Data processing method, device and computer-readable storage medium

Abstract

The present invention discloses a data processing method, apparatus, and computer-readable storage medium. The method comprises: obtaining target data, wherein the target data is used to represent data related to a target object; and predicting the target data using a target model to obtain a target probability for the target object, wherein the target model includes multiple different types of pooling modules, and the target probability is used to represent the probability that the target object is at risk. The present invention solves the technical problem of low accuracy in determining whether a user is at risk in the prior art.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a data processing method, device, and computer-readable storage medium. Background Art

[0002] Generally speaking, when users handle financial business, they only need to provide materials related to the financial business they need to handle and wait for approval. However, the current method of judging whether a user has financial risks only based on the materials provided by the user or personal historical information is too simple. It may miss a lot of important judgment information, resulting in an incomplete judgment process, and thus resulting in poor accuracy in judging whether a user has financial risks.

[0003] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0004] Embodiments of the present invention provide a data processing method, apparatus, and computer-readable storage medium to at least solve the technical problem of low accuracy in determining whether a user has financial risks in the prior art.

[0005] Used to represent the relationship between the target user and other objects.

[0006] Optionally, using a target prediction model to process the target data to obtain a target probability of the target object includes: extracting features from the target data to obtain a first feature and at least one second feature, wherein the first feature is used to characterize the features of the target object, the second feature is used to characterize the features of other objects, and the other objects are used to characterize objects that have an associated relationship with the target object; using multiple pooling modules to pool the first feature and at least one second feature to obtain multiple third features, wherein the pooling modules correspond one-to-one to the third features; and obtaining the target probability based on the first feature and the multiple third features.

[0007] Optionally, performing feature extraction on the target data to determine the first feature and at least one second feature includes: performing feature extraction on the target data to obtain the first feature and at least one second initial feature; obtaining at least one first similarity based on the first feature and the at least one second initial feature, wherein the first similarity is used to characterize the similarity between the first feature and the second initial feature; and multiplying the second initial feature by the first similarity associated with the second initial feature to determine the at least one second feature.

[0008] Optionally, obtaining the target probability based on the first feature and multiple third features includes: performing feature fusion on the first feature and multiple third features to obtain initial features; performing dimensionality reduction processing on the initial features to obtain target features; and obtaining the target probability based on the target features.

[0009] Optionally, obtaining the target probability based on the target feature includes: normalizing the target feature to obtain the target probability.

[0010] Optionally, the multiple different types of pooling modules include: a global maximum pooling module, a global minimum pooling module and a global average pooling module.

[0011] Optionally, the method further includes: constructing a loss function based on the target probability; and optimizing the target model based on the loss function.

[0012] According to another aspect of an embodiment of the present invention, a data processing device is also provided, including: an acquisition module for acquiring target data, wherein the target data is used to characterize data related to a target object; a prediction module for predicting the target data using a target model to obtain a target probability of the target object, wherein the target model includes multiple different types of pooling modules, and the target probability is used to characterize the probability that there is a risk in the target object.

[0013] Optionally, the target data includes target object data and other object data, and the acquisition module includes: a first acquisition unit, used to acquire the target object data; a second acquisition unit, used to acquire other object data based on a preset relationship and the target object data, wherein the preset relationship is used to characterize the association relationship between the target user and other objects.

[0014] Optionally, the prediction module includes: a feature extraction unit, used to extract features from the target data to obtain a first feature and at least one second feature, wherein the first feature is used to characterize the features of the target object, the second feature is used to characterize the features of other objects, and the other objects are used to characterize objects that have an association relationship with the target object; a feature pooling unit, used to pool the first feature and at least one second feature using multiple pooling modules to obtain multiple third features, wherein the pooling modules correspond one-to-one to the third features; a target probability determination unit, used to obtain the target probability based on the first feature and the multiple third features.

[0015] Optionally, the feature extraction unit includes: a feature extraction subunit, used to extract features from the target data to obtain a first feature and at least one second initial feature; a similarity calculation subunit, used to obtain at least one first similarity based on the first feature and at least one second initial feature, wherein the first similarity is used to characterize the similarity between the first feature and the second initial feature; and a feature determination subunit, used to multiply the second initial feature by the first similarity related to the second initial feature to determine at least one second feature.

[0016] Optionally, the target probability determination unit includes: a feature fusion subunit, used to perform feature fusion on the first feature and multiple third features to obtain an initial feature; a feature dimension reduction subunit, used to perform dimension reduction processing on the initial feature to obtain a target feature; and a probability determination subunit, used to obtain a target probability based on the target feature.

[0017] Optionally, the target determination subunit is further configured to perform normalization processing on the target features to obtain a target probability.

[0018] Optionally, the pooling module in the prediction module includes: a global maximum pooling module, a global minimum pooling module and a global average pooling module.

[0019] Optionally, the device further includes: a loss function construction module, used to construct a loss function based on the target probability; and a model optimization module, used to optimize the target model based on the loss function.

[0020] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned data processing method.

[0021] According to another aspect of an embodiment of the present invention, an electronic device is also provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned data processing method.

[0022] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is configured to run a program, wherein the program executes the data processing method of the above method embodiment when the program is running.

[0023] In an embodiment of the present invention, a method of acquiring target data and predicting the target data using a target model comprising multiple different types of pooling modules is adopted. The method is not limited to the material data provided by the user, but also combines the social relationship data related to the user to perform multi-mode pooling processing on the acquired data. Finally, a multi-angle prediction of the user's authorization probability is performed, making the prediction result more comprehensive and detailed, so as to improve the accuracy of determining whether the user has financial risks, thereby solving the technical problem of low recognition accuracy in determining whether a user has financial risks in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0025] Figure 1 is a flow chart of a data processing method according to an embodiment of the present invention;

[0026] Figure 2 is a schematic diagram illustrating a data processing principle according to an embodiment of the present invention;

[0027] Figure 3 FIG. 4 is a structural block diagram of a data processing device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] With the development of network technology, more and more network relationships are being beneficially utilized in daily life. Mining the correlation of these network data can effectively improve the user's risk identification ability. The most commonly used method is to use graph convolutional neural network learning, combined with the user's network data to determine whether the user has financial risks. However, the shortcomings of this method are obvious. On the one hand, the performance of the model trained by this method will be greatly affected by big data, especially the process of Attention training of the GAT (Graph Attention Network) model, which is proportional to the data dimension used for training. The training process is time-consuming and may even cause a large amount of memory overflow. On the other hand, this method mostly uses single-angle information to represent the user's risk rate. However, with the decrease in the user's financial overdue rate, it means that the method of considering the problem from a single angle cannot fully and effectively represent the user's risk rate, which also leads to the problem of low accuracy in determining the user's financial risk using this method.

[0031] Example 1

[0032] According to an embodiment of the present invention, an embodiment of a method for data processing is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0033] Figure 1 FIG. 1 is a flow chart of a data processing method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0034] Step S102: acquiring target data, wherein the target data is used to represent data related to the target object.

[0035] The above-mentioned target objects generally refer to users who conduct financial business. The above-mentioned target data generally includes target object data and other object data. The target object data may refer to personal material data provided by the above-mentioned users. Other object data may refer to data related to the target object obtained through other people or things related to the target object. Considering the convenience and efficiency of obtaining data, it can generally be obtained through other people related to the target object.

[0036] Before predicting the probability of a user having financial risk, all relationship data related to the user may be obtained first, including but not limited to the user's social relationship data, personal material data, etc.

[0037] Optionally, acquiring the target data includes: acquiring target object data; and acquiring other object data according to a preset relationship and the target object data, wherein the preset relationship is used to represent an association relationship between the target user and other objects.

[0038] The aforementioned preset relationships generally include, but are not limited to, social relationships, purchasing relationships, geographic location information, and WiFi connection relationships related to the target object. Other objects can be friends of the target object, users who purchased the same item as the target object, users who live in the same area as the target object, or even users who are connected to the same WiFi signal source as the target object. It should be noted that in actual applications, personnel can mine key information from data related to the target object based on the acquired data, and this application does not impose any restrictions thereon.

[0039] In an optional solution of this embodiment, while receiving the personal material data provided by the user, other material data related to the user can be obtained from other objects based on the user's social relationships, and the other material data can be matched and compared with the personal material data provided by the user applying for financial services, so as to determine the authenticity of the user and the personal material data provided by the user, including whether the user is capable of conducting the financial business, whether false materials are provided, etc., thereby improving the accuracy of risk identification for the user.

[0040] Step S104: Use the target model to predict the target data to obtain the target probability of the target object.

[0041] Among them, the target model includes multiple different types of pooling modules, and the target probability is used to characterize the probability that the target object has risks.

[0042] Since a large amount of inconsistent related data may be obtained when acquiring other object data, resulting in poor precision and low accuracy of the trained target model, in order to improve the predictive ability of the target model, multiple pooling models can be set in the target model to perform pooling processing on the above target data respectively, and then the dimensionality reduction data obtained after processing is integrated to obtain the risk probability of the target object. While extracting effective data to improve the predictive ability of the target model, it also greatly reduces the consumption of computing resources and improves the efficiency of determining user financial risks.

[0043] In an optional solution of this embodiment, in order to confirm the risk of the target object in multiple dimensions, multiple different types of pooling modules are used to learn the features extracted from the target data and perform multi-angle risk characterization, wherein the above-mentioned multiple different types of pooling modules may include but are not limited to: global maximum pooling module, global minimum pooling module and global average pooling module, etc.

[0044] Through the above steps, the target data is obtained and the target model containing multiple different types of pooling modules is used to predict the target data. The method is not limited to the material data provided by the user, but also combines the social relationship data related to the user to perform multi-way pooling processing on the obtained data. Finally, the authorization probability of the user is predicted from multiple angles, making the prediction result more comprehensive and detailed, so as to improve the accuracy of determining whether the user has financial risks, thereby solving the technical problem of low recognition accuracy in determining whether a user has financial risks in the existing technology.

[0045] Optionally, using a target prediction model to process the target data to obtain a target probability of the target object includes: extracting features from the target data to obtain a first feature and at least one second feature, wherein the first feature is used to characterize the features of the target object, the second feature is used to characterize the features of other objects, and the other objects are used to characterize objects that have an associated relationship with the target object; using multiple pooling modules to pool the first feature and at least one second feature to obtain multiple third features, wherein the pooling modules correspond one-to-one to the third features; and obtaining the target probability based on the first feature and the multiple third features.

[0046] After obtaining the target object data and other object data, the risk activation method constructed based on the node-based attention mechanism can be used to extract features from these data. The above-mentioned first feature can refer to the feature corresponding to the target object data, and the above-mentioned second feature can refer to the feature corresponding to the other object data.

[0047] In an optional solution of this embodiment, the method of extracting features from the acquired data may include but is not limited to: expected cross entropy, information gain method, etc.

[0048] The third feature mentioned above can refer to a dimensionality reduction feature after pooling. That is, after obtaining the target object data and the other object data, these two data can be pooled to reduce their data dimensionality and improve their data accuracy. To further improve the correlation between the target data and the other data, multiple different types of pooling modules can be used to perform multiple different pooling processes on these data to obtain multiple different types of dimensionality reduction feature data.

[0049] After determining the third feature, the first feature of the target object data can be combined again to perform secondary processing such as feature fusion and secondary dimensionality reduction to obtain a target probability that can reflect whether the user has financial risks.

[0050] Optionally, performing feature extraction on the target data to determine the first feature and at least one second feature includes: performing feature extraction on the target data to obtain the first feature and at least one second initial feature; obtaining at least one first similarity based on the first feature and the at least one second initial feature, wherein the first similarity is used to characterize the similarity between the first feature and the second initial feature; and multiplying the second initial feature by the first similarity associated with the second initial feature to determine the at least one second feature.

[0051] When extracting the first and second features, since the other object data is large and complex, if the model is directly constructed using the extracted second initial features after feature extraction, it may consume a lot of time and memory resources. Therefore, to avoid this problem, the extracted first features can be used to match and filter the second initial features to obtain accurate and concise second features.

[0052] Specifically, since the target object data is generally clear and single, we can first use some preset feature extraction methods to extract features from the target object data to obtain the first feature, and then use the first feature to match and screen the second initial feature, and calculate the feature similarity between any second initial feature and the first feature, that is, the above-mentioned first similarity, and then multiply the first similarity with the corresponding second initial feature to obtain the second feature, so as to improve the correlation between the determined second feature and the first feature, thereby indirectly improving the accuracy of the trained target model.

[0053] For example, if the target object data has a first initial feature A that is "able to complete financial business in full as required", then after obtaining other object data, such as other object data provided by its neighbors, a second initial feature B with a high similarity to the first initial feature A is matched and screened from the other object data as the second feature corresponding to the first feature A, such as "the target object often completes its promises on time", "the target object never lies", etc.

[0054] Optionally, using multiple pooling modules to pool the first feature and at least one second feature to obtain multiple third features includes: obtaining multiple pooling parameters, wherein the pooling parameters include pooling window parameters, and the pooling parameters correspond one-to-one to the pooling modules; based on the multiple pooling parameters, downsampling the first feature and at least one second feature by multiple pooling modules to obtain multiple third features.

[0055] In an optional scheme of this embodiment, the various pooling modules shown in the aforementioned step S102: the global maximum pooling module, the global minimum pooling module, and the global average pooling module, etc., can be directly connected to the Dense layer (fully connected layer) without setting pooling parameters, and deep propagation is performed through the Dense layer to improve the adaptability and operating efficiency of the target model finally constructed, thereby making the determined results more accurate.

[0056] In an optional scheme of this embodiment, in addition to the above-mentioned multiple pooling modules, multiple different types of pooling modules may also include: a maximum pooling module, a minimum pooling module and an average pooling module. The above-mentioned pooling modules can downsample the first feature and at least one second feature after dividing the pooling window to obtain multiple third features. In this application, the pooling window size can be taken as 1×n, which is not specifically limited here. The specific pooling process can refer to relevant literature.

[0057] It should be noted that in the process of pooling, the Attention mechanism can be introduced, that is, some important features can be preset, such as "task completion rate of the target object", "current status of the target object", etc., and higher weight coefficients can be set for these features, so that in the process of pooling, features with high weight coefficients are limitedly considered to ensure that the third feature after feature dimensionality reduction includes at least the above-mentioned important types of features.

[0058] In an optional solution of this embodiment, different pooling window sizes, moving steps, etc. can be set for different pooling modules, without specific limitation.

[0059] In an optional solution of this embodiment, when determining multiple different types of pooling modules, a different Attention weight coefficient can be set for each pooling module to improve the diversity of the determined third day and to improve the versatility of the determined target model.

[0060] Optionally, obtaining the target probability based on the first feature and multiple third features includes: performing feature fusion on the first feature and multiple third features to obtain initial features; performing dimensionality reduction processing on the initial features to obtain target features; and obtaining the target probability based on the target features.

[0061] Since what is being determined is whether there is a risk in the target object, after obtaining the third feature, the first and third features mentioned above can be fused. On the one hand, after the feature fusion, the processing calculation amount can be reduced to ensure the efficiency of real-time determination of the target probability; on the other hand, the compressed features required for determining the target probability can be given to the maximum extent to ensure the accuracy of the determined target probability.

[0062] In an optional embodiment of this embodiment, the first feature and the third feature can be fused by using AGG (Attention Guided Graph) Activation to improve the correlation between the fused compressed features and the financial business conducted by the target object, thereby improving the accuracy of the determined target features.

[0063] In an optional solution of this embodiment, after performing the above-mentioned feature fusion to determine the compressed features, the compressed features can be further subjected to dimensionality reduction processing to obtain the final target features, further refine the compressed features, and improve the efficiency of determining the target probability.

[0064] Optionally, obtaining the target probability based on the target feature includes: normalizing the target feature to obtain the target probability.

[0065] In an optional solution of this embodiment, the above target features may be normalized using a softmax activation function (normalization function).

[0066] Optionally, the method further includes: constructing a loss function based on the target probability; and optimizing the target model based on the loss function.

[0067] After determining the target probability, the target model can be optimized using a loss function to determine the cumulative error in the previous steps to ensure that the determined target probability is more consistent with the actual situation. For specific methods, please refer to relevant literature and no specific limitations are given.

[0068] In an optional solution of this embodiment, the above-mentioned loss function may refer to a cross-entropy loss function. Generally, the Risk Prop (risk support value) in the aforementioned step can be predicted using a softmax activation function, and then the above-mentioned cross-entropy loss function can be constructed using RiskProp. For the specific construction method, please refer to relevant literature.

[0069] In order to facilitate the understanding of the above data processing method, Figure 2 FIG. 1 is a schematic diagram showing a data processing principle according to an embodiment of the present invention. Figure 2 As shown, the data processing system can be divided into four parts, including: a feature extraction part 202, a pooling processing part 204, a probability determination part 206 and a probability optimization part 208.

[0070] Specifically, in the feature extraction part 202, the first feature and the second feature can be extracted from the obtained multiple corresponding target object data and other object data through Risk Activation recognition; in the pooling processing part 204, different Attention weight coefficients can be set for the extracted first and second features, and multiple different pooling processes can be performed to obtain multiple third features after preliminary dimensionality reduction; in the probability determination part 206, the first and third features can be subjected to feature fusion processing, activated by AGG Activation, and normalized to determine the target probability; finally, in the probability optimization part 208, the cross entropy loss function can be constructed by Risk Prop to optimize the determined target probability and further improve its accuracy.

[0071] By combining this data processing method with the additional material data provided by other subjects related to the target, and matching and comparing this additional material data with the personal material data provided by the target, we can effectively prevent the target subject from falsifying their personal material data, which could lead to the risk identification model making incorrect risk identification judgments. This method abandons traditional fixed, single-formula identification methods, such as simply judging whether the user-provided materials are complete and meet regulatory requirements. This allows the risk identification model to adopt a more human-like thinking mode, taking into account more comprehensive factors when determining the risk probability of the target subject, and improving the risk identification model's ability to identify risks from multiple perspectives.

[0072] Example 2

[0073] According to another aspect of the embodiments of the present invention, corresponding to the embodiments of the aforementioned data processing method, the present invention also provides a data processing device. The specific implementation and application scenarios are the same as those of the aforementioned embodiments and are not described in detail here. The serial numbers of the aforementioned embodiments of the present invention are for descriptive purposes only and do not represent the advantages or disadvantages of the embodiments.

[0074] Please refer to Figure 3 , Figure 3 This is a structural block diagram of a data processing device according to an embodiment of the present invention, which includes: an acquisition module 302 for acquiring target data, wherein the target data is used to characterize data related to a target object; a prediction module 304 for predicting the target data using a target model to obtain a target probability of the target object, wherein the target model includes multiple different types of pooling modules, and the target probability is used to characterize the probability that a risk exists in the target object.

[0075] Optionally, the target data includes target object data and other object data, and the acquisition module 302 includes: a first acquisition unit, used to acquire the target object data; a second acquisition unit, used to acquire other object data based on a preset relationship and the target object data, wherein the preset relationship is used to characterize the association relationship between the target user and other objects.

[0076] Optionally, the prediction module 304 includes: a feature extraction unit, used to extract features from the target data to obtain a first feature and at least one second feature, wherein the first feature is used to characterize the features of the target object, the second feature is used to characterize the features of other objects, and the other objects are used to characterize objects that have an association relationship with the target object; a feature pooling unit, used to pool the first feature and at least one second feature using multiple pooling modules to obtain multiple third features, wherein the pooling modules correspond one-to-one to the third features; a target probability determination unit, used to obtain the target probability based on the first feature and the multiple third features.

[0077] Optionally, the feature extraction unit includes: a feature extraction subunit, used to extract features from the target data to obtain a first feature and at least one second initial feature; a similarity calculation subunit, used to obtain at least one first similarity based on the first feature and at least one second initial feature, wherein the first similarity is used to characterize the similarity between the first feature and the second initial feature; and a feature determination subunit, used to multiply the second initial feature by the first similarity related to the second initial feature to determine at least one second feature.

[0078] Optionally, the target probability determination unit includes: a feature fusion subunit, used to perform feature fusion on the first feature and multiple third features to obtain an initial feature; a feature dimension reduction subunit, used to perform dimension reduction processing on the initial feature to obtain a target feature; and a probability determination subunit, used to obtain a target probability based on the target feature.

[0079] Optionally, the target determination subunit is further configured to perform normalization processing on the target features to obtain a target probability.

[0080] Optionally, the pooling module in the prediction module 304 includes: a global minimum pooling module and a global average pooling module.

[0081] Optionally, the device further includes: a loss function construction module, used to construct a loss function based on the target probability; and a model optimization module, used to optimize the target model based on the loss function.

[0082] Example 3

[0083] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to perform the data processing method of the above method embodiment.

[0084] Example 4

[0085] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is configured to run a program, wherein the program executes the data processing method of the above method embodiment when the program is running.

[0086] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0087] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0088] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0089] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected to achieve the purpose of the present embodiment according to actual needs.

[0090] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0091] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0092] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A data processing method, characterized in that: include: Acquiring target data, wherein the target data is used to represent data related to a target object, the target data includes target object data and other object data, and acquiring the target data includes: acquiring the target object data; acquiring the other object data based on a preset relationship and the target object data, wherein the preset relationship is used to represent an association relationship between the target user and other objects, and the other object data is used to verify the authenticity of the target object data by matching and comparing with the target object data; The target data is predicted using a target model to obtain a target probability of the target object, wherein the target model includes multiple different types of pooling modules, the target probability is used to characterize the probability that the target object has a risk, the target probability is obtained based on a first feature and multiple third features, the multiple third features are obtained by pooling the first feature and at least one second feature through multiple pooling modules, the first feature and the at least one second feature are obtained by performing feature extraction on the target data, the first feature is used to characterize the feature corresponding to the target object, the second feature is used to characterize the feature corresponding to the other object data, and the third feature is used to characterize the dimensionality reduction feature after pooling the first feature and the at least one second feature, and the pooling module includes at least: a global maximum pooling module, a global minimum pooling module, and a global average pooling module; The method further includes: performing feature fusion on the first feature and the multiple third features to obtain an initial feature; performing dimensionality reduction processing on the initial feature to obtain a target feature; and obtaining the target probability based on the target feature.

2. The method according to claim 1, characterized in that Processing the target data using a target prediction model to obtain the target probability of the target object includes: Performing feature extraction on the target data to obtain a first feature and at least one second feature, wherein the first feature is used to characterize a feature of the target object, the second feature is used to characterize a feature of other objects, and the other objects are used to characterize objects that have an associated relationship with the target object; pooling the first feature and the at least one second feature using a plurality of the pooling modules to obtain a plurality of third features, wherein the pooling modules correspond one to one to the third features; The target probability is obtained based on the first feature and the multiple third features.

3. The method according to claim 2, characterized in that Performing feature extraction on the target data to determine a first feature and at least one second feature includes: Performing feature extraction on the target data to obtain the first feature and at least one second initial feature; Obtaining at least one first similarity based on the first feature and at least one second initial feature, wherein the first similarity is used to represent a similarity between the first feature and the second initial feature; The second initial feature is multiplied by a first similarity associated with the second initial feature to determine at least one second feature.

4. The method according to claim 1, wherein Obtaining the target probability based on the target feature includes: Normalizing the target features to obtain the target probability.

5. The method according to claim 1, wherein The method further comprises: Based on the target probability, construct a loss function; The target model is optimized based on the loss function.

6. A data processing device, characterized in that: include: an acquisition module, configured to acquire target data, wherein the target data is used to represent data related to a target object, and the target data includes target object data and other object data. The acquisition module comprises: a first acquisition unit, configured to acquire the target object data; and a second acquisition unit, configured to acquire the other object data based on a preset relationship and the target object data, wherein the preset relationship is used to represent an association relationship between a target user and other objects, and the other object data is used to verify the authenticity of the target object data by matching and comparing with the target object data. A prediction module is used to predict the target data using a target model to obtain a target probability of the target object, wherein the target model includes multiple different types of pooling modules, the target probability is used to characterize the probability that the target object has a risk, and the target probability is obtained based on a first feature and multiple third features, the multiple third features are obtained by pooling the first feature and at least one second feature through multiple pooling modules, the first feature and the at least one second feature are obtained by feature extraction of the target data, the first feature is used to characterize the feature corresponding to the target object, the second feature is used to characterize the feature corresponding to the other object data, and the third feature is used to characterize the dimensionality reduction feature after pooling the first feature and the at least one second feature, and the pooling module includes at least: a global maximum pooling module, a global minimum pooling module, and a global average pooling module; The device is further configured to: perform feature fusion on the first feature and the plurality of third features to obtain an initial feature; perform dimensionality reduction processing on the initial feature to obtain a target feature; and obtain the target probability based on the target feature.

7. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 5.

Images