characteristic obtained after fusion processing is performed on the target transaction characteristic of the client and the basic characteristic of the client. Different fusion features correspond to different client reputation parameters.
Specifically, the server inputs the client features corresponding to the client identifications into a pre-trained reputation prediction model, performs feature extraction processing on the client features through the pre-trained reputation prediction model to obtain client transaction features and client basic features, acquires the attention distribution value of each client transaction feature, and calculates attention weights of each client transaction feature according to the attention distribution value of each client transaction feature; according to the attention weight corresponding to each customer transaction feature, carrying out pooling summation processing on each customer transaction feature to obtain a total customer transaction feature serving as a target transaction feature corresponding to the customer identifier; performing splicing processing on the target transaction characteristic corresponding to the client identifier and the client basic characteristic to obtain a spliced characteristic serving as a fusion characteristic corresponding to the client identifier; and determining the reputation parameter matched with the fusion feature as a client reputation parameter corresponding to the client identifier by utilizing the self-adaptive activation function.
For example, referring to fig. 6, the server first obtains the client information, and performs preprocessing (such as removing redundant data) on the client information to obtain preprocessed client information; performing feature extraction processing on the preprocessed customer information to obtain customer consumption features and customer fixed features; the attention weight of each customer consumption characteristic is obtained through an attention mechanism algorithm; and carrying out pooling summation treatment on each customer consumption characteristic according to the attention weight of each customer consumption characteristic to obtain a total customer consumption characteristic, carrying out splicing and smoothing treatment on the total customer consumption characteristic and a customer fixed characteristic to obtain a client splicing characteristic, and processing the client splicing characteristic by utilizing a self-adaptive activation function to obtain a client credit value. Specifically, the reputation value of the client related to the transaction is calculated by the following steps:
step 6.1: acquiring specific information of a client on a blockchain, and calculating a client credit value according to the specific information of the client on the chain; the customer features are divided into customer fixed features and customer consumption features.
Step 6.2: in order to ensure the robustness and generalization capability of the model, the reputation value of the client is calculated based on the historical consumption characteristics and the current consumption characteristics of the client through an attention mechanism algorithm. The data is first vector transformed.
Step 6.2.1: the local activating unit is utilized to extract the historical business characteristics of the clients, and then the historical business characteristics are weighted and pooled, so that the reputation value of the clients related to the current transaction can be calculated adaptively, and the reputation value is expressed by the following formula:
wherein { e } 1 ,e 2 ,...,e H Feature vector list representing historical consumer features of a customer of length H, v A A vector representing current transaction information. a is a feed forward network, and its output is the activation weight.
Step 6.2.2: the business characteristics of a customer are selected as the attention signal.
Step 6.2.3: the input related sequence is encoded, and the weight w (x) corresponding to the attention signal is obtained by comparing the similarity function i T), namely calculating to obtain weights according to a scaling dot product model of scale transformation in the attention distribution model; the attention profile is calculated specifically by the following form:
wherein x is i Representing the consumption characteristics of the i-th client, t representing the fixed characteristic vector of the i-th client, W, U, v being a learnable network parameter, d being the dimension of the input information. Scaling dot product models that can employ scaling hereinWeight w (x i Normalization processing is carried out on t) to obtain the attention distribution f i ：
Step 6.2.4: it is converted into an attention weight according to the similarity. The obtained attention distribution value and the corresponding area are weighted and summed to obtain the attention value of the attention signal As a weight for attention. />
Where Vi refers to customer i, which corresponds to the transaction characteristics of the customer's historical transactions.
Step 6.2.5: and weighting the input related sequence through the weight to obtain a vector which is used as an input aggregation characteristic.
Wherein the weight refers to the attention weight of step 6.2.4; the input correlation sequence is the customer's most initial feature vector.
Step 6.3: and calculating to obtain the reputation value of the client by using the self-adaptive activation function.
Based on a client, a plurality of attention values are obtained, an aggregate feature is obtained through step 6.2.5, and is spliced and fused with other features of the client (such as client fixed features), and then a client reputation value can be obtained through step 6.3.
Wherein the adaptive activation function is represented by the following formula:
i.e., f (w) =p (w) ·w+ (1-p (w))·αw; where w represents the input of an activation function f (·) such as the characteristics of the original customer, the characteristics processed by the attention mechanism, p (w) =i (w > 0) for controlling f (w) to switch between f (w) =w and f (w) =αw.
Wherein E is w And V w Respectively represent eachThe mean and variance of the batch input, ε, is a constant. P (w) is between 0 and 1; p (w) is a variable function introduced for controlling f (w) to switch between f (w) =w and f (w) =αw, and the conventional activation function α=0 cannot be adapted.
In this embodiment, the client features corresponding to the client identifier are input into the pre-trained reputation prediction model, and the pre-trained reputation prediction model is utilized to identify the merchant features corresponding to the client identifier, so as to obtain the client reputation parameter, which is beneficial to introducing the client reputation parameter when the transaction anomaly identification is performed subsequently, so that the transaction anomaly identification accuracy is improved.
In one embodiment, the pre-trained reputation prediction model is trained by: acquiring sample transaction information and an actual transaction identification result corresponding to the sample transaction information; sample trade information carries sample merchant identification and sample customer identification; acquiring sample merchant characteristics corresponding to the sample merchant identifications and sample customer characteristics corresponding to the sample customer identifications; respectively inputting the sample merchant characteristics and the sample client characteristics into a reputation prediction model to be trained to obtain sample merchant reputation parameters corresponding to sample merchant identifications and sample client reputation parameters corresponding to sample client identifications; performing full connection processing on the sample merchant reputation parameter and the sample customer reputation parameter to obtain a predicted transaction identification result corresponding to sample transaction information; training the reputation prediction model to be trained according to the difference between the predicted transaction recognition result and the actual transaction recognition result to obtain a trained reputation prediction model, and taking the trained reputation prediction model as a pre-trained reputation prediction model.
The sample transaction information comprises normal transaction information and abnormal transaction information; the actual transaction identification result is a normal transaction or an abnormal transaction. The predicted transaction recognition result is a normal transaction or an abnormal transaction.
The sample merchant identifier refers to a sample merchant name, and the sample customer identifier refers to a sample customer name. Sample merchant features may also be categorized into sample merchant transaction features and sample merchant fixed features, and sample customer features may also be categorized into sample customer transaction features and sample customer fixed features.
The sample merchant reputation parameter refers to a sample merchant reputation value, and the sample customer reputation parameter refers to a sample customer reputation value.
Specifically, the server acquires sample transaction information and an actual transaction identification result corresponding to the sample transaction information from the database; identifying sample transaction information to obtain a sample merchant identifier and a sample customer identifier carried by the sample transaction information; acquiring sample merchant features corresponding to the sample merchant identifications and sample customer features corresponding to the sample customer identifications from the blockchain; and inputting the sample merchant characteristics into the reputation prediction model to be trained to obtain sample merchant reputation parameters corresponding to the sample merchant identifications, and inputting the sample client characteristics into the reputation prediction model to be trained to obtain sample client reputation parameters corresponding to the sample client identifications. For example, the server performs feature extraction processing on the sample merchant features through a reputation prediction model to be trained to obtain sample merchant transaction features and sample merchant basic features, obtains attention distribution values of the sample merchant transaction features, and calculates attention weights of the sample merchant transaction features according to the attention distribution values of the sample merchant transaction features; according to the attention weight corresponding to the transaction characteristics of each sample merchant, carrying out pooling summation processing on the transaction characteristics of each sample merchant to obtain the transaction characteristics of the total sample merchant, wherein the transaction characteristics are used as target transaction characteristics corresponding to the sample merchant identifications; performing splicing processing on the target transaction characteristic corresponding to the sample merchant identifier and the sample merchant basic characteristic to obtain a spliced characteristic serving as a fusion characteristic corresponding to the sample merchant identifier; and determining the reputation parameter matched with the fusion feature by using the self-adaptive activation function as a sample merchant reputation parameter corresponding to the sample merchant identification.
Then, the server carries out full connection processing on the sample merchant reputation parameter and the sample customer reputation parameter to obtain a predicted transaction identification result corresponding to the sample transaction information; for example, the server calculates the probability that the sample transaction information belongs to the abnormal transaction according to the sample merchant reputation parameter and the sample customer reputation parameter, and if the probability is larger than the preset probability, the sample transaction information is judged to belong to the abnormal transaction; according to the difference between the predicted transaction recognition result and the actual transaction recognition result, combining a loss function, and calculating to obtain a first loss value; when the first loss value is larger than a first preset threshold value, the model parameters of the reputation prediction model to be trained are adjusted by using the first loss value, the training process is repeated, the reputation prediction model after the model parameters are adjusted is trained until the first loss value obtained according to the reputation prediction model after training is smaller than or equal to the first preset threshold value, the training is stopped, and the reputation prediction model after training is used as a reputation prediction model after training, so that a prestrain reputation prediction model is obtained.
For example, referring to the above examples, after step 5.3 or 6.3, step 5.4 or 6.4 is introduced: calculating the output result of the step 5.3 or 6.3 through a full connection layer to obtain a transaction validity prediction result, namely predicting whether the transaction is suspicious according to the merchant reputation value and the client reputation value, and judging that the transaction is suspicious if the reputation value is lower than a threshold value; and then returning a loss function based on the predicted result and the true value, and continuously correcting the weight and the predicted result of the reputation prediction model through the back propagation of the loss function, thereby obtaining the trained reputation prediction model.
In the embodiment, the reputation prediction model to be trained is trained for a plurality of times by utilizing sample transaction information, so that the reputation prediction model after training is obtained, the accuracy of merchant reputation parameters and client reputation parameters which are output by the reputation prediction model later is improved, the accuracy of transaction anomaly identification results obtained on the basis of the merchant reputation parameters and the client reputation parameters is further improved, and the transaction anomaly identification accuracy is further improved.
In one embodiment, the pre-trained transaction anomaly identification model is trained by: sample transaction information is obtained, and the sample transaction information is divided to obtain a training data set and a test data set; deleting abnormal data in the training data set by using the isolated forest model to obtain a target training data set; training the transaction anomaly recognition model to be trained according to the target training data set to obtain a trained transaction anomaly recognition model; and retraining the trained transaction anomaly recognition model according to the test data set to obtain a trained transaction anomaly recognition model which is used as a pre-trained transaction anomaly recognition model.
Wherein the sample transaction information includes normal transaction information and abnormal transaction information. The training data set is used for training the transaction abnormality recognition model, and the test data set is used for detecting the recognition accuracy of the transaction abnormality recognition model. It should be noted that, in the actual scenario, the ratio of the training data set to the test data set is 3:1.
wherein the outlier data in the training dataset is outlier data; the isolated forest model refers to a model corresponding to an isolated forest algorithm and is used for identifying and deleting outlier data in the training data set. The target training data set refers to a training data set after deleting abnormal data, that is, the target training data set includes normal data.
The trained transaction anomaly recognition model is retrained, and model parameters of the transaction anomaly recognition model are further optimized, so that recognition accuracy of the transaction anomaly recognition model is higher.
Specifically, the server acquires a plurality of sample transaction information from a financial database, and divides the plurality of sample transaction information to obtain a training data set and a test data set; identifying abnormal data in the training data set through the isolated forest model, deleting the abnormal data in the training data set to obtain a processed training data set as a target training data set; sample transaction information in the target training data set is input into a transaction anomaly identification model to be trained, and a predicted transaction anomaly identification result aiming at the sample transaction information is obtained; acquiring an actual transaction abnormality recognition result aiming at sample transaction information from a financial database, and calculating to obtain a second loss value according to the difference between the predicted transaction abnormality recognition result and the actual transaction abnormality recognition result and a loss function (such as a cross entropy loss function); when the second loss value is larger than a second preset threshold value, the model parameters of the transaction anomaly identification model to be trained are adjusted by using the second loss value, the training process is repeated, the transaction anomaly identification model with the adjusted model parameters is trained, and training is stopped until the second loss value obtained according to the trained transaction anomaly identification model is smaller than or equal to the second preset threshold value; inputting sample transaction information in the test data set into a trained transaction anomaly identification model to obtain a transaction anomaly identification result aiming at the sample transaction information, and confirming the identification accuracy of the trained transaction anomaly identification model according to the transaction anomaly identification result; continuously adjusting model parameters of the trained transaction anomaly identification model according to the identification accuracy rate until the identification accuracy rate of the adjusted transaction anomaly identification model is greater than the preset accuracy rate, stopping the test, and confirming the transaction anomaly identification model at the moment as a trained transaction anomaly identification model; and finally, taking the trained transaction abnormality recognition model as a pre-trained transaction abnormality recognition model.
For example, the server mainly utilizes a method of combining an isolated forest and a single-class support vector machine to identify transaction information in advance; the outliers in the training set are detected and removed in advance through an isolated forest algorithm, so that the influence of the outliers on the decision accuracy of the single-class support vector machine can be reduced; and then training a single-class support vector machine model based on normal data, and further improving the detection accuracy of the transaction anomaly identification model by combining feature selection and parameter optimization. The specific flow is shown in fig. 7, and mainly comprises the following steps:
step 7.1: acquiring historical abnormal transaction data, preprocessing the historical abnormal transaction data to obtain a training data set and a test data set, and selecting proper characteristic parameters; wherein the ratio between the training data set and the test data set is 3:1.
Step 7.2: and detecting and removing outliers in the training data set by using an isolated forest algorithm, and extracting normal data in the training data set to serve as training data, so that a training data set of a single-class support vector machine classification model is obtained.
Step 7.3: training a single classification detection model based on a training set after outliers are removed by an isolated forest algorithm, and then testing the detection effect of the trained model by using a test data set. And (3) carrying out parameter optimization on the model according to the evaluation index, and meanwhile, taking care to avoid overfitting. And after multiple parameter optimization and result comparison, selecting optimal model parameters, thereby obtaining a trained model.
For another example, the server may also build a transaction anomaly recognition model by pre-filtering merchants or customers for whom history of anomalous transactions has not been recorded.
Step one: the isolated forest is an isolated sample by utilizing a binary tree search structure, and consists of t binary trees. Because the density of the abnormal samples of the data set is low, the abnormal samples are displayed in isolation earlier in the binary tree structure, and the abnormal samples are close to the root node; in contrast, the normal samples would be farther from the root node. h (x) is the path length from the root node through the binary tree to the external node, and c (n) is the average of h (x) given n, used to normalize h (x). Outlier score s (x, n) is defined as s (x, n) =2 -E(h(x))/c(n) Wherein:
where H (i) is the sum of the sums estimated by ln (i) + 0.5772156649. S approaches 0.5 when E (h (x)) tends to c (n); s approaches 1 when E (h (x)) approaches 0; s approaches 0 when E (h (x)) tends to n-1; if s is close to 1, then x is outlier data.
Step two: let training samples d= { x 1 ,x 2 ,...,x i Solving a quadratic programming problem:
s.t.(ω·Φ(x n ))≥ρ-ξ n ,ξ n ≥0(n＝1,2,...,i)
wherein i is the size of set D; zeta type toy n A non-zero relaxation variable, penalty term added to avoid overfitting; v is a regularization parameter, typically v ε (0, 1). Assuming ω and ρ are solutions to the problem (i.e., the corresponding result of solving the quadratic programming problem to a minimum under constraint conditions), a decision function, i.e., a transaction anomaly identification model, can be constructed.
If the result obtained by the decision function is positive, the transaction data is normal, otherwise, the transaction data is abnormal, and the transaction data can be removed.
In this embodiment, the sample transaction information is divided to obtain a training data set and a test data set, the training data set is used for model training to obtain a trained transaction anomaly identification model, and then the test data set is used for retraining the trained transaction anomaly identification model to obtain a trained transaction anomaly identification model, so that the accuracy of the transaction anomaly identification result output by the trained transaction anomaly identification model is improved, and the transaction anomaly identification accuracy is further improved.
In one embodiment, as shown in fig. 8, after inputting the transaction information to be identified into a pre-trained transaction anomaly identification model to obtain a transaction anomaly identification result for the transaction information to be identified, the method further includes a step of updating the transaction anomaly identification result, specifically including the following steps:
step S801, in the case that the transaction anomaly recognition result is an anomaly transaction, determining a merchant transaction information set corresponding to a merchant identifier and a customer transaction information set corresponding to a customer identifier according to the transaction information to be recognized.
Step S802, a first association rule matched with a merchant transaction information set and a second association rule matched with a customer transaction information set are identified from a plurality of association rules stored in an abnormal transaction association rule library.
Step 803, screening out a first target association rule with the largest association degree from the first association rules, and taking the association degree corresponding to the first target association rule as the first association degree between the transaction information to be identified and the merchant identification.
Step S804, screening out a second target association rule with the largest corresponding association degree from the second association rules, and taking the association degree corresponding to the second target association rule as the second association degree between the transaction information to be identified and the client to which the client identifier belongs.
Step S805, updating the transaction anomaly identification result according to the first association degree and the second association degree to obtain a target transaction anomaly identification result aiming at the transaction information to be identified.
The merchant transaction information set refers to an information set formed by transaction information related to merchants in the transaction information to be identified; the client transaction information set refers to an information set formed by transaction information related to clients in the transaction information to be identified.
The abnormal transaction association rule base stores a plurality of association rules related to abnormal transactions, for example, for each transaction record of a merchant, the transaction amount is an integer of 100 and the transaction time can be an association rule from 11 pm to 7 pm. As shown in fig. 3, the server performs rule mining processing on the abnormal transaction information stored in the historical abnormal transaction information base, so that a plurality of association rules related to abnormal transactions can be obtained and stored in the abnormal transaction association rule base, and the abnormal transaction association rule base is convenient to identify the abnormal transactions in the follow-up process.
It should be noted that, each association rule is matched with a support degree, a confidence degree and an association degree, and the association degree=1-confidence degree; the confidence level is used for representing the probability that the current transaction is an abnormal transaction, and the association level is used for representing the association level between the current transaction and a merchant or a customer. That is, the smaller the association degree, the greater the probability that the transaction is abnormal.
The first association rule refers to an association rule in the abnormal transaction association rule base, wherein the matching degree between the association rule and the merchant transaction information set is greater than the first matching degree, that is, the first association rule may be multiple. The second association rule refers to an association rule with a matching degree with the client transaction information set greater than the second matching degree in the abnormal transaction association rule base, that is, the second association rule may be multiple.
The first target association rule refers to an association rule with the largest association degree corresponding to the first association rule; the second target association rule refers to an association rule with the largest association degree corresponding to the second association rule.
The first association degree between the transaction information to be identified and the merchant to which the merchant identification belongs is used for representing the association condition between the transaction information to be identified and the merchant, and if the first association degree is lower, the probability that the transaction information to be identified is abnormal transaction is higher, namely, the first association degree has a corresponding relationship with the abnormal transaction probability. In an actual scene, the first association degree refers to an association degree corresponding to a first target association rule.
The second association degree between the transaction information to be identified and the client to which the client identifier belongs is used for representing the association condition between the transaction information to be identified and the client, and if the second association degree is lower, the probability that the transaction information to be identified is abnormal transaction is higher, namely, the second association degree has a corresponding relationship with the abnormal transaction probability. In the actual scene, the second association degree refers to the association degree corresponding to the second target association rule.
It should be noted that, the first association degree and the second association degree may represent the probability that the transaction information to be identified is an abnormal transaction, so that in the case that the transaction abnormal identification result is an abnormal transaction, the server updates the transaction abnormal identification result according to the first association degree and the second association degree, so that the transaction abnormal identification accuracy can be further improved.
The target transaction anomaly recognition result refers to a final transaction anomaly recognition result.
Specifically, under the condition that the abnormal transaction identification result is abnormal transaction, the server identifies transaction information related to the merchant from the transaction information to be identified, and constructs a merchant transaction information set according to the transaction information related to the merchant; identifying transaction information related to a client from the transaction information to be identified, and constructing a client transaction information set according to the transaction information related to the client; identifying an association rule with a matching degree with the merchant transaction information set being greater than a first matching degree from a plurality of association rules stored in an abnormal transaction association rule library as a first association rule; identifying an association rule with the matching degree with the client transaction information set being larger than a second matching degree from a plurality of association rules stored in the abnormal transaction association rule library as a second association rule; according to the association degree corresponding to the first association rule, the association rule with the largest association degree is screened out from the first association rule to be used as a first target association rule, and the association degree corresponding to the first target association rule is used as the first association degree between the transaction information to be identified and the merchant to which the merchant identification belongs; and according to the association degree corresponding to the second association rule, screening the association rule with the largest association degree from the second association rule as a second target association rule, and taking the association degree corresponding to the second target association rule as a second association degree between the transaction information to be identified and the client to which the client identifier belongs. Finally, under the condition that the first association degree and the second association degree are smaller than the preset association degree, determining the transaction information to be identified as abnormal transaction, thereby confirming that the target transaction abnormal identification result is abnormal transaction; and otherwise, determining the transaction information to be identified as normal transaction, thereby confirming that the abnormal identification result of the target transaction is normal transaction.
For example, referring to fig. 3, after determining that the present transaction is an abnormal transaction through the transaction abnormal recognition model, the server calculates the association degree between the present transaction and the merchant and the association degree between the present transaction and the client by using the association rule stored in the abnormal transaction association rule library, determines the probability that the present transaction is an abnormal transaction by using the association degree between the present transaction and the merchant and the association degree between the present transaction and the client, and if the probability is greater than the preset probability, determines that the present transaction is an abnormal transaction, and stores the transaction information of the present transaction into the abnormal transaction information library; if the probability is smaller than or equal to the preset probability, confirming that the transaction is normal, and storing the transaction information of the transaction into a normal transaction information base.
It should be noted that, referring to fig. 3, the association rules stored in the abnormal transaction association rule base are mined from the historical abnormal transaction information base and the abnormal transaction information base. The specific digging process is as follows:
step 1: and (5) preprocessing data. For the data samples required by the process, the data samples are obtained from historical credit card abnormal transaction data in advance, wherein the data samples comprise business articles, unit price, legal age of merchants, unit price, total price, quantity and transaction time of credit card transaction articles, quantity of consumer credit cards, overdue payment amount and the like. Because of the redundant data, the data needs to be preprocessed, including missing value filling, type conversion, data integration and the like, and finally the attribute set { the transaction article quantity is 1 (a-no, b-yes), the total transaction amount is an integer of 100 (c-no, d-yes), the transaction time is from 11 pm to 7 am (e-no, f-yes), the transaction time period has no transaction (g-no, h-yes) which is not the client, the client has transaction (i-no, j-yes) in the historical transaction time period, the client has transaction (k-no, l-yes) in the time period, the number of the credit cards of the consumer exceeds 1 (m-no, n-yes), the consumption amount is within 1 standard deviation from the average value in the normal distribution of the transaction data of the merchant (o-no, p-yes), and the consumption amount is within 1 standard deviation from the average value in the normal distribution of the consumer consumption data of the merchant (q-no, r-yes).
Step 2: and (3) obtaining a data set by utilizing the related data information obtained after the data processing in the step one. And (3) carrying out equal conversion on the real-time records, and calculating the support degree of the corresponding clients and merchants of the transaction. The 9-dimensional feature dimension is high, so that features are extracted into two items according to the association between variables and the relation between a merchant and a cardholder, wherein the two items are { the number of transaction objects is 1, the total transaction amount is an integer of 100, the transaction time is in 11 pm to 7 am, no transaction is carried out by the customer in a transaction time period, the consumption amount is within 1 standard deviation from the average value in normal distribution of business data of the merchant }, the number of transaction objects is 1, the total transaction amount is an integer of 100, the transaction time is in 11 pm to 7 am, the transaction is carried out by the customer in a historical transaction time period, the number of credit cards of the customer exceeds 1 in the time period, and the consumption amount is within 1 standard deviation from the average value in normal distribution of consumption data of the customer }.
Step 3: and obtaining the highest association degree. And generating association rules through transaction data, obtaining the corresponding association degree and taking the highest value. If the association rule X-Y exists, the corresponding support and confidence are respectively as follows: the number of times/total number of records that an item within sup (x→y) = |x n|= { X, Y is present in a certain record at the same time }, the number of times/number of records that an item within cond (x→y) = |x N y|/|x|= { X, Y is present in a certain record at the same time/number of times X. The first implementation is as follows:
Step 3.1: at the minimum support sp min On the premise of using a recursion method to obtain a required frequent item set; initial candidate 1 set l 1 I.e. all data that has occurred; then pruning is carried out to remove the candidate item set with the support degree less than sp min Is a set of frequent k entries.
For example, if a merchant has a transaction data set of { ' adfhjkmoq ', ' acfgiknoq ', ' bcegilnor ', ' abcdefghijkllmnoqr }, a candidate 1 set { abcdefghijlmnoqr }, a support degree is calculated for each element in the set, for example, the corresponding support degree of element a is 2/3, the corresponding support degree of b is 1/3, and if 50% is used as the minimum support degree at this time, pruning is performed for element b. Taking this as an example, the final frequent 1-term set { acfgikoq } can be obtained. And connecting according to the frequent 1 item set to obtain a candidate 2 item set { ac, af, ag, ai, ak, ao, aq, cf, cg, ci, ck, & gt, oq }, and performing pruning on the candidate 2 item set in the same way to obtain the frequent 2 item set, and iterating pruning until only one item remains in the frequent item set.
Step 3.2: at the minimum confidence od is satisfied min Under the condition of (2), according to the frequent item set obtained in step 3.1, obtaining a corresponding association rule, for example, an integer with a transaction amount of 100 and a transaction time which can be an association rule between 11 pm and 7 pm.
Step 3.3: and (3) taking the processed data (i.e. the data processed in the steps 1 and 2) as input, and obtaining a corresponding confidence degree od through the association rule in the step 3.2, wherein the association degree of the transaction with a merchant or a client is 1-od, and the probability of abnormality exists in the transaction.
The second implementation is as follows:
at the minimum support sp min On the premise of using a recursion method to obtain a required frequent item set; assuming that a minimum support sp is defined min The maximum available set of five items with minimum support =1 (a), total transaction amount 100 integer (b), transaction time between 11 pm and 7 am (c), no transaction by the customer (d), and consumption amount within 1 standard deviation from the average in normal distribution of merchant business data (e).
At the minimum confidence od is satisfied min Under the condition of=50%, according to the frequent item set obtained in the above process, a corresponding association rule is obtained. The specific results are shown in the following table:
table 1 association rules
The rule states that: when the transaction amount is an integer of 100 and the transaction time is between 11 pm and 7 am, if the merchant has no transaction other than the customer in the transaction time period and the consumption amount is within 1 standard deviation from the average value in the normal distribution of merchant business data, the transaction may be abnormal transaction, and the association degree is ms=1-od max 。
In this embodiment, after the transaction information to be identified is input into the pre-trained transaction anomaly identification model to obtain the transaction anomaly identification result for the transaction information to be identified, the transaction anomaly identification result is updated by using the first association degree between the transaction information to be identified and the merchant to which the merchant identifier belongs and the second association degree between the transaction information to be identified and the customer to which the customer identifier belongs, so that the first association degree between the transaction information to be identified and the merchant to which the merchant identifier belongs and the second association degree between the transaction information to be identified and the customer to which the customer identifier belongs are comprehensively considered, thereby being beneficial to further improving the transaction anomaly identification accuracy.
In one embodiment, the step S805 updates the transaction anomaly identification result according to the first association degree and the second association degree to obtain a target transaction anomaly identification result for the transaction information to be identified, which specifically includes the following contents: comparing the first association degree with the second association degree with a preset association degree to obtain a comparison result; the comparison result is used for representing whether the transaction information to be identified is abnormal transaction or not; and updating the transaction abnormality recognition result according to the comparison result to obtain a target transaction abnormality recognition result aiming at the transaction information to be recognized.
The first association degree and the second association degree are smaller than the preset association degree, or the first association degree is smaller than the preset association degree, the second association degree is larger than or equal to the preset association degree, or the first association degree is larger than or equal to the preset association degree, and the second association degree is smaller than the preset association degree, so that the transaction information to be identified is abnormal transaction.
It should be noted that, according to the first association degree and the second association degree, the association degree between the transaction information to be identified and the merchant and the association degree between the transaction information to be identified and the customer are determined, if the association degree between the transaction information to be identified and the merchant and the association degree between the transaction information to be identified and the customer are smaller, the transaction information to be identified is not used for real requirements, and the transaction information to be identified is abnormal.
Specifically, the server compares the first association degree and the second association degree with a preset association degree to obtain a comparison result; if the comparison result is that the first association degree and the second association degree are smaller than the preset association degree, confirming that the transaction information to be identified is abnormal transaction, and confirming that the target transaction abnormal identification result is abnormal transaction.
Of course, the server may also confirm that the target transaction anomaly recognition result is an abnormal transaction when the first association degree is less than the preset association degree, the second association degree is greater than or equal to the preset association degree, or the first association degree is greater than or equal to the preset association degree, and the second association degree is less than the preset association degree.
In this embodiment, the transaction anomaly recognition result is updated according to the first association degree and the second association degree, so as to obtain a target transaction anomaly recognition result for the transaction information to be recognized, and on the basis of considering the merchant reputation parameter, the customer reputation parameter and the recognition result of the transaction anomaly recognition model, the association degree between the transaction to be recognized and the merchant and the customer is continuously considered, so that the determination accuracy of the transaction anomaly recognition result is improved, erroneous judgment is avoided, and the transaction anomaly recognition accuracy is further improved.
In one embodiment, as shown in fig. 9, there is provided still another transaction anomaly identification method, which is described by taking the application of the method to a server as an example, and includes the following steps:
step S901, obtaining transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier.
Step S902, acquiring a merchant feature corresponding to the merchant identifier and a customer feature corresponding to the customer identifier.
Step S903, identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant feature corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client feature corresponding to the client identifier.
Step S904, inputting the transaction information to be identified into a pre-trained transaction anomaly identification model to obtain a transaction anomaly identification result for the transaction information to be identified under the condition that the merchant reputation parameter and the client reputation parameter meet a preset reputation condition.
In step S905, in the case that the transaction anomaly recognition result is an anomaly transaction, a merchant transaction information set corresponding to the merchant identifier and a customer transaction information set corresponding to the customer identifier are determined according to the transaction information to be recognized.
Step S906, identifying a first association rule matching the merchant transaction information set and a second association rule matching the customer transaction information set from a plurality of association rules stored in the abnormal transaction association rule library.
Step S907, screening out a first target association rule with the largest corresponding association degree from the first association rules, and taking the association degree corresponding to the first target association rule as the first association degree between the transaction information to be identified and the merchant identification.
Step S908, screening out a second target association rule with the largest corresponding association degree from the second association rules, and taking the association degree corresponding to the second target association rule as a second association degree between the transaction information to be identified and the client to which the client identifier belongs.
Step S909, comparing the first association degree and the second association degree with a preset association degree to obtain a comparison result; the comparison result is used for representing whether the transaction information to be identified is an abnormal transaction or not.
Step S910, updating the transaction anomaly identification result according to the comparison result to obtain a target transaction anomaly identification result aiming at the transaction information to be identified.
In the transaction anomaly identification method provided in the above embodiment, when transaction anomaly identification is performed, firstly, according to the merchant characteristics corresponding to the merchant identifier carried by the transaction information to be identified and the client characteristics corresponding to the client identifier, identifying the merchant reputation parameter and the client reputation parameter, and under the condition that the merchant reputation parameter and the client reputation parameter meet the preset reputation condition, outputting the transaction anomaly identification result aiming at the transaction information to be identified by utilizing a pre-trained transaction anomaly identification model; the merchant reputation parameter, the customer reputation parameter and the transaction information to be identified are comprehensively considered, and the transaction anomaly identification model trained in advance is utilized for identification, so that manual verification is not needed, the determination accuracy of the transaction anomaly identification result is improved, and the transaction anomaly identification accuracy is improved.
In one embodiment, in order to more clearly clarify the transaction abnormality recognition method provided by the embodiment of the present application, a specific embodiment is described below specifically. In one embodiment, the application also provides a credit card transaction anomaly identification method based on the blockchain, as shown in fig. 3, the server calculates a merchant credit value and a client credit value by using the merchant characteristics corresponding to the merchant related to the current transaction information and the client characteristics corresponding to the client related to the current transaction information; under the condition that the credit value of the merchant and the credit value of the client are both larger than the preset credit value, judging that the current transaction information is normal transaction, and storing the current transaction information into a normal transaction information base; under the condition that the credit value of the merchant and the credit value of the client are smaller than the preset credit value, judging that the current transaction information is suspicious transaction, inputting the current transaction information into a transaction abnormality recognition model trained based on historical data, and recognizing the current transaction information through the transaction abnormality recognition model to judge whether the current transaction information is abnormal transaction or not; if the current transaction information is normal transaction, storing the current transaction information into a normal transaction information base; if the current transaction information is abnormal transaction, calculating the association degree between the current transaction and the merchant and the association degree between the current transaction and the customer by using association rules stored in an abnormal transaction association rule library, determining the probability that the current transaction is abnormal transaction by using the association degree between the current transaction and the merchant and the association degree between the current transaction and the customer, and if the probability is greater than the preset probability (namely, the association degree between the current transaction and the merchant and the association degree between the current transaction and the customer are smaller than the preset association degree), confirming that the current transaction is abnormal transaction and storing the transaction information of the current transaction into an abnormal transaction information library; if the probability is smaller than or equal to the preset probability, confirming that the transaction is normal, and storing the transaction information of the transaction into a normal transaction information base. Meanwhile, the server extracts association rules related to abnormal transactions from the historical abnormal transaction information base and the abnormal transaction information base, and stores the association rules into an abnormal transaction association rule base.
Further, as shown in fig. 2, the present application also provides an architecture diagram of a transaction blockchain network, where the blockchain network is composed of a plurality of credit card consumption channel nodes, a plurality of merchant nodes, and a plurality of banking nodes.
The block chain network is characterized in that a plurality of consumption channel nodes set in the block chain network are accessed as light nodes, and only user interaction, transaction information generation and request initiation operations are performed, so that block packaging and consensus are not involved.
The merchant node is responsible for receiving information broadcast in a consumption channel block chain, wherein the information comprises a unique consumption channel identifier of the whole network and one or more credit card transaction information recorded by the channel, and the credit card transaction information comprises a unique transaction time stamp, commodity information of the transaction, transaction amount of the transaction, customer information, bank information corresponding to the customer and merchant information.
Wherein the consumption channel node is responsible for receiving relevant information and consumption information of the customer transaction. The bank node is responsible for receiving a verification request (the verification request can comprise short message verification, biological identification verification and the like and is mainly used for verifying the authenticity of the identity of the customer) sent by the customer in the consumption channel block chain, acquiring information from a corresponding merchant based on the transaction request information, performing association calculation, credit value calculation and transaction legitimacy verification on the information, wherein the successful verification is that the merchant node is returned to be successful and the corresponding amount of money is paid to the merchant account, then performing block packing and whole network broadcast consensus on the related information of the transaction, and writing the information into a storage layer after the consensus is successful.
The application further provides a credit card transaction anomaly identification system based on the blockchain, which is arranged in a consumption channel node on a transaction blockchain network, wherein the consumption channel node is used as a light node to access the channel blockchain, and a main node of the network comprises a plurality of merchant nodes and a plurality of bank nodes. The system comprises: an encryption module and a communication module, wherein:
the encryption module is used for combining credit card transaction, client, merchant and other information with the unique network identifier of the consumption channel and the unique network transaction time stamp which is requested to be acquired in advance, and encrypting by adopting a contracted encryption algorithm to generate encryption information; for example, the transaction time stamp is 20230302192302, and information such as a network card, CPU (Central Processing Unit ) data, IP (Internet Protocol, protocol of interconnection between networks), city attribution and the like is added, and then a binary digital identifier with a length of 128 bits is generated through an encryption algorithm, which is basically unique in the effective case.
The communication module is used for broadcasting the encryption information generated by the encryption module in the consumption channel block chain on one hand, so that the corresponding merchant node can normally receive the broadcasted transaction information; and on the other hand, the method is used for initiating a request to a full-network public timestamp server of the consumption channel block chain to acquire the full-network unique transaction timestamp of the current transaction.
The application further provides another credit card transaction anomaly identification system based on the blockchain, which is arranged on a merchant node in a transaction blockchain network, and the merchant node is used as a master node to access a consumption channel blockchain to participate in operations such as block packing and consensus. The system comprises: communication module, authority management module, block generation module and consensus module, wherein:
the communication module is used for receiving the encrypted transaction information broadcast by all credit card transactions in the consumption channel block chain and the blocks broadcast by other nodes in the block chain on one hand; and on the other hand, the method is used for sending the block generated by the block generating module to a corresponding bank node in the block chain, so that the bank node can normally receive and process the transaction request.
The authority management module is used for decrypting the encrypted transaction information broadcast by all channel terminals in the received consumption channel block chain; and on the other hand, the method is used for setting a business data isolation area between a merchant on the block chain of the consumption channel and different bank nodes on the block chain. The business account book data in the business authority isolation area is only opened to the node users with corresponding authorities, and the node users without the authorities of the isolation area cannot access the business account book data of the isolation area.
And the block generation module is used for generating a block with a unique block ID according to the transaction result.
And the consensus module is used for carrying out consensus operation on the blocks broadcasted on the consumption channel block chain.
The application further provides a credit card transaction anomaly identification system based on the blockchain, which is arranged on a banking node in a transaction blockchain network, and the banking node is used as a master node to access a consumption channel blockchain and participate in operations such as block packing and consensus. The system comprises: the system comprises a reputation value calculation module, a relevance calculation module, a transaction validity prediction module, a communication module, a permission management module, a contract execution module, a block generation module and a consensus module, wherein:
and the reputation value calculation module is used for calculating reputation values corresponding to the clients and the merchant information corresponding to the transaction based on the historical transaction information of the clients and the merchant respectively, namely, the merchant reputation value and the client reputation value are obtained through self-adaptive learning of the self-adaptive activation function.
And the association degree calculating module is used for calculating association degrees of clients and merchants corresponding to the transaction based on the client information and the merchant information corresponding to the credit card transaction which actually occurs.
And the transaction validity prediction module is used for predicting the validity of the transaction so as to improve the accuracy of transaction abnormality prediction.
The communication module is used for receiving encrypted transaction information and blocks broadcast by other nodes in the consumption channel block chain on one hand; and on the other hand, the method is used for sending the block generated by the block generating module to other master nodes in the block chain, so that the other master nodes can normally share the block.
The authority management module is used for decrypting the encrypted transaction information broadcast by all the consumption channels in the received consumption channel block chain; and on the other hand, the method is used for setting a business data isolation area between a merchant on the block chain of the consumption channel and different bank nodes on the block chain. The business account book data in the business authority isolation area is only opened to the node users with corresponding authorities, and the node users without the authorities of the isolation area cannot access the business account book data of the isolation area.
And the contract execution module is used for executing the preset intelligent contract according to the obtained transaction information.
And the block generation module is used for generating a block with a unique block ID according to the transaction result.
And the consensus module is used for carrying out consensus operation on the blocks broadcasted on the consumption channel block chain.
As shown in fig. 10, the present application also provides a process flow diagram of a consumption channel node, which mainly includes the following steps:
step 10.1: after the consumption channel interacts with the customer, the customer information, the merchant information of the transaction and the bank information preset by the transaction customer and used for paying the amount are acquired.
Step 10.2: after successful interaction with the channel terminal, the consumer starts payment, and the consumer channel initiates a request to the whole network public timestamp server through an internal preset communication unit to acquire the whole network unique transaction timestamp of the transaction.
Step 10.3: the consumption channel combines the obtained consumption amount, transaction client information, bank information preset by the transaction client and the unique network identification information of the consumption channel, and the encryption information is obtained by encryption by using an encryption algorithm appointed by a merchant in advance.
Step 10.4: and sending the encrypted information obtained in the step 10.3 to a blockchain network for broadcasting.
Further, the application also provides a processing flow chart of the merchant node, which mainly comprises the following contents: the merchant node receives encrypted transaction information broadcast by a channel in a consumption channel block chain, wherein the information comprises the unique network identification information of the channel, the unique network transaction time stamp, the transaction information, the transaction amount, the transaction client information and the preset bank information of the client. The merchant node decrypts the received information based on a pre-agreed encryption algorithm to verify the legitimacy of the information. After verifying legal, the merchant node sends a payee transaction request to the corresponding bank node based on the transaction information recorded in the decrypted received information and the bank information set by the client.
The above embodiment can achieve the following technical effects: (1) trusted transaction: the non-tamperable technology, the consensus mechanism, the synchronization mechanism and the like of the blockchain technology ensure that each account book is kept consistent and safe, the transaction can be audited and traced, and the legal compliance of the transaction is ensured. And (2) accuracy and rapidness: based on the uplink data, obtaining the association degree of transaction information with the merchant and the customer, and obtaining the respective credit value according to the merchant and the customer information; and a attention mechanism is introduced, a classification algorithm is optimized, abnormal prediction is carried out on credit card transaction information, and prediction accuracy is improved. (3) data security: the credit card business data between the merchant and the bank are subjected to data isolation through a blockchain data isolation technology, and each blockchain user can only access the data of the isolation area with corresponding authority, so that the security and privacy of the business account book data are improved. The traffic data store is cryptographically protected and even if the blockchain direct access database is bypassed, no access to the traffic data is possible. (4) algorithm security: the encryption transmission is carried out by adopting the SM2, SM3 and SM4 algorithms, so that the security is higher.
It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiment of the application also provides a transaction abnormality recognition device for realizing the transaction abnormality recognition method. The implementation of the solution provided by the device is similar to the implementation described in the above method, so the specific limitation in the embodiments of the transaction anomaly identification device provided below may refer to the limitation of the transaction anomaly identification method hereinabove, and will not be described herein.
In one embodiment, as shown in fig. 11, there is provided a transaction anomaly identification device including: an information acquisition module 1110, a feature acquisition module 1120, a reputation identification module 1130, and a result determination module 1140, wherein:
an information acquisition module 1110, configured to acquire transaction information to be identified, which is broadcast in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier.
The feature acquisition module 1120 is configured to acquire a merchant feature corresponding to the merchant identifier and a customer feature corresponding to the customer identifier.
The reputation recognition module 1130 is configured to recognize a merchant reputation parameter corresponding to the merchant identifier according to the merchant feature corresponding to the merchant identifier, and recognize a client reputation parameter corresponding to the client identifier according to the client feature corresponding to the client identifier.
The result determining module 1140 is configured to input the transaction information to be identified into a pre-trained transaction anomaly identification model to obtain a transaction anomaly identification result for the transaction information to be identified when the merchant reputation parameter and the customer reputation parameter satisfy a preset reputation condition.
In one embodiment, the reputation recognition module 1130 is further configured to input the merchant feature corresponding to the merchant identifier into a pre-trained reputation prediction model, and extract the merchant transaction feature and the merchant basic feature from the merchant feature through the pre-trained reputation prediction model; acquiring attention weights corresponding to the transaction characteristics of all merchants, and carrying out fusion processing on the transaction characteristics of all merchants according to the attention weights corresponding to the transaction characteristics of all merchants to obtain target transaction characteristics corresponding to merchant identifications; carrying out fusion processing on the target transaction characteristic corresponding to the merchant identifier and the merchant basic characteristic to obtain a fusion characteristic corresponding to the merchant identifier; and performing reputation prediction processing according to the fusion characteristics corresponding to the merchant identifications to obtain merchant reputation parameters corresponding to the merchant identifications.
In one embodiment, the reputation recognition module 1130 is further configured to input a client feature corresponding to the client identifier into a pre-trained reputation prediction model, and extract the client transaction feature and the client basic feature from the client feature through the pre-trained reputation prediction model; acquiring attention weights corresponding to the transaction characteristics of all clients, and carrying out fusion processing on the transaction characteristics of all clients according to the attention weights corresponding to the transaction characteristics of all clients to obtain target transaction characteristics corresponding to the client identifications; carrying out fusion processing on the target transaction characteristic corresponding to the client identifier and the client basic characteristic to obtain a fusion characteristic corresponding to the client identifier; and performing reputation prediction processing according to the fusion characteristics corresponding to the client identifications to obtain client reputation parameters corresponding to the client identifications.
In one embodiment, the transaction anomaly identification device further comprises a first training module, configured to obtain sample transaction information and an actual transaction identification result corresponding to the sample transaction information; sample trade information carries sample merchant identification and sample customer identification; acquiring sample merchant characteristics corresponding to the sample merchant identifications and sample customer characteristics corresponding to the sample customer identifications; respectively inputting the sample merchant characteristics and the sample client characteristics into a reputation prediction model to be trained to obtain sample merchant reputation parameters corresponding to sample merchant identifications and sample client reputation parameters corresponding to sample client identifications; performing full connection processing on the sample merchant reputation parameter and the sample customer reputation parameter to obtain a predicted transaction identification result corresponding to sample transaction information; training the reputation prediction model to be trained according to the difference between the predicted transaction recognition result and the actual transaction recognition result to obtain a trained reputation prediction model, and taking the trained reputation prediction model as a pre-trained reputation prediction model.
In one embodiment, the transaction anomaly identification device further comprises a second training module, which is used for obtaining sample transaction information and dividing the sample transaction information to obtain a training data set and a test data set; deleting abnormal data in the training data set by using the isolated forest model to obtain a target training data set; training the transaction anomaly recognition model to be trained according to the target training data set to obtain a trained transaction anomaly recognition model; and retraining the trained transaction anomaly recognition model according to the test data set to obtain a trained transaction anomaly recognition model which is used as a pre-trained transaction anomaly recognition model.
In one embodiment, the transaction anomaly identification device further comprises a result updating module, configured to determine, according to the transaction information to be identified, a merchant transaction information set corresponding to the merchant identifier and a customer transaction information set corresponding to the customer identifier, in the case that the transaction anomaly identification result is an anomaly transaction; identifying a first association rule matched with the merchant transaction information set and a second association rule matched with the customer transaction information set from a plurality of association rules stored in an abnormal transaction association rule library; screening a first target association rule with the largest corresponding association degree from the first association rules, and taking the association degree corresponding to the first target association rule as the first association degree between the transaction information to be identified and the merchant to which the merchant identification belongs; screening a second target association rule with the largest corresponding association degree from the second association rules, and taking the association degree corresponding to the second target association rule as a second association degree between the transaction information to be identified and the client to which the client identifier belongs; and updating the transaction anomaly identification result according to the first association degree and the second association degree to obtain a target transaction anomaly identification result aiming at the transaction information to be identified.
In one embodiment, the result updating module is further configured to compare the first association degree and the second association degree with a preset association degree to obtain a comparison result; the comparison result is used for representing whether the transaction information to be identified is abnormal transaction or not; and updating the transaction abnormality recognition result according to the comparison result to obtain a target transaction abnormality recognition result aiming at the transaction information to be recognized.
The respective modules in the transaction abnormality recognition device described above may be implemented in whole or in part by software, hardware, or a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 12. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used for storing data of merchant characteristics, customer characteristics and the like. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by a processor, implements a transaction anomaly identification method.
It will be appreciated by those skilled in the art that the structure shown in FIG. 12 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
In an embodiment, there is also provided a computer device comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, carries out the steps of the method embodiments described above.
In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method embodiments described above.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims (11)

1. A transaction anomaly identification method, the method comprising:
acquiring transaction information to be identified, which is broadcasted in a blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
acquiring merchant characteristics corresponding to the merchant identifications and customer characteristics corresponding to the customer identifications;
identifying a merchant reputation parameter corresponding to the merchant identifier according to the merchant characteristic corresponding to the merchant identifier, and identifying a client reputation parameter corresponding to the client identifier according to the client characteristic corresponding to the client identifier;
And under the condition that the merchant credit parameter and the client credit parameter meet the preset credit condition, inputting the transaction information to be identified into a pre-trained transaction abnormality identification model to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
2. The method of claim 1, wherein the identifying a merchant reputation parameter corresponding to the merchant identification based on a merchant characteristic corresponding to the merchant identification comprises:
inputting the merchant characteristics corresponding to the merchant identification into a pre-trained reputation prediction model, and extracting merchant transaction characteristics and merchant basic characteristics from the merchant characteristics through the pre-trained reputation prediction model;
acquiring attention weights corresponding to the transaction characteristics of all the merchants, and carrying out fusion processing on the transaction characteristics of all the merchants according to the attention weights corresponding to the transaction characteristics of all the merchants to obtain target transaction characteristics corresponding to the merchant identifications;
carrying out fusion processing on the target transaction characteristics corresponding to the merchant identifications and the merchant basic characteristics to obtain fusion characteristics corresponding to the merchant identifications;
And performing reputation prediction processing according to the fusion characteristics corresponding to the merchant identifications to obtain merchant reputation parameters corresponding to the merchant identifications.
3. The method of claim 1, wherein the identifying a client reputation parameter corresponding to the client identification based on a client characteristic corresponding to the client identification comprises:
inputting the client characteristics corresponding to the client identifications into a pre-trained reputation prediction model, and extracting client transaction characteristics and client basic characteristics from the client characteristics through the pre-trained reputation prediction model;
acquiring attention weights corresponding to the client transaction characteristics, and carrying out fusion processing on the client transaction characteristics according to the attention weights corresponding to the client transaction characteristics to obtain target transaction characteristics corresponding to the client identifications;
carrying out fusion processing on the target transaction characteristic corresponding to the client identifier and the client basic characteristic to obtain a fusion characteristic corresponding to the client identifier;
and performing reputation prediction processing according to the fusion characteristics corresponding to the client identifications to obtain client reputation parameters corresponding to the client identifications.
4. A method according to claim 2 or 3, wherein the pre-trained reputation prediction model is trained by:
acquiring sample transaction information and an actual transaction identification result corresponding to the sample transaction information; sample trade information carries sample merchant identification and sample customer identification;
acquiring sample merchant characteristics corresponding to the sample merchant identifications and sample customer characteristics corresponding to the sample customer identifications;
respectively inputting the sample merchant characteristics and the sample client characteristics into a reputation prediction model to be trained to obtain sample merchant reputation parameters corresponding to the sample merchant identifications and sample client reputation parameters corresponding to the sample client identifications;
performing full connection processing on the sample merchant reputation parameter and the sample customer reputation parameter to obtain a predicted transaction identification result corresponding to the sample transaction information;
and training the reputation prediction model to be trained according to the difference between the predicted transaction recognition result and the actual transaction recognition result to obtain a trained reputation prediction model, and taking the trained reputation prediction model as the pre-trained reputation prediction model.
5. The method of claim 1, wherein the pre-trained transaction anomaly identification model is trained by:
sample transaction information is obtained, and the sample transaction information is divided to obtain a training data set and a test data set;
deleting abnormal data in the training data set by using an isolated forest model to obtain a target training data set;
training the transaction anomaly recognition model to be trained according to the target training data set to obtain a trained transaction anomaly recognition model;
and retraining the trained transaction anomaly recognition model according to the test data set to obtain a trained transaction anomaly recognition model serving as the pre-trained transaction anomaly recognition model.
6. The method according to claim 1, further comprising, after inputting the transaction information to be identified into a pre-trained transaction anomaly recognition model to obtain a transaction anomaly recognition result for the transaction information to be identified:
under the condition that the transaction abnormality identification result is abnormal transaction, determining a merchant transaction information set corresponding to the merchant identification and a customer transaction information set corresponding to the customer identification according to the transaction information to be identified;
Identifying a first association rule matched with the merchant transaction information set and a second association rule matched with the customer transaction information set from a plurality of association rules stored in an abnormal transaction association rule library;
screening a first target association rule with the largest corresponding association degree from the first association rules, and taking the association degree corresponding to the first target association rule as the first association degree between the transaction information to be identified and the merchant of which the merchant identification belongs;
screening a second target association rule with the largest corresponding association degree from the second association rules, and taking the association degree corresponding to the second target association rule as a second association degree between the transaction information to be identified and the client to which the client identifier belongs;
and updating the transaction anomaly identification result according to the first association degree and the second association degree to obtain a target transaction anomaly identification result aiming at the transaction information to be identified.
7. The method of claim 6, wherein updating the transaction anomaly identification result according to the first degree of association and the second degree of association to obtain a target transaction anomaly identification result for the transaction information to be identified comprises:
Comparing the first association degree with the second association degree with a preset association degree to obtain a comparison result; the comparison result is used for representing whether the transaction information to be identified is abnormal transaction or not;
and updating the transaction abnormality identification result according to the comparison result to obtain a target transaction abnormality identification result aiming at the transaction information to be identified.
8. A transaction anomaly identification device, the device comprising:
the information acquisition module is used for acquiring transaction information to be identified, which is broadcasted in the blockchain network; the transaction information to be identified carries a merchant identifier and a customer identifier;
the feature acquisition module is used for acquiring the merchant features corresponding to the merchant identifications and the customer features corresponding to the customer identifications;
the credit recognition module is used for recognizing merchant credit parameters corresponding to the merchant identifications according to the merchant characteristics corresponding to the merchant identifications and recognizing client credit parameters corresponding to the client identifications according to the client characteristics corresponding to the client identifications;
and the result determining module is used for inputting the transaction information to be identified into a pre-trained transaction abnormality identification model under the condition that the merchant reputation parameter and the client reputation parameter meet the preset reputation condition to obtain a transaction abnormality identification result aiming at the transaction information to be identified.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
11. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
CN202310648422.9A 2023-06-02 2023-06-02 Transaction abnormality identification method, apparatus, computer device, and storage medium Pending CN116611895A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310648422.9A CN116611895A (en) 2023-06-02 2023-06-02 Transaction abnormality identification method, apparatus, computer device, and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310648422.9A CN116611895A (en) 2023-06-02 2023-06-02 Transaction abnormality identification method, apparatus, computer device, and storage medium

Publications (1)

Publication Number Publication Date
CN116611895A true CN116611895A (en) 2023-08-18

Family

ID=87681662

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310648422.9A Pending CN116611895A (en) 2023-06-02 2023-06-02 Transaction abnormality identification method, apparatus, computer device, and storage medium

Country Status (1)

Country Link
CN (1) CN116611895A (en)

Similar Documents

Publication Publication Date Title
Benchaji et al. Enhanced credit card fraud detection based on attention mechanism and LSTM deep model
CN108876133A (en) Risk assessment processing method, device, server and medium based on business information
CN116235183A (en) Neural network architecture for transaction data processing
CN113011646B (en) Data processing method, device and readable storage medium
Arora et al. Facilitating user authorization from imbalanced data logs of credit cards using artificial intelligence
US20210174367A1 (en) System and method including accurate scoring and response
CN116307671A (en) Risk early warning method, risk early warning device, computer equipment and storage medium
US20210089899A1 (en) Federated learning system and method for detecting financial crime behavior across participating entities
O’Kane Detecting patterns in the Ethereum transactional data using unsupervised learning
US11694208B2 (en) Self learning machine learning transaction scores adjustment via normalization thereof accounting for underlying transaction score bases relating to an occurrence of fraud in a transaction
Pandey et al. A review of credit card fraud detection techniques
CN116611895A (en) Transaction abnormality identification method, apparatus, computer device, and storage medium
Sulayman et al. Designing security user profiles via anomaly detection for user authentication
Sherly et al. A improved incremental and interactive frequent pattern mining techniques for market basket analysis and fraud detection in distributed and parallel systems
Zhao et al. Deep federated learning hybrid optimization model based on encrypted aligned data
Adebayo et al. Comparative Review of Credit Card Fraud Detection using Machine Learning and Concept Drift Techniques
Akinje et al. Fraudulent Detection Model Using Machine Learning Techniques for Unstructured Supplementary Service Data
US11989276B2 (en) Intelligent authentication of users in Metaverse leveraging non-fungible tokens and behavior analysis
Leppla et al. Fraud pattern detection for NFT markets
CN117349900B (en) Intelligent prediction system and prediction method based on data mining
Perwej A Potent Technique for Identifying Fake Accounts on Social Platforms
US11270230B1 (en) Self learning machine learning transaction scores adjustment via normalization thereof
Abhinav et al. Bitcoin Price Prediction Using Deep Learning
Ikeda A new feature engineering framework for financial cyber fraud detection using machine learning and deep learning
Herradura et al. Trend Analyses for Blockchain Technology Innovations Using Data Analytics

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