Portrayal generation method and device, and storage medium

Abstract

Embodiments of the invention provide a portrait generation method and apparatus, and a storage medium. The method comprises the steps of obtaining historical log data of a target object on at least two platforms; the historical log data comprises at least one group of identification data used for identity verification of the target object on each platform and at least one group of behavior data of the target object on each platform; based on the at least one group of identification data of each platform, constructing initial portraits of the target object on the at least two platforms; and according to an association relationship between the at least one group of identification data of each platform and the at least one group of behavior data of each platform, adding the at least one group of behavior data of each platform to the initial portrait, and obtaining a unified portrait of the target object on the at least two platforms.

Description

Technical Field

[0001] The present application relates to an image generation method, an apparatus, and a storage medium. Background Art

[0002] In a multi-platform or multi-channel business environment, user data is often scattered across different systems and platforms. To provide a consistent user experience and accurate marketing strategies, it is necessary to effectively integrate and optimize this scattered user data. Traditional data integration methods often ignore the complex relationship network among users, resulting in incomplete and inaccurate user portraits. Summary of the Invention

[0003] Embodiments of the present application provide an image generation method, an apparatus, and a storage medium.

[0004] The technical solution of the present application is implemented as follows:

[0005] In a first aspect, embodiments of the present application provide an image generation method, the method including:

[0006] Obtain historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for identity authentication of the target object on each platform, and at least one set of behavior data of the target object on each platform;

[0007] Based on at least one set of identification data of each platform, construct an initial portrait of the target object on at least two platforms; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on at least two platforms.

[0008] In the above image generation method, after obtaining the historical log data of the target object on at least two platforms, the method further includes:

[0009] For each platform, extract a first association relationship between every two identification data in at least one set of identification data on each platform, a second association relationship between every two behavior data in at least one set of behavior data on each platform, and a third association relationship between one identification data in at least one set of identification data and one behavior data in at least one set of behavior data on each platform.

[0010] In the above image generation method, based on at least one set of identification data of each platform, constructing an initial portrait of the target object on at least two platforms includes:

[0011] According to the first association relationship between every two identification data in at least one set of identification data extracted from each platform, construct an identification relationship network corresponding to the target object;

[0012] Extract features from the identification relationship network to obtain the initial portraits of the target object on at least two platforms.

[0013] In the above portrait generation method, extracting features from the identification relationship network to obtain the initial portraits of the target object on at least two platforms includes:

[0014] Determine the amount of data contained in the identification relationship network;

[0015] If the amount of data is greater than the first threshold, then optimize the identification relationship network until a target identification relationship network that meets the first threshold is obtained;

[0016] Extract features from the target identification relationship network to obtain the initial portrait of the target object.

[0017] In the above portrait generation method, the identification relationship network is composed of at least two nodes and the edges associated with at least two nodes. Each node is used to represent at least one identification data on at least two platforms, and the edges associated with at least two nodes are used to represent the association relationship between every two identification data on at least two platforms.

[0018] In the above portrait generation method, optimizing the identification relationship network includes:

[0019] Input the identification relationship network into a preset graph convolutional network model. Use the preset graph convolutional network model to update each node in the identification relationship network by aggregating the features of adjacent nodes, and remove some discrete nodes in the identification relationship network to obtain the first identification relationship network;

[0020] If the amount of data contained in the first identification relationship network is greater than the first threshold, then input the first identification relationship network into the preset graph convolutional network model again until the target identification relationship network is obtained.

[0021] In the above portrait generation method, according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on at least two platforms, including:

[0022] According to the third association relationship of each platform extracted, add the first part of the behavior data related to the identification data in the initial portrait to the corresponding identification data in the initial portrait;

[0023] According to the second association relationship of each platform extracted, add the second part of the behavior data related to the first part of the behavior data to the corresponding identification data in the initial portrait to obtain the unified portrait; at least one set of behavior data on each platform includes the first part of the behavior data and the second part of the behavior data.

[0024] In the above image generation method, the method further includes:

[0025] Sending the unified image to each of at least two platforms simultaneously for each platform to perform operations related to the target object based on the unified image.

[0026] In a second aspect, an embodiment of the present application provides an image generation device, the device includes:

[0027] An acquisition unit, configured to acquire historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used by the target object for authentication on each platform, and at least one set of behavior data of the target object on each platform;

[0028] A construction unit, configured to construct an initial image of the target object on at least two platforms based on at least one set of identification data of each platform; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial image to obtain a unified image of the target object on at least two platforms.

[0029] An embodiment of the present application provides an image generation method, device, and storage medium. The method includes: acquiring historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used by the target object for authentication on each platform, and at least one set of behavior data of the target object on each platform; constructing an initial image of the target object on at least two platforms based on at least one set of identification data of each platform; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial image to obtain a unified image of the target object on at least two platforms. Description of the Drawings

[0030] Figure 1 It is a flowchart of an image generation method provided by an embodiment of the present application;

[0031] Figure 2 It is a flowchart of an exemplary association relationship extraction method provided by an embodiment of the present application;

[0032] Figure 3 It is a flowchart of an exemplary method for constructing an identification relationship network provided by an embodiment of the present application;

[0033] Figure 4 It is a flowchart of an exemplary method for constructing an initial image provided by an embodiment of the present application;

[0034] Figure 5A flowchart of an exemplary optimized identification relationship network method provided by an embodiment of the present application;

[0035] Figure 6 A flowchart of an exemplary unified portrait construction method provided by an embodiment of the present application;

[0036] Figure 7 A flowchart of an exemplary portrait generation process provided by an embodiment of the present application;

[0037] Figure 8 A schematic diagram of an exemplary user relationship network provided by an embodiment of the present application;

[0038] Figure 9 A schematic diagram of the composition structure of a portrait generation device provided by an embodiment of the present application;

[0039] Figure 10 A schematic diagram of the composition structure of a portrait generation device provided by an embodiment of the present application. Detailed implementation manners

[0040] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0041] An embodiment of the present application provides a portrait generation method, which is applied to a portrait generation device. Figure 1 A flowchart of a portrait generation method provided by an embodiment of the present application, as Figure 1 shown, the portrait generation method may include the following steps S100 to step S101:

[0042] Step S100, obtain historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for identity authentication of the target object on each platform, and at least one set of behavior data of the target object on each platform.

[0043] In an embodiment of the present application, the portrait generation device obtains historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for identity authentication of the target object on each platform, and at least one set of behavior data of the target object on each platform.

[0044] In some embodiments, the platform can be understood as a shopping platform where the target object shops, a music platform where the target object listens to music, a video platform where the target object watches videos, or a retrieval platform where the target object queries or retrieves, etc. The specific platform can be selected according to the actual situation, and this embodiment does not limit it here.

[0045] In some embodiments, the identification data can be understood as the login identification used by the target object when logging in to each platform. For example, logging in through the account of a social software, logging in through a mobile phone number, or logging in through the account password assigned by the platform for the target object; since each platform usually supports at least one login method, therefore, the target object has at least one set of identification data for authentication on each platform.

[0046] In some embodiments, for a shopping platform, the behavior data of the target object can be historical purchase data or historical browsing data, etc. on the shopping platform; for a music platform, the behavior data of the target object can be historical music listening data or historical collection data, etc. on the music platform; for a video platform, the behavior data of the target object can be historical playback data or historical download data, etc. on the video platform; for a retrieval platform, the behavior data of the target object can be historical search data or historical click data, etc. on the retrieval platform; thus, the target object has at least one set of behavior data on each platform.

[0047] Step S101: Based on at least one set of identification data of each platform, construct an initial portrait of the target object on at least two platforms; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on at least two platforms.

[0048] In the embodiments of the present application, after the portrait generation device obtains the historical log data of the target object on at least two platforms, based on at least one set of identification data of each platform, construct an initial portrait of the target object on at least two platforms; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on at least two platforms.

[0049] In some embodiments, after obtaining at least one set of identification data of the target object on each of at least two platforms, first, according to at least one set of identification data on each platform, construct an initial portrait of the target object on at least two platforms. At this time, the initial portrait represents the association relationship between all the identification data of the target object on at least two platforms. Then, according to the association relationship between the identification data and the behavior data on each platform, add the behavior data associated with the identification data in the initial portrait to the corresponding identification data in the initial portrait, and finally obtain a unified portrait of the target object on at least two platforms. At this time, the unified portrait is the unity of the identification data and the behavior data of the target object on at least two platforms.

[0050] In an optional implementation manner of the present application, after the portrait generation device obtains the unified portrait of the target object on at least two platforms, the following steps may further be performed: sending the unified portrait to each of the at least two platforms simultaneously for each platform to perform operations related to the target object based on the unified portrait.

[0051] In some embodiments, each platform can more comprehensively and accurately understand the target object according to the unified portrait of the target object on at least two platforms, so as to be able to provide more accurate service support for the target object on this platform.

[0052] In an optional implementation manner of the present application, referring to Figure 2 , after the portrait generation device obtains the historical log data of the target object on at least two platforms, the following step S200 is further included:

[0053] Step S200: For each platform, extract the first association relationship between every two identification data in at least one group of identification data on each platform, the second association relationship between every two behavior data in at least one group of behavior data, and the third association relationship between one identification data in at least one group of identification data and one behavior data in at least one group of behavior data.

[0054] In some embodiments, since there is an association between the identification data of the user on each platform, therefore, the first association relationship between every two identification data can be extracted from at least one group of identification data on each platform, and each platform corresponds to a group of first association relationships. Thus, at least two groups of first association relationships corresponding to at least two platforms are obtained.

[0055] In some embodiments, since there is an association between the behavior data of the user on each platform, therefore, the second association relationship between every two behavior data can be extracted from at least one group of behavior data on each platform, and each platform corresponds to a group of second association relationships. Thus, at least two groups of second association relationships corresponding to at least two platforms are obtained.

[0056] In some embodiments, since there is an association between the behavior data and the identification data of the user on each platform, therefore, one behavior data and one identification data on each platform can be associated to obtain the third association relationship, and each platform corresponds to a group of third association relationships. Thus, at least two groups of third association relationships corresponding to at least two platforms are obtained.

[0057] In an optional implementation manner of the present application, referring to Figure 3 , the portrait generation device constructs the initial portrait of the target object on at least two platforms based on at least one group of identification data of each platform, including the following steps S300 to step S301:

[0058] Step S300: Construct an identification relationship network corresponding to the target object according to the first association relationship between every two pieces of identification data in at least one set of identification data extracted from each platform.

[0059] In an alternative embodiment of the present application, the identification relationship network is composed of at least two nodes and the edges associated among at least two nodes. Each node is used to represent at least one piece of identification data on at least two platforms, and the edges associated among at least two nodes are used to represent the association relationship between every two pieces of identification data on at least two platforms.

[0060] In some embodiments, the identification relationship network can be understood as a relationship cluster composed of multiple nodes. Each node corresponds to an identification information of the target object. The connection relationship between every two nodes is determined according to the first association relationship. The weight corresponding to the edge between every two connected nodes represents the relevance between the two nodes. If there are multiple first association relationships between two pieces of identification data of the target object, at this time, it indicates that the relevance between these two pieces of identification data is relatively large. In the identification relationship network, the weight corresponding to the edge connecting the two nodes corresponding to these two pieces of identification data is relatively large.

[0061] Step S301: Extract features from the identification relationship network to obtain an initial portrait of the target object on at least two platforms.

[0062] In some embodiments, after obtaining the identification relationship network of the target object, since the identification relationship network contains the identification data corresponding to each node, at this time, feature extraction can be performed on the identification data included in the identification relationship network to obtain a vector space corresponding to the identification relationship network, that is, the initial portrait mentioned in this embodiment.

[0063] In an alternative embodiment of the present application, with reference to Figure 4 , the portrait generation device extracts features from the identification relationship network to obtain an initial portrait of the target object on at least two platforms, including the following steps S400 to S402:

[0064] Step S400: Determine the amount of data included in the identification relationship network.

[0065] In some embodiments, after obtaining the identification relationship network, it is necessary to determine whether the identification relationship network needs to be further optimized. The trigger condition for optimization is that the amount of data in the identification relationship network is greater than the first threshold value. Therefore, it is necessary to first determine the amount of data included in the identification relationship network.

[0066] Step S401: If the amount of data is greater than the first threshold value, then optimize the identification relationship network until a target identification relationship network that meets the first threshold value is obtained.

[0067] In some embodiments, when the amount of data included in the identity relationship network is greater than a first threshold value, optimization of the identity relationship network is triggered until a target identity relationship network with a data amount less than or equal to the first threshold value is obtained.

[0068] Step S402: Extract features from the target identity relationship network to obtain an initial portrait of the target object.

[0069] In some embodiments, features are extracted from a target identity relationship network with a data amount less than or equal to the first threshold value to obtain an initial portrait of the target object.

[0070] It can be understood that optimizing the identity relationship network when the amount of data in the identity relationship network is large can reduce the amount of features in the finally obtained initial portrait, thereby reducing the space required for storing the initial portrait.

[0071] In an alternative embodiment of the present application, referring to Figure 5 , the portrait generation device optimizes the identity relationship network, including the following steps S500 to S501:

[0072] Step S500: Input the identity relationship network into a preset graph convolutional network model, and use the preset graph convolutional network model to update each node in the identity relationship network by aggregating the features of adjacent nodes, and remove some discrete nodes in the identity relationship network to obtain a first identity relationship network.

[0073] In some embodiments, the preset graph convolutional network model can be a Cluster-Graph Convolutional Network (Cluster-GCN). Cluster-GCN can be multi-layered, and each layer updates the features of the central node by aggregating the features of neighbor nodes. Based on multi-layer Cluster-GCN, node aggregation is performed on the identity relationship network, discrete nodes are aggregated, and discrete nodes that cannot be aggregated are removed to obtain a processed first identity relationship network.

[0074] It can be understood that through the diagonal enhancement and normalization strategy, multi-layer Cluster-GCN can be trained to avoid the problem of gradient disappearance or explosion in deep network training.

[0075] Step S501: If the amount of data included in the first identity relationship network is greater than the first threshold value, input the first identity relationship network into the preset graph convolutional network model again until a target identity relationship network is obtained.

[0076] In some embodiments, after obtaining the first identification relationship network, the data volume included in the first identification relationship network is determined again. If the data volume included in the first identification relationship network is still greater than the first threshold, then the first identification relationship network is processed again using a preset graph convolutional network model to obtain a second identification relationship network. If the data volume included in the second identification relationship network is less than or equal to the first threshold, then the second identification relationship network is the obtained target identification relationship network. If the data volume included in the second identification relationship network is greater than the first threshold, then the second identification relationship network needs to be processed again using the preset graph convolutional network model until a target identification relationship network that meets the first threshold is obtained.

[0077] In an alternative embodiment of the present application, with reference to Figure 6 , the portrait generation device adds at least one set of behavior data of each platform to the initial portrait according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, to obtain a unified portrait of the target object on at least two platforms, including the following steps S600 to step S601:

[0078] Step S600: According to the third association relationship of each platform extracted, add the first part of the behavior data related to the identification data in the initial portrait to the corresponding identification data in the initial portrait.

[0079] In some embodiments, since there is an association between the behavior data and the identification data of the user on each platform, therefore, based on this association relationship, one behavior data and one identification data on each platform can be associated to obtain a third association relationship. Each platform corresponds to a set of third association relationships. Thus, according to a set of third association relationships corresponding to each platform, the first part of the behavior data related to the identification data in the initial portrait can be added to the corresponding identification data in the initial portrait.

[0080] Step S601: According to the second association relationship of each platform extracted, add the second part of the behavior data related to the first part of the behavior data to the corresponding identification data in the initial portrait to obtain a unified portrait; at least one set of behavior data on each platform includes the first part of the behavior data and the second part of the behavior data.

[0081] In some embodiments, since there is an association between the behavior data of the user on each platform, therefore, based on this association relationship, the second association relationship between every two behavior data can be extracted from at least one set of behavior data on each platform. Each platform corresponds to a set of second association relationships. Thus, according to a set of second association relationships corresponding to each platform, the second part of the behavior data related to the first part of the behavior data in the initial portrait can be added to the corresponding identification data in the initial portrait. Thus, a unified portrait of the target object on at least two platforms is obtained.

[0082] Based on the above embodiments, the present application explains the portrait generation method in the above embodiments through an exemplary portrait generation process, with reference to Figure 7 , including the following steps S701 to S707:

[0083] Step S700, Data collection and preprocessing.

[0084] In some embodiments, collect information such as the behavior data, transaction records, and social interactions of the user (the target object in the above embodiments) on different platforms, and perform data cleaning and formatting.

[0085] Step S701, Construct a user relationship graph.

[0086] In some embodiments, use the collected data to construct a user relationship graph, where the nodes represent a certain identification information of the user, and the edges represent the association relationships between the identification information of the users, with reference to Figure 8 .

[0087] Step S702, Design a graph convolutional network model.

[0088] In some embodiments, design a multi-layer graph convolutional network model, and each layer updates the feature representation of the central node by aggregating the features of neighboring nodes.

[0089] Step S703, Feature learning and optimization.

[0090] In some embodiments, learn the deep feature representation of user nodes through a multi-layer graph convolutional network model, and at the same time optimize the user relationship network by combining the behavior data and attribute information of the users.

[0091] Step S704, User portrait integration.

[0092] In some embodiments, according to the learned user feature representation, integrate the portraits of the user on different platforms to generate a unified and comprehensive user portrait of the user on different platforms.

[0093] Step S705, Deploy the user portrait to the platform.

[0094] In some embodiments, deploy the user portrait to the platform. Each platform can more comprehensively and accurately understand the target object based on the unified portrait of the target object on at least two platforms, so as to be able to provide more accurate service support for the target object on this platform.

[0095] An embodiment of the present application provides an image generation method, which includes: obtaining historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for authentication of the target object on each platform, and at least one set of behavior data of the target object on each platform; constructing an initial image of the target object on at least two platforms based on at least one set of identification data of each platform; adding at least one set of behavior data of each platform to the initial image according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, to obtain a unified image of the target object on at least two platforms; adopting the above implementation solution can better capture and utilize the data of the target object on different platforms, so as to construct a more comprehensive user image and provide more accurate data support for the platform to provide services.

[0096] Based on the above embodiment, in another embodiment of the present application, an image generation device 900 is provided. Figure 9 It is a schematic structural diagram of a composition of an image generation device provided by the present application, as Figure 9 shown, the image generation device 900 includes:

[0097] An acquisition unit 901, configured to obtain historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for authentication of the target object on each platform, and at least one set of behavior data of the target object on each platform;

[0098] A construction unit 902, configured to construct an initial image of the target object on the at least two platforms based on at least one set of identification data of each platform; and add at least one set of behavior data of each platform to the initial image according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, to obtain a unified image of the target object on the at least two platforms.

[0099] In some embodiments, the image generation device 1 further includes: an extraction unit;

[0100] The extraction unit is configured to extract, for each platform, a first association relationship between every two identification data in at least one set of identification data on each platform, a second association relationship between every two behavior data in at least one set of behavior data, and a third association relationship between one identification data in at least one set of identification data and one behavior data in at least one set of behavior data.

[0101] In some embodiments, the building unit 902 is further configured to construct an identification relationship network corresponding to the target object according to the first association relationship between every two pieces of identification data in at least one set of identification data extracted from each of the platforms.

[0102] The extraction unit is further configured to perform feature extraction on the identification relationship network to obtain an initial portrait of the target object on the at least two platforms.

[0103] In some embodiments, the portrait generation device 1 further includes: a determination unit and an optimization unit;

[0104] The determination unit is configured to determine the amount of data included in the identification relationship network;

[0105] The optimization unit is configured to, if the amount of data is greater than a first threshold, optimize the identification relationship network until a target identification relationship network that meets the first threshold is obtained;

[0106] The extraction unit is further configured to perform feature extraction on the target identification relationship network to obtain the initial portrait of the target object.

[0107] In some embodiments, the identification relationship network is composed of at least two nodes and the edges associated among the at least two nodes, each node is used to represent at least one piece of identification data on the at least two platforms, and the edges associated among the at least two nodes are used to represent the association relationship between every two pieces of identification data on the at least two platforms.

[0108] In some embodiments, the portrait generation device 1 further includes: an input unit;

[0109] The input unit is configured to input the identification relationship network into a preset graph convolutional network model, and use the preset graph convolutional network model to update each node in the identification relationship network by aggregating the features of adjacent nodes, and remove some discrete nodes in the identification relationship network to obtain a first identification relationship network;

[0110] The input unit is further configured to, if the amount of data included in the first identification relationship network is greater than the first threshold, input the first identification relationship network into the preset graph convolutional network model again until the target identification relationship network is obtained.

[0111] In some embodiments, the portrait generation device 1 further includes: an addition unit;

[0112] The addition unit is configured to add a first part of behavior data related to the identification data in the initial portrait to the corresponding identification data in the initial portrait according to the third association relationship of each of the platforms extracted.

[0113] The adding unit is further configured to add the second part of behavior data related to the first part of behavior data to the corresponding identification data in the initial portrait according to the second association relationship of each of the platforms extracted, so as to obtain the unified portrait; at least one set of behavior data on each of the platforms includes the first part of behavior data and the second part of behavior data.

[0114] In some embodiments, the portrait generating device 1 further includes: a sending unit;

[0115] The sending unit is configured to send the unified portrait to each of the at least two platforms simultaneously, so that each of the platforms can perform operations related to the target object based on the unified portrait.

[0116] An embodiment of the present application provides a portrait generating device, which includes: an obtaining unit and a constructing unit. The obtaining unit is configured to obtain historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for identity authentication of the target object on each platform and at least one set of behavior data of the target object on each platform; the constructing unit is configured to construct an initial portrait of the target object on at least two platforms based on at least one set of identification data of each platform; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on at least two platforms. By adopting the above implementation solution, it is possible to better capture and utilize the data of the target object on different platforms, thereby constructing a more comprehensive user portrait and providing more accurate data support for the platform to provide services.

[0117] Figure 10 As shown in the following figure, which is a schematic structural diagram of a portrait generating device provided by an embodiment of the present application. In practical applications, based on the same inventive concept of the above embodiments, Figure 10 As shown, the portrait generating device 1000 in this embodiment includes: a processor 1001, a transmitter 1002, a memory 1003, and a communication bus 1004.

[0118] In the process of a specific embodiment, the above-mentioned acquisition unit 1601, construction unit 1602, extraction unit, determination unit, optimization unit, input unit, and addition unit can be implemented by the processor 1001 located on the portrait generation device 1000. The above-mentioned sending unit can be implemented by the transmitter 1002 located on the portrait generation device 1000. The above-mentioned processor 1001 can be at least one of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Image Processing Device (DSPD), a Programmable Logic Image Processing Device (PLD), a Field Programmable Gate Array (FPGA), a CPU, a controller, a microcontroller, and a microprocessor. It can be understood that for different portrait generation devices 1000, the electronic devices used to implement the above-mentioned processor functions can also be others, and this embodiment does not make specific limitations.

[0119] In the embodiment of the present application, the above-mentioned processor 1001 implements the following portrait generation method:

[0120] Obtain historical log data of the target object on at least two platforms; the historical log data includes at least one set of identification data used by the target object for identity authentication on each of the platforms, and at least one set of behavior data of the target object on each of the platforms;

[0121] Based on at least one set of identification data of each platform, construct an initial portrait of the target object on the at least two platforms; according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, add at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on the at least two platforms. In some embodiments, the processor 1001 is further configured to, for each platform, extract a first association relationship between every two identification data in at least one set of identification data on each platform, a second association relationship between every two behavior data in at least one set of behavior data, and a third association relationship between one identification data in at least one set of identification data and one behavior data in at least one set of behavior data.

[0122] In some embodiments, the processor 1001 is further configured to construct an identification relationship network corresponding to the target object according to a first association relationship between every two pieces of identification data in at least one set of identification data extracted from each of the platforms; perform feature extraction on the identification relationship network to obtain an initial portrait of the target object on the at least two platforms.

[0123] In some embodiments, the processor 1001 is further configured to determine the amount of data included in the identification relationship network; if the amount of data is greater than a first threshold, then optimize the identification relationship network until a target identification relationship network that meets the first threshold is obtained; perform feature extraction on the target identification relationship network to obtain the initial portrait of the target object.

[0124] In some embodiments, the identification relationship network is composed of at least two nodes and edges associated among the at least two nodes, each node is used to represent at least one piece of identification data on the at least two platforms, and the edges associated among the at least two nodes are used to represent the association relationship between every two pieces of identification data on the at least two platforms.

[0125] In some embodiments, the processor 1001 is further configured to input the identification relationship network into a preset graph convolutional network model, use the preset graph convolutional network model to update each node in the identification relationship network by aggregating the features of adjacent nodes, and remove some discrete nodes in the identification relationship network to obtain a first identification relationship network; if the amount of data included in the first identification relationship network is greater than the first threshold, then input the first identification relationship network into the preset graph convolutional network model again until the target identification relationship network is obtained.

[0126] In some embodiments, the processor 1001 is further configured to add a first part of behavior data related to the identification data in the initial portrait to the corresponding identification data in the initial portrait according to a third association relationship extracted from each of the platforms; add a second part of behavior data related to the first part of behavior data to the corresponding identification data in the initial portrait according to a second association relationship extracted from each of the platforms to obtain the unified portrait; at least one set of behavior data on each platform includes the first part of behavior data and the second part of behavior data.

[0127] In some embodiments, the processor 1001 is further configured to send the unified portrait to each of the at least two platforms simultaneously for each platform to perform operations related to the target object based on the unified portrait.

[0128] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors. The programs are applied to an image generation device, and the computer program implements the image generation method as described above.

[0129] It should be noted that in this article, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more limitations, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, article or device including that element.

[0130] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present disclosure, in essence, or the part that contributes to the related art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing an image generation device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the image generation methods of various embodiments of the present disclosure.

[0131] The above is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application.

Claims

1. An image generation method, the method comprising: Obtaining historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for authentication of the target object on each of the platforms, and at least one set of behavior data of the target object on each of the platforms; Based on at least one set of identification data of each platform, constructing an initial portrait of the target object on the at least two platforms; According to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform, adding at least one set of behavior data of each platform to the initial portrait to obtain a unified portrait of the target object on the at least two platforms.

2. The method according to claim 1, after obtaining the historical log data of the target object on at least two platforms, the method further comprising: For each platform, extracting a first association relationship between every two pieces of identification data in at least one set of identification data on each platform, a second association relationship between every two pieces of behavior data in at least one set of behavior data, and a third association relationship between one piece of identification data in at least one set of identification data and one piece of behavior data in at least one set of behavior data.

3. The method according to claim 2, the constructing an initial portrait of the target object on the at least two platforms based on at least one set of identification data of each platform, comprising: According to the first association relationship between every two pieces of identification data in at least one set of identification data extracted from each platform, constructing an identification relationship network corresponding to the target object; Performing feature extraction on the identification relationship network to obtain an initial portrait of the target object on the at least two platforms.

4. The method according to claim 3, the performing feature extraction on the identification relationship network to obtain an initial portrait of the target object on the at least two platforms, comprising: Determining the amount of data contained in the identification relationship network; If the amount of data is greater than a first threshold, then optimizing the identification relationship network until a target identification relationship network that meets the first threshold is obtained; Performing feature extraction on the target identification relationship network to obtain the initial portrait of the target object.

5. The method according to claim 4, the identification relationship network is composed of at least two nodes and the associated edges among the at least two nodes, each node is used to represent at least one piece of identification data on the at least two platforms, and the associated edges among the at least two nodes are used to represent the association relationship between every two pieces of identification data on the at least two platforms.

6. The method according to claim 5, the optimizing the identification relationship network, comprising: Inputting the identification relationship network into a preset graph convolutional network model, and using the preset graph convolutional network model to update each node in the identification relationship network by aggregating the features of adjacent nodes, and removing some discrete nodes in the identification relationship network to obtain a first identification relationship network; If the amount of data contained in the first identification relationship network is greater than the first threshold value, the first identification relationship network is input into the preset graph convolutional network model again until the target identification relationship network is obtained.

7. The method according to any one of claims 4-6, wherein adding at least one set of behavior data of each platform to the initial portrait according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform to obtain a unified portrait of the target object on the at least two platforms includes: Adding the first part of the behavior data related to the identification data in the initial portrait to the corresponding identification data in the initial portrait according to the third association relationship of each platform extracted; Adding the second part of the behavior data related to the first part of the behavior data to the corresponding identification data in the initial portrait according to the second association relationship of each platform extracted to obtain the unified portrait; at least one set of behavior data on each platform includes the first part of the behavior data and the second part of the behavior data.

8. The method according to claim 1, the method further includes: Sending the unified portrait to each of the at least two platforms simultaneously for each platform to perform operations related to the target object based on the unified portrait.

9. A portrait generation device, the device includes: An acquisition unit, configured to acquire historical log data of a target object on at least two platforms; the historical log data includes at least one set of identification data used for identity authentication of the target object on each platform, and at least one set of behavior data of the target object on each platform; A construction unit, configured to construct an initial portrait of the target object on the at least two platforms based on at least one set of identification data of each platform; Adding at least one set of behavior data of each platform to the initial portrait according to the association relationship between at least one set of identification data of each platform and at least one set of behavior data of each platform to obtain a unified portrait of the target object on the at least two platforms.

10. A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method according to any one of claims 1-8 is implemented.

Images