CN115545712A - Fraud prediction method, device, equipment and storage medium for transaction behaviors - Google Patents
Fraud prediction method, device, equipment and storage medium for transaction behaviors Download PDFInfo
- Publication number
- CN115545712A CN115545712A CN202211158596.9A CN202211158596A CN115545712A CN 115545712 A CN115545712 A CN 115545712A CN 202211158596 A CN202211158596 A CN 202211158596A CN 115545712 A CN115545712 A CN 115545712A
- Authority
- CN
- China
- Prior art keywords
- fraud
- transaction
- current transaction
- matrix
- behavior
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
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
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/38—Payment protocols; Details thereof
- G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
- G06Q20/401—Transaction verification
- G06Q20/4016—Transaction verification involving fraud or risk level assessment in transaction processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Theoretical Computer Science (AREA)
- Finance (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Technology Law (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Security & Cryptography (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a fraud prediction method, a fraud prediction device, fraud prediction equipment and a storage medium for transaction behaviors. The method comprises the following steps: acquiring current transaction flow data corresponding to current transaction behaviors; performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior; inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types; and determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model. The technical scheme of the invention can realize automatic fraud prediction of transaction behaviors without manual participation, and improve the accuracy and efficiency of fraud prediction of transaction behaviors.
Description
Technical Field
The present invention relates to the field of computer technologies, and in particular, to a fraud prediction method, apparatus, device, and storage medium for transaction behavior.
Background
With the development of computer technology, the transaction behavior is more and more diversified, and accordingly, the fraud form of the transaction behavior is more and more diversified.
Currently, fraud risk prediction of transaction behavior through human experience is often required. And directly allowing the transaction for the transaction behavior predicted to have a lower fraud risk degree, performing key check and check on the transaction behavior predicted to have a higher fraud risk degree, and allowing the transaction to be performed after the security is confirmed through the check.
However, the fraud risk prediction mode based on manual experience is time-consuming and labor-consuming, misjudgment may occur due to subjective factors, accuracy of fraud risk prediction results cannot be effectively guaranteed, and fraud risk prediction efficiency is reduced.
Disclosure of Invention
The invention provides a transaction behavior fraud prediction method, a transaction behavior fraud prediction device, transaction behavior fraud prediction equipment and a storage medium, so that automatic fraud prediction of transaction behaviors is realized without manual participation, and the accuracy and efficiency of transaction behavior fraud prediction are improved.
According to an aspect of the invention, there is provided a fraud prediction method for transaction behaviour, the method comprising:
acquiring current transaction flow data corresponding to current transaction behaviors;
performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior;
inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types;
and determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type.
According to another aspect of the present invention, there is provided a fraud prediction apparatus for transaction behaviour, the apparatus comprising:
the current transaction flow data acquisition module is used for acquiring current transaction flow data corresponding to the current transaction behavior;
the current transaction characteristic matrix construction module is used for extracting the characteristics of the current transaction running data and constructing a current transaction characteristic matrix corresponding to the current transaction behavior;
the fraud prediction module is used for inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to perform fraud prediction of a current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types;
and the fraud prediction result determining module is used for determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type.
According to another aspect of the present invention, there is provided an electronic apparatus including:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,
the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform a method of fraud prediction of transaction behaviour according to any embodiment of the invention.
According to another aspect of the present invention, there is provided a computer-readable storage medium having stored thereon computer instructions for causing a processor to perform a method for fraud prediction of transaction behaviour according to any of the embodiments of the present invention.
According to the technical scheme of the embodiment of the invention, the current transaction running data corresponding to the current transaction behavior is obtained; performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior; inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types; determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type; according to the technical scheme of the embodiment of the invention, the current transaction characteristic matrix and the target transaction characteristic matrix can be input into the trained twin convolutional neural network model, the fraud probability of the current transaction behavior is determined based on the output result, the possibility that the current transaction behavior belongs to the fraud transaction behavior can be determined through the fraud probability, and the specific fraud type of the current transaction behavior can be judged according to the position distributed by the maximum probability value, so that the actual fraud type and the twin convolutional neural network model are closely combined, the interpretability of the fraud prediction result output by the twin convolutional neural network model is increased, the automatic fraud prediction of the transaction behavior is further realized, manual participation is not needed, and the accuracy and efficiency of the fraud prediction of the transaction behavior are improved.
It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a flowchart of a fraud prediction method for transaction behavior according to an embodiment of the present invention;
fig. 2 is a flowchart of a fraud prediction method for transaction behavior according to a second embodiment of the present invention;
FIG. 3 is a diagram of a twin convolutional neural network model according to a second embodiment of the present invention;
fig. 4 is a schematic structural diagram of a fraud prediction apparatus for transaction behavior according to a third embodiment of the present invention;
fig. 5 is a schematic structural diagram of an electronic device implementing the fraud prediction method for transaction behavior according to the embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Moreover, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Example one
Fig. 1 is a flowchart of a fraud prediction method for transaction behaviors, which is applicable to a situation where fraud prediction is performed on a generated transaction behavior according to an embodiment of the present invention, and the method may be performed by a fraud prediction apparatus for transaction behaviors, which may be implemented in hardware and/or software, and the fraud prediction apparatus for transaction behaviors may be configured in an electronic device. As shown in fig. 1, the method includes:
and S110, acquiring current transaction running data corresponding to the current transaction behavior.
The transaction behavior may refer to a behavior of a buyer and a seller performing a transaction on valuable goods and services, and is used to replace a thing with another thing. The current transaction behavior may refer to a transaction behavior that a transaction application is proposed at this time, but the transaction behavior is not completed and fraud prediction is to be performed. Transaction pipeline data may refer to transaction data that includes a variety of transaction characteristics. For example, the banking transaction flow data may refer to a deposit and withdrawal transaction record of a banking account (including a bankbook and a bank card), wherein the deposit and withdrawal transaction record includes data of transaction time, transaction date and the like. The current transaction flow data may refer to transaction flow data corresponding to a current transaction activity.
Specifically, the user wants to implement the current transaction behavior through the credit card, and the transaction information required by the current transaction behavior needs to be filled in on the front-end transaction interface of the bank transaction processing server. The back end of the bank transaction processing server can acquire current transaction running data corresponding to the current transaction behavior before the transaction is completed based on the transaction information filled by the user and information generated by submitting the transaction behavior, such as transaction date, transaction time and the like.
Illustratively, S110 may include: and acquiring current transaction flow time sequence data corresponding to each transaction behavior generated by the user in the current time period.
The current time period may refer to a time period formed by the current transaction time and a period of historical time before the current transaction time. For example, the current time period may refer to the last month and include the current time of day. There may be a plurality of transaction activities that the user has completed and a current transaction activity during the current time period. The transaction flow time series data may be transaction flow data arranged in a time series. For example, the transaction flow timing data may be transaction flow data arranged from far to near timing from the current time. The current transaction flow timing data may refer to each transaction activity generated during the current time period and arranged according to a certain timing. The current transaction flow sequence data may include current transaction flow data corresponding to current transaction behaviors generated by the user at the current time and historical transaction flow data corresponding to each completed transaction behavior of the user in the current time period.
Specifically, the user wants to implement the current transaction behavior through the credit card, and the transaction information required by the current transaction behavior needs to be filled in on the front-end transaction interface of the bank transaction processing server. The back end of the bank transaction processing server can acquire current transaction flow time sequence data corresponding to each transaction behavior generated by the user in the current time period based on a transaction record acquisition protocol agreed by the user, and suspend the previous transaction behavior, so that the property of the current transaction behavior can be predicted according to each transaction behavior generated by the user in the current time period.
And S120, extracting the characteristics of the current transaction flow data, and constructing a current transaction characteristic matrix corresponding to the current transaction behavior.
The characteristics can refer to the characteristics of something different from other things, and are used for distinguishing data with each representative meaning in the current transaction flow data. For example, the characteristic may be, but is not limited to, industry type, order number, order date, credit card type, line of transaction city code, merchant zip code, merchant number, merchant address.
Exemplarily, feature extraction is performed on current transaction flow time sequence data, normalization processing is performed on each extracted feature data, and a current transaction feature matrix corresponding to a current transaction behavior is constructed, wherein the row number of the current transaction feature matrix represents a time dimension, and the column number represents a feature dimension.
In particular, feature extraction may be performed on current transaction chronological data to which the user is involved. For each transaction flow, each feature data, such as an industry type, an order number, an order date, a credit card type, a transaction line city code, a merchant zip code, a merchant number, a merchant address and other feature data can be extracted, and the extracted feature data is normalized. For date type data, the number of days from a certain fixed year to a specified date can be calculated and based on the following formula
The days are converted to floating point numbers. The above formula can also be used to convert to a number between 0 and 1 (including 0 and 1) for the amount class data. For example, if the transaction amounts are 3 yuan for the first day, 5 yuan for the second day, 1 yuan for the third day, 4 yuan for the fourth day, and 2 yuan for the current transaction amount, the normalized data corresponding to the first day amount after conversion by the formula is (3-1)/(5-1) =0.5; the normalized data corresponding to the money amount on the second day is 1; the normalized data corresponding to the third day money amount is 0; the normalized data corresponding to the amount on the fourth day is 0.75; the normalized data corresponding to the current transaction amount is 0.25. Aiming at non-quantitative data such as industry types and merchant addresses, the non-quantitative data can be converted into data from characters and identifications in a binary coding mode. For example, codes can be preset in a binary coding mode for jurisdictions such as provinces, cities, districts, counties, towns and the like from large to small, so that the jurisdictions can be distinguished conveniently. If the user never remits the money to the outside, but the current transaction behavior is the money to the outside and the amount is huge, the user can temporarily determine the abnormal transaction and carry out detailed investigation on the current transaction behavior, thereby avoiding the user from being induced to remit the money by people. After all data are normalized by the characteristic data, the normalized data are used for constructing a current transaction characteristic matrix A corresponding to the current transaction behavior m×n . The row number m of the current transaction feature matrix represents a time dimension, and the column number n represents a feature dimension. Feature matrix A for current transaction m×n ，{A 1n ，A 2n ，…，A mn May represent current transaction flow characteristics; the remaining columns of the current transaction characteristic matrix may represent transaction running characteristics that have been completed within the current time period.
It should be noted that, a convolution network model can be used to extract the features of the current transaction running data, so that the advantage of strong capability of convolution network to extract high-dimensional features is fully utilized, and the feature extraction efficiency is further improved.
And S130, inputting the current transaction characteristic matrix and the target transaction characteristic matrix into the twin convolutional neural network model after training is finished to predict the fraud of the current transaction behavior.
The target transaction characteristic matrix is a transaction characteristic matrix which is constructed in advance based on normal transaction behavior flow data and fraud transaction behavior flow data corresponding to a plurality of fraud types. The twin Convolutional Neural Network (SCNN), also called as bilateral Convolutional Neural Network, is composed of two Convolutional networks, generally two groups of data are input, and spatial features of high latitude in the two groups of data are output and judged through overall Network feature calculation, so that the similarity degree of the two groups of data is compared through the mode.
For example, the step S130 of "constructing the target transaction characteristic matrix based on the normal transaction flow data and the fraud transaction flow data corresponding to the fraud types" may include: performing feature extraction on the normal transaction behavior flow data to construct at least one normal transaction feature matrix; performing feature extraction on at least one fraud transaction behavior flow data corresponding to each fraud type to construct at least one fraud transaction feature matrix corresponding to each fraud type; and performing matrix splicing processing on the at least one normal transaction characteristic matrix and the at least one fraud transaction characteristic matrix corresponding to each fraud type to construct a target transaction characteristic matrix.
Specifically, normal transaction behavior running data and fraudulent transaction behavior running data corresponding to a plurality of fraud types can be counted in a plurality of ways, such as manual counting, and reference is made to a m×n The construction method of (1) utilizes the statistical data to construct a target transaction characteristic matrix B. If h fraudulent transaction behaviors are pipelined data, a fraudulent transaction matrix B can be constructed 1 Is composed of
If there is (k)-h) normal transaction behavior pipeline data, a normal transaction matrix B can be constructed 2 Is composed of
B 1 And B 2 Together forming a target transaction feature matrix B.
The current transaction characteristic matrix and the constructed target transaction characteristic matrix can be input into the twin convolutional neural network model after the training is finished, and fraud prediction of the current transaction behavior is carried out through the twin convolutional neural network model.
In this embodiment, the twin convolutional neural network model is obtained by training in advance based on the current sample transaction feature matrix, the target sample transaction feature matrix, and the sample fraud tags corresponding to the current sample transaction feature matrix. Wherein the sample fraud tag may be a fraud prediction result obtained by processing based on a plurality of fraud prediction modes. For example, a sample fraud tag may be a tag G = [0,0, …,1, …,0,0 designed according to the target transaction feature matrix B and following One-Hot encoding rules]Or the fraud prediction result obtained after the transaction behavior is processed in a mode of artificially determining whether the transaction behavior has the fraud behavior is used, so that the obtained sample fraud label is an accurate fraud prediction result corresponding to a fraud prediction stage, and the fraud prediction result of the twin convolutional neural network model is further improved. The One-Hot encoding may refer to effective encoding, and the One-Hot encoding rule may be that the same-bit state registers are used to encode the states of the same bits, each state has a unique corresponding register state representation, and only One register is effective at any time. The position of 1 in G is A m×n And inputting the twin convolutional neural network model to obtain the rules needing to be hit.
The training process of the twin convolutional neural network model can be as follows: performing feature extraction on current sample transaction flow data corresponding to current sample transaction behaviors to constructAnd (4) obtaining a current sample transaction characteristic matrix corresponding to the current sample transaction behavior, and determining a sample fraud label corresponding to the current sample transaction behavior. And constructing a target sample transaction characteristic matrix based on the normal transaction behavior flow data and the fraud transaction behavior flow data corresponding to the plurality of fraud types. Inputting the current sample transaction characteristic matrix and the target sample transaction characteristic matrix into a twin convolutional neural network model to be trained to obtain a fraud prediction result output by the twin convolutional neural network model to be trained, and the fraud prediction result can be based on a training function
Determining a fraud prediction result and a sample fraud label, determining a training error, namely a cross entropy loss value, reversely transmitting the training error to a twin convolutional neural network model to be trained, adjusting model parameters in the twin convolutional neural network model until a preset convergence condition is met, such as the iteration times reach a preset number or the training error converges, determining that the training of the twin convolutional neural network model is finished, and at the moment, taking the trained twin convolutional neural network model as the twin convolutional neural network model corresponding to a fraud prediction stage, so that the trained twin convolutional neural network model can have the prediction capability of predicting various fraud types, and the fraud behavior can be predicted more accurately by using the twin convolutional neural network model.
It should be noted that, when a new fraud type occurs, the target transaction feature matrix may be updated and iterated based on the new fraud transaction behavior pipelining data of the fraud type; meanwhile, the twin convolutional neural network model also needs to be retrained, so that the twin convolutional neural network model can have the capability of predicting a newly-appeared fraud type, the twin convolutional neural network model can be updated and iterated in time, the training and the prediction are carried out simultaneously, and the learning capability of the network is enhanced.
And S140, determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model.
The fraud prediction result may include whether the current transaction behavior is normal behavior or fraud behavior. The fraud prediction result includes the predicted target fraud type. When the fraud prediction result is fraud, a specific target fraud type can also be predicted.
Specifically, the output result of the twin convolutional neural network model is a fraud probability matrix X = [ X ] 1 ，x 2 ，x 3 ，…，x l ]And analyzing the fraud probability matrix X based on the arrangement sequence of the normal behaviors in the target transaction characteristic matrix B input into the twin convolutional neural network model and the fraud behaviors of each fraud type to obtain the target type corresponding to the maximum probability value in the fraud probability matrix X, namely a fraud prediction result. And the specific fraud type is judged according to the position distributed by the maximum probability value, so that the interpretability is enhanced.
According to the technical scheme of the embodiment of the invention, the current transaction running data corresponding to the current transaction behavior is obtained; performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior; inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types; determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type; according to the technical scheme of the embodiment of the invention, the current transaction characteristic matrix and the target transaction characteristic matrix can be input into the trained twin convolutional neural network model, the fraud probability of the current transaction behavior is determined based on the output result, the possibility that the current transaction behavior belongs to the fraud transaction behavior can be determined through the fraud probability, and the specific fraud type of the current transaction behavior can be judged according to the position distributed by the maximum probability value, so that the actual fraud type and the twin convolutional neural network model are closely combined, the interpretability of the fraud prediction result output by the twin convolutional neural network model is increased, the automatic fraud prediction of the transaction behavior is further realized, manual participation is not needed, and the accuracy and efficiency of the fraud prediction of the transaction behavior are improved.
Example two
Fig. 2 is a flowchart of a fraud prediction method for transaction behaviors according to a second embodiment of the present invention, and this embodiment describes in detail a process of determining a fraud probability matrix through a twin convolutional neural network model based on the above embodiment. Wherein explanations of the same or corresponding terms as those of the above embodiments are omitted. As shown in fig. 2, the method includes:
and S210, acquiring current transaction running data corresponding to the current transaction behavior.
And S220, performing feature extraction on the current transaction flow data, and constructing a current transaction feature matrix corresponding to the current transaction behavior.
And S230, inputting the current transaction characteristic matrix into a first convolution network submodel for characteristic convolution processing, and determining a first characteristic vector.
The target transaction characteristic matrix is a transaction characteristic matrix which is constructed in advance based on normal transaction behavior flow data and fraud transaction behavior flow data corresponding to a plurality of fraud types. FIG. 3 is a schematic diagram of a twin convolutional neural network model. Illustratively, referring to fig. 3, the twin convolutional neural network model may include: the system comprises a first convolution network submodel, a second convolution network submodel and a fusion submodel, wherein the first convolution network submodel and the second convolution network submodel share weight values. The first convolutional network submodel and the second convolutional network submodel may have the same weight value.
Specifically, the feature matrix A for the current transaction m×n Because the convolution parameters such as the optimal convolution kernel size, the number of convolution steps, the number of convolution channels, and the like need to be determined through multiple experiments, the activation function in the convolution layer adopts a commonly used modified Linear Unit (ReLU) activation function. Since the convolution output is a multi-channel matrix, squareAnd performing dot product calculation, and elongating the two-dimensional vector in the convolved single channel into a one-dimensional vector. Wherein the functional form of ReLU is
Referring to FIG. 3, a current transaction feature matrix A may be formed m×n Inputting into a first convolution network sub-model of the twin convolution neural network model for feature convolution processing to determine a first feature vector, such as
And S240, inputting the target transaction characteristic matrix into a second convolution network submodel for characteristic convolution processing, and determining a vector matrix formed by a plurality of second characteristic vectors.
Specifically, referring to fig. 3, the target transaction feature matrix B may be input into a second convolution network sub-model of the twin convolution neural network model for feature convolution processing to obtain a plurality of second feature vectors, such as
And
and the vector matrix is formed based on a plurality of second feature vectors
And S250, inputting the first feature vector and the vector matrix into a fusion sub-model, carrying out fusion processing on the first feature vector and the vector matrix, and determining a fraud probability matrix.
Exemplarily, the "performing the fusion process on the first feature vector and the vector matrix and determining the fraud probability matrix" in S250 may include: performing dot product processing on the first feature vector and each second feature vector in the vector matrix, performing normalization processing on the obtained dot product result, and determining fraud probability corresponding to each type in the target transaction feature matrix, wherein the types comprise a normal type and multiple fraud types; adding the fraud probabilities belonging to the same type in the target transaction characteristic matrix to determine the fraud probability corresponding to each type; and ranking the fraud probabilities corresponding to the types based on the type ranking order to obtain a fraud probability matrix.
In particular, referring to fig. 3, the first feature vector O may be divided into A Sum vector matrix O B Inputting the first feature vector O into the fusion sub-model A Sum vector matrix O B Is subjected to a dot product process r. For example, the dot product result r may be
And normalizing the obtained dot product result r based on a Softmax function, and determining the fraud probability corresponding to each type in the target transaction characteristic matrix. Wherein the Softmax function is
And all of p j The addition result is 1. The fraud probabilities belonging to the same type in the target transaction characteristic matrix can be added to determine the fraud probability corresponding to each type; the fraud probabilities corresponding to the types can be ranked based on the type ranking order to obtain a fraud probability matrix. For example, if in the target transaction characteristics matrix
Fraud of the same type can be classified as the same rule as if a credit card had a record of transactions in different countries in a short period of time, or if in a target transaction signature matrix
Of another type, the same rule may be classified as having more than 6 identical Bank identification codes (Bank Ident) within twenty minutesidentification Number, BIN) Number credit cards are consumed by the same online merchant, or if the target transaction feature matrix has running data which really belongs to fraudulent transaction behaviors but cannot be classified into particularly obvious rule attributes, the running data can be classified into other types of rules. And adding the fraud probabilities of the same type to determine the corresponding fraud probability of each type, such as x 1 The fraud probability, x, corresponding to the transaction records of different countries in a short time for the same credit card can be 2 The fraud probability corresponding to the consumption of the same online merchant by more than 6 credit cards with the same BIN number can be calculated within twenty minutes. The fraud probabilities corresponding to the respective types may be ranked based on the type ranking order to obtain a fraud probability matrix X = [ X ] 1 ，x 2 ，x 3 ，…，x l ]。
And S260, determining the maximum fraud probability in the fraud probability matrix according to the fraud probability matrix output by the twin convolutional neural network model.
In particular, a fraud probability matrix X = [ X ] that may be output based on a twin convolutional neural network model 1 ，x 2 ，x 3 ，…，x l ]If x 1 If the probability is greater than the rest probabilities in the fraud probability matrix, the maximum fraud probability in the fraud probability matrix is determined to be x 1 。
S270, determining a target type corresponding to the target element position based on the type arrangement sequence and the target element position corresponding to the maximum fraud probability, and determining the target type and the maximum fraud probability as a fraud prediction result corresponding to the current transaction behavior.
Specifically, there is a correspondence between element positions and types. There is a correspondence between the type arrangement order and the element position. If the target element position corresponding to the maximum fraud probability is based on the type arrangement sequence, the target type corresponding to the target element position can be determined, and the target type and the maximum fraud probability are determined as fraud prediction results corresponding to the current transaction behaviors, so that not only is automatic fraud prediction of the transaction behaviors realized, but also the interpretability of the fraud prediction results output by the twin convolutional neural network model is enhanced.
According to the technical scheme of the embodiment of the invention, the fraud probabilities belonging to the same type in the target transaction characteristic matrix are fused (such as addition processing), so that the fraud probability corresponding to each type is determined; and based on the type arrangement sequence, arranging the fraud probabilities corresponding to the types to obtain a fraud probability matrix, determining the target type corresponding to the target element position based on the type arrangement sequence and the target element position corresponding to the maximum fraud probability, and determining the target type and the maximum fraud probability as a fraud prediction result corresponding to the current transaction behavior, so that automatic fraud prediction of the transaction behavior is realized, the interpretability of the fraud prediction result output by the twin convolutional neural network model is further enhanced, and the accuracy and efficiency of fraud prediction of the transaction behavior are improved.
The following is an embodiment of the fraud prediction apparatus for transaction behaviors provided in an embodiment of the present invention, which belongs to the same inventive concept as the fraud prediction method for transaction behaviors of the above embodiments, and reference may be made to the embodiment of the fraud prediction method for transaction behaviors, for details that are not described in detail in the embodiment of the fraud prediction apparatus for transaction behaviors.
EXAMPLE III
Fig. 4 is a schematic structural diagram of a fraud prediction apparatus for transaction behaviors according to a third embodiment of the present invention. As shown in fig. 4, the apparatus includes: the system comprises a current transaction flow data acquisition module 310, a current transaction characteristic matrix construction module 320, a fraud prediction module 330 and a fraud prediction result determination module 340.
The current transaction running data acquiring module 310 is configured to acquire current transaction running data corresponding to a current transaction behavior; the current transaction feature matrix construction module 320 is configured to perform feature extraction on current transaction flow data to construct a current transaction feature matrix corresponding to a current transaction behavior; the fraud prediction module 330 is configured to input the current transaction feature matrix and the target transaction feature matrix into the twin convolutional neural network model after training is completed to perform fraud prediction on the current transaction behavior, where the target transaction feature matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types; and the fraud prediction result determining module 340 is configured to determine a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, where the fraud prediction result includes the predicted target fraud type.
According to the technical scheme of the embodiment of the invention, the current transaction running data corresponding to the current transaction behavior is obtained; performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior; inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types; determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type; according to the technical scheme of the embodiment of the invention, the current transaction characteristic matrix and the target transaction characteristic matrix can be input into the trained twin convolutional neural network model, the fraud probability of the current transaction behavior is determined based on the output result, the possibility that the current transaction behavior belongs to the fraud transaction behavior can be determined through the fraud probability, and the specific fraud type of the current transaction behavior can be judged according to the position distributed by the maximum probability value, so that the actual fraud type and the twin convolutional neural network model are closely combined, the interpretability of the fraud prediction result output by the twin convolutional neural network model is increased, the automatic fraud prediction of the transaction behavior is further realized, manual participation is not needed, and the accuracy and efficiency of the fraud prediction of the transaction behavior are improved.
Optionally, the current transaction flow data obtaining module 310 is specifically configured to: and acquiring current transaction flow time sequence data corresponding to each transaction behavior generated by the user in the current time period, wherein the current transaction flow time sequence data comprises the current transaction flow data corresponding to the current transaction behavior generated by the user at the current moment.
Optionally, the current transaction feature matrix building module 320 is specifically configured to: and performing feature extraction on the current transaction flow time sequence data, performing normalization processing on each extracted feature data, and constructing a current transaction feature matrix corresponding to the current transaction behavior, wherein the row number of the current transaction feature matrix represents a time dimension, and the column number represents a feature dimension.
Optionally, fraud prediction module 330 may include:
the normal transaction characteristic matrix construction submodule is used for extracting the characteristics of the running data of at least one normal transaction behavior and constructing at least one normal transaction characteristic matrix;
the fraud transaction characteristic matrix construction submodule is used for carrying out characteristic extraction on at least one fraud transaction behavior running data corresponding to each fraud type and constructing at least one fraud transaction characteristic matrix corresponding to each fraud type;
and the target transaction characteristic matrix constructing submodule is used for performing matrix splicing processing on at least one normal transaction characteristic matrix and at least one fraud transaction characteristic matrix corresponding to each fraud type to construct a target transaction characteristic matrix.
Optionally, the twin convolutional neural network model may include: the system comprises a first convolution network submodel, a second convolution network submodel and a fusion submodel, wherein the first convolution network submodel and the second convolution network submodel share weight values;
the first feature vector determining submodule is used for inputting the current transaction feature matrix into a first convolution network submodel for feature convolution processing to determine a first feature vector;
the vector matrix determining submodule is used for inputting the target transaction characteristic matrix into the second convolution network submodel for characteristic convolution processing, and determining a vector matrix formed by a plurality of second characteristic vectors;
and the fraud probability matrix determining submodule is used for inputting the first feature vector and the vector matrix into the fusion submodel, carrying out fusion processing on the first feature vector and the vector matrix and determining the fraud probability matrix.
Optionally, the fraud probability matrix determination sub-module is specifically configured to: performing dot product processing on the first feature vector and each second feature vector in the vector matrix, performing normalization processing on the obtained dot product result, and determining fraud probability corresponding to each type in the target transaction feature matrix, wherein the types comprise a normal type and multiple fraud types; adding the fraud probabilities belonging to the same type in the target transaction characteristic matrix to determine the fraud probability corresponding to each type; and ranking the fraud probabilities corresponding to the types based on the type ranking order to obtain a fraud probability matrix.
Optionally, the fraud prediction result determining module 340 is specifically configured to: determining the maximum fraud probability in the fraud probability matrix according to the fraud probability matrix output by the twin convolutional neural network model; and determining a target type corresponding to the target element position based on the type arrangement sequence and the target element position corresponding to the maximum fraud probability, and determining the target type and the maximum fraud probability as a fraud prediction result corresponding to the current transaction behavior.
The transaction behavior fraud prediction device provided by the embodiment of the invention can execute the transaction behavior fraud prediction method provided by any embodiment of the invention, and has the corresponding functional module and beneficial effect of the execution method.
It should be noted that, in the embodiment of the fraud prediction apparatus for transaction behaviors, the included modules are merely divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be implemented; in addition, the specific names of the functional modules are only for convenience of distinguishing from each other and are not used for limiting the protection scope of the present invention.
Example four
FIG. 5 illustrates a block diagram of an electronic device 10 that may be used to implement an embodiment of the invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.
As shown in fig. 5, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a Read Only Memory (ROM) 12, a Random Access Memory (RAM) 13, and the like, wherein the memory stores a computer program executable by the at least one processor, and the processor 11 can perform various suitable actions and processes according to the computer program stored in the Read Only Memory (ROM) 12 or the computer program loaded from a storage unit 18 into the Random Access Memory (RAM) 13. In the RAM 13, various programs and data necessary for the operation of the electronic apparatus 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input/output (I/O) interface 15 is also connected to the bus 14.
A number of components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16 such as a keyboard, a mouse, or the like; an output unit 17 such as various types of displays, speakers, and the like; a storage unit 18 such as a magnetic disk, an optical disk, or the like; and a communication unit 19 such as a network card, modem, wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.
In some embodiments, the fraud prediction method of transaction behavior may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 10 via the ROM 12 and/or the communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the above described fraud prediction method of transaction behaviour may be performed. Alternatively, in other embodiments, the processor 11 may be configured to perform fraud prediction methods of transaction behaviour by any other suitable means (e.g. by means of firmware).
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
A computer program for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be performed. A computer program can execute entirely on a machine, partly on a machine, as a stand-alone software package partly on a machine and partly on a remote machine or entirely on a remote machine or server.
In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user may provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), blockchain networks, and the internet.
The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical host and VPS service are overcome.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired results of the technical solution of the present invention can be achieved.
The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method for fraud prediction of transaction behavior, comprising:
acquiring current transaction flow data corresponding to current transaction behaviors;
performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior;
inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types;
and determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type.
2. The method of claim 1, wherein obtaining current transaction flow data corresponding to current transaction activity comprises:
acquiring current transaction flow time sequence data corresponding to each transaction behavior generated by a user in a current time period, wherein the current transaction flow time sequence data comprises current transaction flow data corresponding to the current transaction behavior generated by the user at the current moment.
3. The method of claim 2, wherein performing feature extraction on the current transaction flow data to construct a current transaction feature matrix corresponding to the current transaction behavior comprises:
and performing feature extraction on the current transaction flow time sequence data, performing normalization processing on each extracted feature data, and constructing a current transaction feature matrix corresponding to the current transaction behavior, wherein the line number of the current transaction feature matrix represents a time dimension, and the column number represents a feature dimension.
4. The method of claim 1, wherein constructing the target transaction feature matrix based on the normal transaction behavior pipeline data and the fraud transaction behavior pipeline data corresponding to the plurality of fraud types comprises:
performing feature extraction on the running data of at least one normal transaction behavior to construct at least one normal transaction feature matrix;
performing feature extraction on at least one fraud transaction behavior flow data corresponding to each fraud type to construct at least one fraud transaction feature matrix corresponding to each fraud type;
and performing matrix splicing processing on the at least one normal transaction characteristic matrix and the at least one fraud transaction characteristic matrix corresponding to each fraud type to construct a target transaction characteristic matrix.
5. The method of claim 1, wherein the twin convolutional neural network model comprises: the system comprises a first convolution network submodel, a second convolution network submodel and a fusion submodel, wherein the first convolution network submodel and the second convolution network submodel share a weight value;
inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to predict fraud of the current transaction behavior, wherein the fraud prediction comprises the following steps:
inputting the current transaction feature matrix into the first convolution network sub-model for feature convolution processing to determine a first feature vector;
inputting the target transaction characteristic matrix into the second convolution network submodel for characteristic convolution processing, and determining a vector matrix consisting of a plurality of second characteristic vectors;
and inputting the first feature vector and the vector matrix into the fusion sub-model, and performing fusion processing on the first feature vector and the vector matrix to determine a fraud probability matrix.
6. The method of claim 5, wherein fusing the first feature vector and the vector matrix to determine a fraud probability matrix comprises:
performing dot product processing on the first feature vector and each second feature vector in the vector matrix, performing normalization processing on an obtained dot product result, and determining fraud probability corresponding to each type in the target transaction feature matrix, wherein the types comprise a normal type and multiple fraud types;
adding the fraud probabilities belonging to the same type in the target transaction characteristic matrix to determine the fraud probability corresponding to each type;
and ranking the fraud probabilities corresponding to the types based on the type ranking order to obtain a fraud probability matrix.
7. The method of claim 5, wherein determining a fraud prediction corresponding to the current transaction behavior based on an output of the twin convolutional neural network model comprises:
determining the maximum fraud probability in the fraud probability matrix according to the fraud probability matrix output by the twin convolutional neural network model;
and determining a target type corresponding to the target element position based on the type arrangement sequence and the target element position corresponding to the maximum fraud probability, and determining the target type and the maximum fraud probability as a fraud prediction result corresponding to the current transaction behavior.
8. An apparatus for predicting fraud in a transaction, comprising:
the current transaction flow data acquisition module is used for acquiring current transaction flow data corresponding to the current transaction behavior;
the current transaction characteristic matrix construction module is used for extracting the characteristics of the current transaction running data and constructing a current transaction characteristic matrix corresponding to the current transaction behavior;
the fraud prediction module is used for inputting the current transaction characteristic matrix and the target transaction characteristic matrix into a twin convolutional neural network model after training is finished to perform fraud prediction of a current transaction behavior, wherein the target transaction characteristic matrix is constructed in advance based on normal transaction behavior running data and fraud transaction behavior running data corresponding to multiple fraud types;
and the fraud prediction result determining module is used for determining a fraud prediction result corresponding to the current transaction behavior according to the output of the twin convolutional neural network model, wherein the fraud prediction result comprises a predicted target fraud type.
9. An electronic device, characterized in that the electronic device comprises:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the method of fraud prediction of transaction behaviour of any of claims 1-7.
10. A computer-readable storage medium having stored thereon computer instructions for causing a processor to perform the method of fraud prediction of transaction behaviour of any of claims 1-7 when executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211158596.9A CN115545712A (en) | 2022-09-22 | 2022-09-22 | Fraud prediction method, device, equipment and storage medium for transaction behaviors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211158596.9A CN115545712A (en) | 2022-09-22 | 2022-09-22 | Fraud prediction method, device, equipment and storage medium for transaction behaviors |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115545712A true CN115545712A (en) | 2022-12-30 |
Family
ID=84728818
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211158596.9A Pending CN115545712A (en) | 2022-09-22 | 2022-09-22 | Fraud prediction method, device, equipment and storage medium for transaction behaviors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115545712A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237114A (en) * | 2023-11-10 | 2023-12-15 | 深圳市迪博企业风险管理技术有限公司 | Financing trade compliance detection method based on twin evolution |
-
2022
- 2022-09-22 CN CN202211158596.9A patent/CN115545712A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237114A (en) * | 2023-11-10 | 2023-12-15 | 深圳市迪博企业风险管理技术有限公司 | Financing trade compliance detection method based on twin evolution |
| CN117237114B (en) * | 2023-11-10 | 2024-03-08 | 深圳市迪博企业风险管理技术有限公司 | Financing trade compliance detection method based on twin evolution |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110009174B (en) | Risk recognition model training method and device and server | |
| CN110210900B (en) | Reference product determination method, device and equipment based on transaction relevance | |
| CN109740642A (en) | Invoice category recognition methods, device, electronic equipment and readable storage medium storing program for executing | |
| CN112016967B (en) | Transaction data processing method and device | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN112669143A (en) | Risk assessment method, device and equipment based on associated network and storage medium | |
| CN115409419A (en) | Value evaluation method and device of business data, electronic equipment and storage medium | |
| CN115545712A (en) | Fraud prediction method, device, equipment and storage medium for transaction behaviors | |
| CN113435900A (en) | Transaction risk determination method and device and server | |
| CN111464337B (en) | Resource allocation method and device and electronic equipment | |
| CN110213239B (en) | Suspicious transaction message generation method and device and server | |
| Arief et al. | Fraud detection based-on data mining on Indonesian E-Procurement System (SPSE) | |
| CN116800831A (en) | Service data pushing method, device, storage medium and processor | |
| CN110619564B (en) | Anti-fraud feature generation method and device | |
| CN116071150A (en) | Data processing method, bank product popularization, wind control system, server and medium | |
| CN115545909A (en) | Approval method, device, equipment and storage medium | |
| CN114943608A (en) | Fraud risk assessment method, device, equipment and storage medium | |
| CN113379464A (en) | Block chain-based site selection method, device, equipment and storage medium | |
| CN111429257A (en) | Transaction monitoring method and device | |
| CN113837183B (en) | Multi-stage certificate intelligent generation method, system and medium based on real-time mining | |
| CN114418752B (en) | Method and device for processing user data without type label, electronic equipment and medium | |
| CN115731001A (en) | Product recommendation method, device, equipment and medium | |
| CN115526708A (en) | Overdue risk probability determination method and device, electronic equipment and medium | |
| CN117436931A (en) | Marketing amount distribution method, marketing amount distribution device, marketing amount distribution equipment and marketing amount distribution medium | |
| CN115640461A (en) | Product recommendation method and device, electronic equipment and storage medium thereof |
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 |