Transaction risk prediction method, device, equipment, medium and product

Abstract

The invention discloses a transaction risk prediction method, device and equipment, a medium and a product. The method comprises the following steps: acquiring at least one piece of historical transaction information of at least one first object of a business platform within a first historical duration before a current moment, and at least one piece of first media resource which is published within the first historical duration and is associated with a financial abnormal event; for at least one first object, according to at least one piece of historical transaction information of the first object, at least one first media resource and a transaction risk prediction model obtained by pre-training, predicting an abnormal risk probability corresponding to at least one future transaction executed by the first object in a future duration, and processing the future transaction according to the abnormal risk probability corresponding to the future transaction when the business platform receives the future transaction. According to the technical scheme, the effect of predicting the risk probability of the future transaction in combination with the historical transaction and the external financial event information is achieved, and the prediction accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of computer data processing technology, and in particular to a method, apparatus, equipment, medium and product for predicting transaction risks. Background Technology

[0002] With the rapid growth of payment services, the scale of global financial transactions continues to increase, but the probability of abnormal transaction operations is also becoming increasingly serious.

[0003] Currently, the methods for predicting abnormal transactions still mainly rely on the subjective experience of users (such as managers) and simple transaction data statistics. It is difficult to achieve standardized, objective and scientific results, and it is time-consuming and labor-intensive, which can easily lead to missed or incorrect judgments and has a low level of automation. Summary of the Invention

[0004] This invention provides a method, apparatus, device, medium, and product for predicting transaction risks, which can combine historical transaction and external financial event information to predict the risk probability of future transactions, thereby improving prediction accuracy and reducing the false positive and false negative rates.

[0005] According to one aspect of the present invention, a method for predicting transaction risk is provided, the method comprising:

[0006] Obtain at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, and at least one first media resource related to a financial anomaly published within the first historical period; wherein, the first object is an object that has a transaction record on the business platform within the first historical period;

[0007] For at least one first object, based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model, the probability of abnormal risk corresponding to at least one future transaction executed by the first object within a future time period is predicted, so that when the business platform receives the future transaction, the future transaction is processed according to the probability of abnormal risk corresponding to the future transaction.

[0008] The transaction risk prediction model is trained on a machine learning model based on historical transaction information of multiple sample objects, sample media resources associated with historical financial events, and the actual risk status of the transactions of the sample objects at the prediction time.

[0009] According to another aspect of the present invention, a transaction risk prediction device is provided, the device comprising:

[0010] The transaction information acquisition module is used to acquire at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, and at least one first media resource related to a financial anomaly published within the first historical period; wherein, the first object is an object that has a transaction record on the business platform within the first historical period;

[0011] The transaction risk prediction module is used to predict the probability of abnormal risk for at least one future transaction executed by at least one first object within a future time period, based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model. This is so that when the business platform receives the future transaction, it can process the future transaction based on the probability of abnormal risk corresponding to the future transaction. The transaction risk prediction model is trained on a machine learning model based on historical transaction information of multiple sample objects, sample media resources associated with historical financial events, and the actual risk status of the transactions occurring at the prediction time for the sample objects.

[0012] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0013] at least one processor; and

[0014] A memory communicatively connected to the at least one processor; wherein,

[0015] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the transaction risk prediction method according to any embodiment of the present invention.

[0016] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the transaction risk prediction method according to any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer program product is provided, the computer program product comprising a computer program that, when executed by a processor, implements the transaction risk prediction method according to any embodiment of the present invention.

[0018] The technical solution of this invention obtains at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, as well as at least one first media resource associated with a financial anomaly published within the first historical period. Further, for at least one first object, based on the at least one historical transaction information, the at least one first media resource, and a pre-trained transaction risk prediction model, the probability of an anomaly risk corresponding to at least one future transaction executed by the first object within a future period is predicted. When the business platform receives a future transaction, it processes the future transaction based on the probability of an anomaly risk corresponding to the future transaction. This solves the problems of time-consuming and labor-intensive anomaly transaction prediction methods in related technologies, which are prone to missed or false judgments and have low automation levels. It achieves the effect of predicting the risk probability of future transactions by combining historical transaction and external financial event information, improving prediction accuracy, reducing false and missed judgment rates, and enhancing the automation level of anomaly transaction prediction, reducing manual intervention, and improving the efficiency of anomaly transaction risk prediction. Furthermore, by predicting future transaction risks in advance, the business platform can handle them in a targeted manner, enhancing the efficiency and accuracy of risk prevention and control.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of a transaction risk prediction method provided by an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of a transaction risk prediction method provided by an embodiment of the present invention;

[0023] Figure 3 This is a schematic diagram of the structure of a transaction risk prediction device provided in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of an electronic device that implements the transaction risk prediction method of this invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] Figure 1 This is a flowchart of a transaction risk prediction method provided by an embodiment of the present invention. This embodiment is applicable to situations where abnormal risks are predicted for transactions received by a business platform. The method can be executed by a transaction risk prediction device, which can be implemented in hardware and / or software and can be configured in a terminal and / or server. Figure 1 As shown, the method includes:

[0028] S110. Obtain at least one historical transaction information of at least one first object within a first historical period prior to the current moment from the business platform, and at least one first media resource related to the financial anomaly published within the first historical period.

[0029] The business platform can refer to a system or institution that provides transaction services, such as a bank's app or an e-commerce payment platform. Users conduct transactions such as transfers, purchases, and investments on this platform, and the platform needs to manage the security of these transactions. The first historical period can be the historical time range used to collect data. The first historical period before the current moment can be a historical time range including the current moment, or a segment extracted from the historical time range before the current moment. The first historical period can be any length, such as one month, three months, six months, or one year. The first object can be a user or entity (individual, enterprise, etc.) that has a transaction record on the business platform within the historical period. For example, a customer who has a transfer record with a bank in the past six months is the first object on that bank's platform. Historical transaction information can be transaction data of the first object that occurred on the business platform within the first historical period, including transaction time, transaction value attributes, transaction frequency, transaction object, transaction type, and transaction location. At least one historical transaction record corresponding to the first object can reflect the first object's past transaction habits on the business platform. For example, normal objects typically make transfers at fixed times, with a relatively low transfer frequency and a relatively stable transfer value attribute, while abnormal objects may suddenly make large cross-border transfers, with abnormal transfer frequency and transfer value attributes.

[0030] Financial anomalies refer to unusual circumstances that may affect the security of financial transactions. Examples include abnormal trading events, platform system vulnerabilities, policy and regulatory changes, and abnormal trading surges caused by market volatility. Primary media resources refer to information sources related to financial anomalies published within a historical period. Optionally, primary media resources include news reports, industry announcements, information published on social media, and risk warnings issued by regulatory agencies. Primary media resources can be used to assist in assessing potential risks in the trading environment.

[0031] In practical applications, to predict transaction risks for the first party with transaction records within a business platform and determine whether the first party is an abnormal transaction party, historical transaction information of the first party can be analyzed. However, relying solely on historical transaction information to predict future risks for the first party may fail to capture new risks brought about by changes in the external environment, lacking the ability to perceive the external environment. This limits the comprehensiveness of anomaly probability prediction, easily leading to misjudgments or omissions. Furthermore, it is difficult to identify reasonable changes in user trading habits due to objective factors, easily misjudging normal transactions.

[0032] To address the above, in this embodiment, for at least one first object with transaction records existing within the business platform during the historical period, at least one historical transaction record of the first object within the historical period prior to the current moment is retrieved from the transaction database within the business platform. Furthermore, the published media resources within the historical period can be detected through external interfaces within the business platform. If the published media resources contain media resources associated with the financial anomaly, the relevant media resources can be collected, and at least one first media resource associated with the financial anomaly published during the historical period can be obtained. Therefore, at least one historical transaction record of at least one first object within the historical period and at least one first media resource associated with the financial anomaly published during the historical period can be obtained.

[0033] S120. For at least one first object, based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model, predict the probability of abnormal risk corresponding to at least one future transaction executed by the first object within a future time period, so as to process the future transaction according to the probability of abnormal risk corresponding to the future transaction when the business platform receives the future transaction.

[0034] In this embodiment, given at least one historical transaction information of at least one first object within a first historical period and at least one first media resource associated with a financial anomaly published within the first historical period, a pre-trained transaction risk prediction model can be used to predict transaction risk based on at least one historical transaction information and at least one first media resource, thereby predicting the probability of anomaly risk corresponding to at least one future transaction executed by the first object at a future time.

[0035] The transaction risk prediction model can be a machine learning-trained model used to predict the probability of abnormal risks in future transactions based on input historical transaction information and primary media resources. The model is trained using machine learning based on historical transaction information of multiple sample objects, sample media resources associated with historical financial events, and the actual risk status of the transactions occurring at the prediction time. Sample objects are the objects in the historical data used to train the model. These objects can also be those with transaction records on the business platform within a historical timeframe, such as users with transaction records on the platform over the past five years. Sample media resources associated with historical financial events refer to information carriers related to financial risk events published during the period in which the sample objects' transactions occurred, including news reports, regulatory announcements, and social media discussions. The actual risk status characterizes whether a transaction carries risk, categorized as normal or abnormal.

[0036] Here, "future duration" refers to a point in time after the current moment, which is the time frame within which risk needs to be predicted. Optionally, future duration can include the next month, three months, six months, or one year. "Future transactions" refer to transactions planned or to be executed on the business platform by the first party within the future duration. These transactions have not yet occurred, and their risks need to be predicted in advance so that the business platform can take appropriate measures when the transaction is initiated. "Abnormal risk probability" refers to the likelihood of an anomaly in the future transaction, typically represented by a value between 0 and 1 (e.g., 0.8 indicates an 80% probability of an abnormal transaction). A higher abnormal risk probability indicates a greater transaction risk.

[0037] In this embodiment, the model structure of the transaction risk prediction model can be set according to actual needs. Optionally, the transaction risk prediction model may include a feature sequence construction sub-model and an abnormal risk probability prediction sub-model.

[0038] In practical implementation, after obtaining at least one historical transaction information and at least one first media resource for at least one first object, the at least one historical transaction information and at least one first media resource corresponding to the first object can be input into a pre-trained transaction risk prediction model. Furthermore, the feature sequence sub-model and the abnormal risk probability prediction sub-model in the transaction risk prediction model can predict transaction risk based on the at least one historical transaction information and at least one first media resource, predicting the abnormal risk probability corresponding to at least one future transaction executed by the first object within a future timeframe. Consequently, when the business platform receives a future transaction initiated by the first object at a future time, it can process the future transaction based on the abnormal risk probability corresponding to the future transaction.

[0039] In this embodiment, when processing future transactions based on the probability of abnormal risks, at least one processing method may be included, including: comparing the probability of abnormal risks with a preset probability threshold to determine the corresponding transaction processing method; comparing the probability of abnormal risks with a probability range corresponding to a preset risk level to determine the risk level to which the transaction belongs, and determining the transaction processing method according to the risk level.

[0040] Optionally, future transactions can be processed based on the probability of abnormal risks corresponding to them, including: allowing future transactions if the probability of abnormal risks is within a preset low-risk level; performing light authentication on future transactions if the probability of abnormal risks is within a preset low-to-medium risk level; performing enhanced authentication on future transactions if the probability of abnormal risks is within a preset high-to-medium risk level; and blocking future transactions if the probability of abnormal risks is within a preset high-risk level.

[0041] Typically, business platforms can pre-define probability ranges to differentiate between high and low risks based on business needs and risk control objectives. These ranges can be categorized as low-risk, low-to-medium-risk, medium-to-high-risk, and high-risk. Optionally, the probability range for the low-risk level can be set to 0%-5%, indicating extremely low risk; the probability range for the low-to-medium-risk level can be set to 5%-20%, indicating low risk but requiring slight attention; the probability range for the medium-to-high-risk level can be set to 20%-70%, indicating high risk requiring rigorous verification; and the probability range for the high-risk level can be set to 70%-100%, indicating extremely high risk requiring immediate prevention.

[0042] The three types of authentication processes are as follows: **Acceptance Processing:** Approval can be a method where the business platform directly approves a future transaction without additional verification. The core of approval processing is that the system automatically completes the transaction, requiring no additional user action, thus ensuring transaction efficiency and user experience. **Light Authentication Processing:** This method requires users to complete simple verifications to confirm the authenticity of the transaction. This method uses simple verification methods with few steps, aiming to eliminate low-probability risks with only a slight increase in user intervention. For example, light authentication processing may include at least one of the following: entering SMS verification codes or email verification codes; answering preset security questions; or rapid facial recognition. **Enhanced Authentication Processing:** This method uses more stringent verification methods to confirm the legitimacy of the transaction. This method involves more complex verification steps and more detailed information, used to screen for medium to high risks and reduce the probability of misjudging legitimate transactions. For example, enhanced authentication processing may include at least one of the following: uploading identity information for identity verification; verifying transaction details via customer service phone; or providing relevant transaction vouchers. **Blocking Processing:** This method involves the business platform directly blocking the execution of a future transaction, preventing its completion, and usually informing the first party of the risk reason. The core of blocking processing is to prevent high-risk transactions and avoid loss of user resources.

[0043] In practical implementation, when the business platform receives a future transaction initiated by the first object within a future timeframe, the abnormal risk probability corresponding to that future transaction can be determined from a pre-defined abnormal risk probability. Furthermore, this abnormal risk probability can be compared with the probability range corresponding to a preset risk level, and the transaction processing method corresponding to the future transaction can be determined based on the comparison result. If the abnormal risk probability is within the probability range corresponding to a low-risk level, the future transaction can be allowed; if it is within the probability range corresponding to a low-to-medium risk level, a light authentication method can be used, and the future transaction can be allowed if authentication is successful; if it is within the probability range corresponding to a medium-to-high risk level, a strengthened authentication method can be used, and the future transaction can be allowed if authentication is successful; if the abnormal risk probability is within the probability range corresponding to a high-risk level, the future transaction can be blocked.

[0044] Optionally, future transactions can be processed based on the probability of abnormal risks corresponding to the future transactions, including: intercepting future transactions when the probability of abnormal risks is greater than a preset probability threshold; and allowing future transactions when the probability of abnormal risks is not greater than the preset probability threshold.

[0045] The preset probability threshold can be a critical probability value pre-set by the business platform based on risk control needs and business rules. This probability value can be used to determine whether a transaction needs to be blocked. The preset probability threshold is the "watershed" for judging the degree of transaction risk, determined by the business platform in combination with historical data, compliance requirements, and other factors, and can be dynamically adjusted. Optionally, the preset probability threshold can be 50%, 55%, or 60%, etc.

[0046] In practical implementation, if the business platform receives a future transaction initiated by the first object within a future time period, it can determine the probability of anomaly risk corresponding to that future transaction from a pre-determined probability of anomaly. Furthermore, this probability of anomaly risk can be compared with a preset probability threshold. If the probability of anomaly risk is greater than the preset probability threshold, the future transaction can be blocked; if the probability of anomaly risk is not greater than the preset probability threshold, the future transaction can be allowed.

[0047] The technical solution of this invention obtains at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, as well as at least one first media resource associated with a financial anomaly published within the first historical period. Further, for at least one first object, based on the at least one historical transaction information, the at least one first media resource, and a pre-trained transaction risk prediction model, the probability of an anomaly risk corresponding to at least one future transaction executed by the first object within a future period is predicted. When the business platform receives a future transaction, it processes the future transaction based on the probability of an anomaly risk corresponding to the future transaction. This solves the problems of time-consuming and labor-intensive anomaly transaction prediction methods in related technologies, which are prone to missed or false judgments and have low automation levels. It achieves the effect of predicting the risk probability of future transactions by combining historical transaction and external financial event information, improving prediction accuracy, reducing false and missed judgment rates, and enhancing the automation level of anomaly transaction prediction, reducing manual intervention, and improving the efficiency of anomaly transaction risk prediction. Furthermore, by predicting future transaction risks in advance, the business platform can handle them in a targeted manner, enhancing the efficiency and accuracy of risk prevention and control.

[0048] In this embodiment, before applying the trading risk prediction model, a pre-built machine learning model can be trained using supervised or unsupervised methods. Before training the machine learning model, multiple training samples can be constructed to train the model based on these samples. To improve the prediction accuracy of the trading risk prediction model, as many and varied training samples as possible can be constructed.

[0049] Optionally, training a transaction risk prediction model includes: acquiring multiple training samples, wherein the training samples include historical transaction information of multiple sample objects within a second historical time period, sample media resources associated with historical financial events, and the actual risk status of transactions occurring in the sample objects within the prediction time period; for multiple training samples, inputting the historical transaction information and sample media resources in the training samples into a pre-built machine learning model to obtain the predicted abnormal risk probability corresponding to the training samples; determining the loss value based on the predicted abnormal risk probability and the actual risk status in the training samples, correcting the model parameters in the machine learning model based on the loss value, using the convergence of the loss function in the machine learning model as the training objective, and obtaining at least one model to be verified; verifying at least one model to be verified based on test samples, and using the model to be verified whose verification results meet preset conditions as the transaction risk prediction model.

[0050] The predicted duration must be later than the second historical duration but earlier than the current moment. In other words, the predicted duration is any historical duration prior to the current moment, and the time interval between this historical duration and the current moment must be smaller than the time interval between the second historical duration and the current moment. The machine learning model used for prediction can be a neural network model with default or initial parameters. The predicted probability of abnormal risk can be the probability of abnormal risk output after the machine learning model predicts historical transaction information and sample media resources. The loss value can be a numerical value characterizing the degree of difference between the model output and the actual output. The loss function can be a function determined based on the loss value, used to characterize the degree of difference between the predicted output and the actual output. The model to be validated can be a model obtained by correcting the model parameters in the machine learning model based on the loss value. The number of models to be validated can be consistent with the number of training iterations (or the number of model parameter corrections). The test samples can be a portion of the training samples or sample data reconstructed based on the construction method of the training samples. The preset conditions may include at least one of the following: Mean-Square Error (MSE) reaching a first error threshold, Mean Absolute Error (MAE) reaching a second error threshold, and Mean Absolute Percentage Error (MAPE) reaching a third error threshold. MSE is typically used to measure the average squared difference between the predicted and actual values. MAE is typically used to measure the average absolute difference between the predicted and actual values. MAPE is typically used to measure the average absolute percentage deviation of the predicted value from the actual value.

[0051] In this embodiment, to construct a rich and diverse set of training samples, historical transaction information of multiple sample objects and sample media resources associated with historical financial events can be obtained. Furthermore, the actual risk status corresponding to at least one transaction of a sample object within the prediction period can be acquired. Further, training samples can be constructed based on the historical transaction information of the sample objects, the sample media resources associated with historical financial events, and the actual risk status corresponding to at least one transaction of the sample objects within the prediction period, thus obtaining multiple training samples. Further, for multiple training samples, the historical transaction information and sample media resources in the training samples can be input into a pre-built machine learning model to process the historical transaction information and sample media resources based on the machine learning model, and output the predicted abnormal risk probability corresponding to the training samples. Further, the predicted abnormal risk probability can be compared with the actual risk status in the training samples to obtain a loss value. Further, with the loss value obtained, the model parameters in the machine learning model can be corrected based on the loss value. The corrected model is used as the first model to be processed, and a backup of the first model to be processed is created and saved. Further, the process of processing training samples and determining the loss value can be repeated. Furthermore, the model parameters in the model to be processed can be corrected based on the obtained loss value. The corrected model is then used as the second model to be processed, and a backup model is created and saved. Further, when convergence of the loss function is detected, such as when the training error of the loss function is less than a preset error, or when the error trend stabilizes, iterative training can be stopped, and the model obtained at this point is used as the model to be verified. Then, the saved backup model can be retrieved, and both the retrieved backup model and the model to be verified are used as models to be verified, resulting in at least one model to be verified. Further, to evaluate the performance of at least one model to be verified, test samples can be obtained. For at least one model to be verified, the model to be verified can be tested and verified based on the test samples, and the verification results corresponding to the model to be verified can be obtained. Further, it can be determined whether the verification results of at least one model to be verified meet preset conditions. Finally, the model to be verified whose verification results meet the preset conditions can be used as the trading risk prediction model.

[0052] In this embodiment, to improve the performance of the transaction risk prediction model during model application, and thereby enhance the prediction accuracy and decision-making precision of the business platform, model optimization can be performed during the application of the transaction risk prediction model on the business platform. Optionally, the method further includes: deploying the transaction risk prediction model in the business platform, and updating the model parameters of the transaction risk prediction model based on the historical transaction information of the first object, the first media resources associated with the financial anomaly event, the processing results of future transactions, and the response results.

[0053] The processing result can be the output information or status returned after processing a future transaction. The processing result can be used to represent detailed information about the processing status of the future transaction. The response result can be the user's feedback on the processing status of the future transaction.

[0054] In practical implementation, after obtaining the transaction risk prediction model, it can be deployed on the business platform. Then, based on the historical transaction information of the first target and the first media resources associated with the financial anomaly obtained from the business platform, the transaction risk prediction model can be used to process the historical transaction information and the first media resources to determine the processing results of future transactions and receive user responses regarding the processing status of future transactions. Furthermore, the model parameters of the transaction risk prediction model can be corrected based on the historical transaction information of the first target, the first media resources associated with the financial anomaly, the processing results of future transactions, and the response results. Thus, the model performance can be continuously optimized during the application of the transaction risk prediction model.

[0055] Figure 2 This is a flowchart of a transaction risk prediction method provided by an embodiment of the present invention. Based on the foregoing embodiments, the transaction risk prediction model includes a feature sequence construction sub-model and an abnormal risk probability prediction sub-model. The method of processing at least one historical transaction information and at least one first media resource according to the transaction risk prediction model is further refined. For specific implementation details, please refer to the technical solution of this embodiment. Technical terms that are the same as or similar to those in the above embodiments will not be repeated here. Figure 2 As shown, the method includes:

[0056] S210. Obtain at least one historical transaction information of at least one first object within the historical time period prior to the current moment from the business platform, and at least one first media resource related to the financial anomaly published within the historical time period.

[0057] S220. For at least one first object, construct a sub-model based on the feature sequence to process at least one historical transaction information and at least one first media resource to obtain a transaction feature vector sequence corresponding to the first object.

[0058] The feature sequence construction sub-model can be a neural network model used for feature extraction and feature sequence construction from input data. That is, the feature sequence construction sub-model can be used to transform multi-source heterogeneous input data into a structured, temporally ordered sequence of feature vectors. It can extract valuable features from messy input data and organize them into a model-recognizable format according to time or logical order. The feature sequence construction sub-model can be a neural network model of any structure, optionally including fully connected neural networks, recurrent neural networks, long short-term memory networks, or Transformer models. Preferably, the feature sequence construction sub-model can adopt a Transformer model, as its core objective is to transform unstructured or multi-source heterogeneous data (such as structured historical transaction information, text or images from primary media resources, etc.) into a structured, temporally ordered sequence of feature vectors while preserving the relationships between data. The Transformer's structural design perfectly meets this requirement. Using Transformer as the core structure of the feature sequence sub-model, its output transaction feature vector sequence can not only retain the temporal logic of the original data, but also integrate cross-data correlation information, providing high-quality input features for subsequent anomaly risk prediction.

[0059] In this embodiment, the feature sequence construction sub-model may include multiple modules, each of which is used to perform different operations. Optionally, it may include a transaction feature extraction module, an encoding embedding module, and a feature sequence construction module.

[0060] It should be noted that, in order to improve data quality and thus improve the accuracy of risk probability prediction, the historical transaction information and the first media resource can be preprocessed before inputting at least one historical transaction information and at least one first media resource corresponding to the first object into the feature sequence construction sub-model. Then, the preprocessed historical transaction information and the first media resource can be input into the feature sequence construction sub-model.

[0061] Optionally, preprocessing is performed on at least one historical transaction information and at least one first media resource, including: preprocessing at least one historical transaction information using a first preprocessing method to obtain preprocessed historical transaction information; and preprocessing the first media resource using a second preprocessing method to obtain preprocessed first media resource.

[0062] The first preprocessing method can be a method of preprocessing structured data. Optionally, the first preprocessing method includes at least one of missing value imputation, outlier handling, standardization, and normalization. For example, missing value imputation can be used to fill missing values ​​of transaction amounts with the mean, median, or historical trend; outlier handling can be used to identify, correct, or remove obviously unreasonable transaction data through standard deviation and / or interquartile range methods; standardization or normalization can unify the scale of features of different magnitudes to avoid the impact of magnitude differences on subsequent model processing.

[0063] The second preprocessing method can be a method for preprocessing unstructured data. Optionally, the second preprocessing method includes at least one of text cleaning, text segmentation, and format conversion. Text cleaning can be understood as removing punctuation, stop words, special characters, and case sensitivity from text data; text segmentation is the process of splitting text into words or clauses; and format conversion can be the process of converting text in an image into an editable string.

[0064] In this embodiment, after obtaining the preprocessed historical transaction information and the first media resource, at least one historical transaction and at least one first media resource can be input into the feature sequence construction sub-model. Then, based on the transaction feature extraction module, encoding embedding module, and feature sequence construction module in the feature sequence construction sub-model, the at least one historical transaction and at least one first media resource can be processed to obtain a transaction feature vector sequence corresponding to the first object.

[0065] Optionally, a sub-model based on the feature sequence is used to process at least one historical transaction information and at least one first media resource to obtain a transaction feature vector sequence corresponding to the first object. This includes: extracting features from at least one historical transaction information based on the transaction feature extraction module to obtain a historical transaction feature vector corresponding to at least one historical transaction information; and encoding and embedding at least one first media resource based on the encoding and embedding module to obtain a financial event feature vector corresponding to at least one first media resource; and processing at least one historical transaction feature vector and at least one financial event feature vector based on the feature sequence construction module to obtain a transaction feature vector sequence corresponding to the first object.

[0066] The transaction feature extraction module can be used to analyze historical transaction information and extract key features associated with the transaction behavior. Optionally, the transaction feature extraction module can be an embedding layer in the feature sequence construction sub-model. It can be understood that the embedding layer is a key layer in deep learning used to map discrete inputs into densely connected vectors. Through the embedding layer, transaction-related features in historical transaction information can be extracted, and the extracted structured features can be transformed into an initial vector, thus obtaining the historical transaction feature vector. The historical transaction information feature vector can be a numerical vector obtained after processing by the transaction feature extraction module. Each element in the historical transaction feature vector corresponds to a specific feature. Optionally, the historical transaction feature vector includes at least one feature from the following: average transaction value attribute value, average monthly transaction frequency, transaction frequency, nighttime transaction ratio, and whether it crosses regions. For example, the historical transaction feature vector can be [5000 (average transaction value attribute value), 3 (average monthly transaction frequency), 0.2 (nighttime transaction ratio), 1 (whether it crosses regions)].

[0067] The encoding and embedding module analyzes the first media resource to transform media information, such as text and images, into structured feature vectors. The encoding and embedding module can be a neural network model with any model structure. For example, it can be a BERT (Bidirectional Encoder Representations from Transformers) model, a pre-trained natural language processing (NLP) model based on the Transformer architecture. The financial event feature vector is a numerical vector obtained after processing the first media resource by the encoding and embedding module, representing a quantitative description of the financial event in the media information. Each element in the financial event feature vector corresponds to a key attribute. Optionally, the financial event feature vector includes at least one key attribute from the following: event type, scope of impact, severity, event location, and the objects involved in the event.

[0068] The feature sequence construction module can be used to logically combine historical transaction feature vectors and financial event feature vectors to form an ordered sequence, preserving the correlation between data. The feature sequence construction module can also be used to concatenate historical transaction feature vectors and at least one financial event feature vector, and then combine the concatenated vectors in chronological order to form an ordered sequence. In this embodiment, the feature sequence construction module may include a feature concatenation unit, a self-attention layer, and a feature sequence construction unit.

[0069] It should be noted that the final transaction feature vector sequence is the input data used for predicting the probability of abnormal risks. In order to improve the prediction accuracy and the comprehensiveness of the prediction basis, the historical transaction feature vector and at least one financial event feature vector can be concatenated to form the first transaction feature vector corresponding to a historical transaction. This first transaction feature vector contains both the transaction features in the historical transaction information and the financial attribute features in at least one first media resource.

[0070] Optionally, the feature sequence construction module processes at least one historical transaction feature vector and at least one financial event feature vector to obtain a transaction feature vector sequence corresponding to the first object, including: concatenating at least one historical transaction feature vector and at least one financial event feature vector using a feature concatenation unit to obtain at least one first transaction feature vector; wherein the first transaction feature vector is obtained by concatenating the historical transaction feature vector and at least one financial event feature vector; weighting at least one first transaction feature vector using a self-attention layer to obtain at least one second transaction feature vector; and sorting at least one second transaction feature vector using a feature sequence construction unit to obtain a transaction feature vector sequence corresponding to the first object.

[0071] The feature concatenation unit can be used to concatenate historical transaction feature vectors and financial event feature vectors, merging features from two different sources into a unified vector. The logic of feature concatenation based on the feature concatenation unit typically involves direct concatenation along the feature dimensions. For example, a historical transaction feature vector of length n can be concatenated with at least one financial event feature vector of length m to form a new vector of length n + at least one m. This new vector can then be used as the first transaction feature vector. A first transaction feature vector can be formed by concatenating a historical transaction feature vector with at least one determined financial event feature vector.

[0072] The self-attention layer is a core component of the Transformer model, used to dynamically weight the input feature vectors and learn the relationships between them. Based on this layer, attention weights can be calculated between historical transaction feature vectors and at least one financial event feature vector in the first transaction feature vector. The first transaction feature vector is then updated according to these attention weights, and the updated feature vector is used as the second transaction feature vector. Compared to the first transaction feature vector, the second transaction feature vector captures the degree of correlation between features through a self-attention mechanism, rather than simply concatenating them. The second transaction feature vector more accurately reflects the interaction between trading behavior and external events.

[0073] The feature sequence construction unit can be used to sort at least one first transaction feature vector according to a specific logic (such as time order, correlation strength, etc.) to form an ordered sequence, ensuring that the temporal or logical relationship of the feature elements is preserved. The transaction feature vector sequence can be a set of first transaction feature vectors arranged in a specific order, which contains the fusion features of the historical transaction behavior of the first object and financial anomalies, and preserves the temporal or logical relationship in the sequence.

[0074] In specific implementation, at least one preprocessed historical transaction and at least one first media resource can be input into the feature sequence construction sub-model. Further, features can be extracted from at least one historical transaction based on the transaction feature extraction module in the feature sequence construction sub-model, resulting in a historical transaction feature vector corresponding to the at least one historical transaction. Also, at least one first media resource can be encoded and embedded based on the encoding embedding module in the feature sequence construction sub-model, resulting in a financial event feature vector corresponding to the at least one first media resource. Further, at least one historical transaction feature vector and at least one financial event feature vector can be input into the feature sequence construction module. Then, for at least one historical transaction, the historical transaction feature vector and at least one financial event feature vector corresponding to the historical transaction can be concatenated based on the feature concatenation unit in the feature sequence construction module, resulting in a first transaction feature vector corresponding to the historical transaction. This yields the first transaction feature vector corresponding to at least one historical transaction. Further, each first transaction feature vector can be weighted based on a self-attention layer, resulting in a second transaction feature vector corresponding to the first transaction feature vector. Furthermore, at least one second transaction feature vector can be sorted according to time order by the feature sequence construction unit, and the sorted sequence can be used as the transaction feature vector sequence corresponding to the first object.

[0075] S230. The transaction feature vector sequence is processed based on the abnormal risk probability prediction sub-model to predict the abnormal risk probability corresponding to at least one future transaction of the first object within the future time period, so that when the business platform receives the future transaction, the future transaction is processed according to the abnormal risk probability corresponding to the future transaction.

[0076] The anomaly risk probability prediction sub-model can be used to predict the risk probability of future transactions based on the input feature sequence. The anomaly risk probability prediction sub-model can be a neural network model of any structure, optionally at least one of Long Short-Term Memory networks, gated recurrent units, Transformer models, and convolutional neural networks. Preferably, the model structure of the anomaly risk probability prediction sub-model can be a Transformer model. Based on the self-attention mechanism in the Transformer model, it can simultaneously capture feature associations at any position in the sequence, not limited to dependencies between adjacent time sequences. Furthermore, the efficient computational capability of the Transformer model supports real-time processing, meeting the needs of high-concurrency scenarios.

[0077] In practical implementation, after the feature sequence construction sub-model outputs the transaction feature vector sequence, this sequence can be input into the anomaly risk probability sub-model. Furthermore, the anomaly risk probability sub-model can predict the anomaly risk probability based on the transaction feature vector sequence, obtaining the anomaly risk probability corresponding to at least one future transaction of the first object within a future timeframe. Consequently, when the business platform receives a future transaction, it can process the future transaction based on the corresponding anomaly risk probability.

[0078] The technical solution of this invention processes at least one historical transaction and at least one first media resource based on a feature sequence-based sub-model to obtain a transaction feature vector sequence corresponding to a first object. Further, it processes the transaction feature vector sequence based on an anomaly risk probability prediction sub-model to predict the anomaly risk probability corresponding to at least one future transaction of the first object within a future timeframe. This achieves the integration of historical transaction and media resource information through the feature sequence-based sub-model, transforming multi-source data into a structured feature sequence, thus improving the comprehensiveness of risk prediction information. Furthermore, by leveraging the anomaly risk probability prediction sub-model for in-depth analysis of sequence features, it accurately captures the correlation between transaction behavior and external events, improving the accuracy of future transaction risk prediction. Moreover, the two sub-models work together to automate the transformation from raw data to risk probability, improving the efficiency and intelligence of risk prediction and providing a reliable basis for real-time risk handling on the business platform.

[0079] Figure 3 This is a schematic diagram of the structure of a transaction risk prediction device provided in an embodiment of the present invention. Figure 3 As shown, the device includes a transaction information acquisition module 310 and a transaction risk prediction module 320.

[0080] The transaction information acquisition module 310 is used to acquire at least one historical transaction information of at least one first object within a historical period prior to the current time on the business platform, as well as at least one first media resource associated with a financial anomaly published within the historical period; wherein the first object is an object with transaction records on the business platform within the historical period; the transaction risk prediction module 320 is used to predict the probability of abnormal risk corresponding to at least one future transaction executed by the first object at a future time, based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model, so as to process the future transaction according to the probability of abnormal risk corresponding to the future transaction when the business platform receives the future transaction; wherein the transaction risk prediction model is trained on a machine learning model based on the historical transaction information of multiple sample objects, historical financial event data, and the actual transaction status of the transactions that occurred by the sample objects at the prediction time.

[0081] The technical solution of this invention obtains at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, as well as at least one first media resource associated with a financial anomaly published within the first historical period. Further, for at least one first object, based on the at least one historical transaction information, the at least one first media resource, and a pre-trained transaction risk prediction model, the probability of an anomaly risk corresponding to at least one future transaction executed by the first object within a future period is predicted. When the business platform receives a future transaction, it processes the future transaction based on the probability of an anomaly risk corresponding to the future transaction. This solves the problems of time-consuming and labor-intensive anomaly transaction prediction methods in related technologies, which are prone to missed or false judgments and have low automation levels. It achieves the effect of predicting the risk probability of future transactions by combining historical transaction and external financial event information, improving prediction accuracy, reducing false and missed judgment rates, and enhancing the automation level of anomaly transaction prediction, reducing manual intervention, and improving the efficiency of anomaly transaction risk prediction. Furthermore, by predicting future transaction risks in advance, the business platform can handle them in a targeted manner, enhancing the efficiency and accuracy of risk prevention and control.

[0082] Optionally, the transaction risk prediction model includes a feature sequence construction sub-model and an anomaly risk probability prediction sub-model; the transaction risk prediction module 320 includes a vector sequence construction sub-module and an anomaly risk probability prediction sub-module. The vector sequence construction sub-module is used to process at least one historical transaction information and at least one of the first media resources based on the feature sequence construction sub-model to obtain a transaction feature vector sequence corresponding to the first object; the anomaly risk probability prediction sub-module is used to process the transaction feature vector sequence based on the anomaly risk probability prediction sub-model to predict the anomaly risk probability corresponding to at least one future transaction of the first object within a future time period.

[0083] Optionally, the feature sequence construction sub-model includes a transaction feature extraction module, an encoding embedding module, and a feature sequence construction module; the vector sequence construction sub-module includes a transaction feature extraction unit, a media resource feature extraction unit, and a vector sequence construction unit. Specifically, the transaction feature extraction unit is used to extract features from at least one historical transaction based on the transaction feature extraction module to obtain a historical transaction feature vector corresponding to at least one historical transaction; and the media resource feature extraction unit is used to perform encoding embedding processing on at least one first media resource based on the encoding embedding module to obtain a financial event feature vector corresponding to at least one first media resource; the vector sequence construction unit is used to process at least one historical transaction feature vector and at least one financial event feature vector based on the feature sequence construction module to obtain a transaction feature vector sequence corresponding to the first object.

[0084] Optionally, the feature sequence construction module includes a feature concatenation unit, a self-attention layer, and a feature sequence construction unit; the vector sequence construction unit includes a vector concatenation subunit, a vector weighting subunit, and a sequence construction subunit. The vector concatenation subunit is used to concatenate at least one historical transaction feature vector and at least one financial event feature vector based on the feature concatenation unit to obtain at least one first transaction feature vector; wherein the first transaction feature vector is obtained by concatenating the historical transaction feature vector and at least one financial event feature vector; the vector weighting subunit is used to weight at least one first transaction feature vector based on the self-attention layer to obtain at least one second transaction feature vector; the sequence construction subunit is used to sort at least one second transaction feature vector based on the feature sequence construction unit to obtain a transaction feature vector sequence corresponding to the first object.

[0085] Optionally, the transaction risk prediction module 320 includes: a transaction release unit, a light authentication unit, a reinforced authentication unit, and a transaction interception unit. The transaction release unit is used to release the future transaction if the probability of the abnormal risk is within a preset probability range corresponding to a low-risk level; the light authentication unit is used to perform light authentication on the future transaction if the probability of the abnormal risk is within a preset probability range corresponding to a low-to-medium risk level; the reinforced authentication unit is used to perform reinforced authentication on the future transaction if the probability of the abnormal risk is within a preset probability range corresponding to a high-risk level; and the transaction interception unit is used to intercept the future transaction if the probability of the abnormal risk is within a preset probability range corresponding to a high-risk level.

[0086] Optionally, the apparatus further includes a model training module. The model training module is used to train a transaction risk prediction model.

[0087] The model training module includes: a training sample acquisition unit, an anomaly risk probability prediction unit, a model parameter correction unit, and a model verification unit. The training sample acquisition unit acquires multiple training samples, which include historical transaction information of multiple sample objects within a second historical time period, sample media resources associated with historical financial events, and the actual risk status of transactions occurring within the prediction time period for the sample objects. The anomaly risk probability prediction unit inputs the historical transaction information and sample media resources from the training samples into a pre-built machine learning model to obtain the predicted anomaly risk probability corresponding to each training sample. The model parameter correction unit determines a loss value based on the predicted anomaly risk probability and the actual risk status in the training samples, corrects the model parameters in the machine learning model based on the loss value, and uses the convergence of the loss function in the machine learning model as the training objective to obtain at least one model to be verified. The model verification unit verifies at least one model to be verified based on test samples and uses the model whose verification results meet preset conditions as the transaction risk prediction model.

[0088] Optionally, the apparatus further includes a model parameter update module. This model parameter update module is used to deploy the transaction risk prediction model in the business platform and update the model parameters of the transaction risk prediction model based on the historical transaction information of the first object, the first media resource associated with the financial anomaly, the processing results of the future transaction, and the response results.

[0089] The transaction risk prediction device provided in this embodiment of the invention can execute the transaction risk prediction method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0090] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, 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 merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0091] like Figure 4 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0092] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0093] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as trading risk prediction methods.

[0094] In some embodiments, the transaction risk prediction method may be implemented as a computer program tangibly contained 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 on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the transaction risk prediction method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the transaction risk prediction method by any other suitable means (e.g., by means of firmware).

[0095] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0096] Computer programs 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, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0097] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0098] To provide interaction with a user, the systems and techniques described herein 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 the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the 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 can be received in any form (including sound input, voice input, or tactile input).

[0099] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), target blockchain networks, and the Internet.

[0100] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0101] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.

[0102] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0103] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for predicting transaction risk, characterized in that, include: Obtain at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, and at least one first media resource related to a financial anomaly published within the first historical period; wherein, the first object is an object that has a transaction record on the business platform within the first historical period; For at least one first object, based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model, the probability of abnormal risk corresponding to at least one future transaction executed by the first object within a future time period is predicted, so that when the business platform receives the future transaction, the future transaction is processed according to the probability of abnormal risk corresponding to the future transaction. The transaction risk prediction model is trained on a machine learning model based on historical transaction information of multiple sample objects, sample media resources associated with historical financial events, and the actual risk status of the transactions of the sample objects at the prediction time.

2. The transaction risk prediction method according to claim 1, characterized in that, The transaction risk prediction model includes a feature sequence construction sub-model and an abnormal risk probability prediction sub-model; The step of predicting the probability of abnormal risk corresponding to at least one future transaction of the first object within a future time period based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model includes: Based on the feature sequence, a sub-model is constructed to process at least one historical transaction information and at least one first media resource to obtain a transaction feature vector sequence corresponding to the first object; Based on the abnormal risk probability prediction sub-model, the transaction feature vector sequence is processed to predict the abnormal risk probability corresponding to at least one future transaction of the first object within a future time period.

3. The transaction risk prediction method according to claim 2, characterized in that, The feature sequence construction sub-model includes a transaction feature extraction module, an encoding embedding module, and a feature sequence construction module; The sub-model constructed based on the feature sequence processes at least one historical transaction information and at least one first media resource to obtain a transaction feature vector sequence corresponding to the first object, including: Based on the transaction feature extraction module, features are extracted from at least one of the historical transaction information to obtain a historical transaction feature vector corresponding to the at least one of the historical transaction information; and... Based on the encoding and embedding module, at least one of the first media resources is encoded and embedded to obtain a financial event feature vector corresponding to at least one of the first media resources. The feature sequence construction module processes at least one of the historical transaction feature vectors and at least one of the financial event feature vectors to obtain a transaction feature vector sequence corresponding to the first object.

4. The transaction risk prediction method according to claim 3, characterized in that, The feature sequence construction module includes a feature concatenation unit, a self-attention layer, and a feature sequence construction unit; the step of processing at least one historical transaction feature vector and at least one financial event feature vector based on the feature sequence construction module to obtain a transaction feature vector sequence corresponding to the first object includes: Based on the feature splicing unit, at least one of the historical transaction feature vectors and at least one of the financial event feature vectors are spliced ​​together to obtain at least one first transaction feature vector; wherein, the first transaction feature vector is obtained by splicing the historical transaction feature vectors and at least one of the financial event feature vectors; Based on the self-attention layer, at least one first transaction feature vector is weighted to obtain at least one second transaction feature vector; Based on the feature sequence construction unit, at least one second transaction feature vector is sorted to obtain a transaction feature vector sequence corresponding to the first object.

5. The transaction risk prediction method according to claim 1, characterized in that, The process of processing the future transaction based on the probability of abnormal risk corresponding to the future transaction includes: If the probability of the abnormal risk is within the probability range corresponding to the preset low-risk level, the future transaction will be allowed to proceed. If the probability of the abnormal risk is within the probability range corresponding to the preset low to medium risk level, the future transaction will be subject to light authentication processing. If the probability of the abnormal risk is within the probability range corresponding to the preset medium-high risk level, the future transaction will be subject to enhanced authentication processing. If the probability of the abnormal risk is within the probability range corresponding to the preset high-risk level, the future transaction will be intercepted.

6. The transaction risk prediction method according to claim 1, characterized in that, Also includes: The trading risk prediction model was trained. The training yields a trading risk prediction model, including: Multiple training samples are obtained, wherein the training samples include historical transaction information of multiple sample objects within a second historical period, sample media resources associated with historical financial events, and the actual risk status of transactions of the sample objects within the prediction period; For multiple training samples, the historical transaction information and sample media resources in the training samples are input into a pre-built machine learning model to obtain the predicted abnormal risk probability corresponding to the training samples; Based on the predicted abnormal risk probability and the actual risk status in the training samples, a loss value is determined, and the model parameters in the machine learning model are corrected according to the loss value. The convergence of the loss function in the machine learning model is taken as the training objective to obtain at least one model to be verified. The verification process is performed on at least one of the models to be verified based on test samples, and the model to be verified that meets the preset conditions is used as the transaction risk prediction model.

7. The transaction risk prediction method according to claim 6, characterized in that, Also includes: The transaction risk prediction model is deployed in the business platform, and the model parameters of the transaction risk prediction model are updated based on the historical transaction information of the first object, the first media resources associated with the financial anomaly, the processing results of the future transaction, and the response results.

8. A transaction risk prediction device, characterized in that, include: The transaction information acquisition module is used to acquire at least one historical transaction information of at least one first object within a first historical period prior to the current moment on the business platform, and at least one first media resource related to a financial anomaly published within the first historical period; wherein, the first object is an object that has a transaction record on the business platform within the first historical period; The transaction risk prediction module is used to predict the probability of abnormal risk for at least one future transaction executed by at least one first object within a future time period, based on at least one historical transaction information of the first object, at least one first media resource, and a pre-trained transaction risk prediction model. This is so that when the business platform receives the future transaction, it can process the future transaction based on the probability of abnormal risk corresponding to the future transaction. The transaction risk prediction model is trained on a machine learning model based on historical transaction information of multiple sample objects, sample media resources associated with historical financial events, and the actual risk status of the transactions occurring at the prediction time for the sample objects.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the transaction risk prediction method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the transaction risk prediction method according to any one of claims 1-7.

Images