Data processing method, device and computer readable storage medium

Abstract

The present invention discloses a data processing method, device and computer-readable storage medium, and relates to the field of data processing. The data processing method comprises: inputting feature data of a new product into a pre-trained feature extraction network, obtaining the output reference features of the new product outputted by the last layer of the feature extraction network, wherein the feature extraction network is trained based on a feature extraction loss function, and the feature extraction loss function is constructed based on the maximum mean difference matching results of the output data of multiple layers of the feature extraction network, and the discrimination results of a discriminant network for distinguishing the source of the output reference features; inputting the output reference features into a pre-trained machine learning model, and obtaining the user interaction index of the new product outputted by the machine learning model. By introducing multi-layer transfer learning in the training of the feature extraction network, the difference problem between the attribute distribution of the new product and the historical attribute distribution of the new product is reduced, and the accuracy of the information prediction of the new product is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to a data processing method, apparatus, and computer-readable storage medium. Background Art

[0002] In the related art, there is a need to pre-determine a user interaction index, which may include, for example, the click volume, view volume, sales volume, etc. of items on an e-commerce website. By predicting this information, the network load situation, warehousing pressure situation, and computing resource consumption situation brought about during the order generation process can be predicted in advance.

[0003] In the related art, the following several determination methods can be adopted. One is the Delphi method, which obtains a prediction result by aggregating the experiences of several experts. Another is time series prediction, which makes predictions through historical data. Still another is machine learning model prediction, which comprehensively considers data such as item attributes and historical information of item prediction information for prediction. Summary of the Invention

[0004] After analysis, the inventors found that most of the prediction methods in the related art are to solve the information prediction of items that have been on the market for a period of time. However, for new products to be launched in the future, there is currently no mature method for pre-determining the user interaction index.

[0005] One technical problem to be solved by the embodiments of the present invention is: how to more accurately pre-determine the user interaction index of new products.

[0006] According to the first aspect of some embodiments of the present invention, there is provided a data processing method, including: inputting the feature data of a new product into a pre-trained feature extraction network to obtain the output reference feature of the new product output by the last layer of the feature extraction network, where the feature extraction network is trained based on a feature extraction loss function, and the feature extraction loss function is constructed based on the maximum mean difference matching result of the output data of multiple layers of the feature extraction network and the discrimination result of a discrimination network for distinguishing the source of the output reference feature; inputting the output reference feature into a pre-trained machine learning model to obtain the user interaction index of the new product output by the machine learning model.

[0007] In some embodiments, in the feature extraction loss function, the maximum mean difference matching result of the output data of multiple layers of the generation network is determined according to the gap between the intermediate reference feature of the new product for training and the intermediate reference feature of the associated product of the new product for training, and the intermediate reference feature is the feature output by the intermediate layer of the feature extraction network after inputting the feature data into the feature extraction network.

[0008] In some embodiments, in the feature extraction loss function, the discrimination result of the discriminant network of the generating network is determined based on the correctness of the source discrimination of the discriminant network on the output reference features of the new product used for training and the output reference features of the related products used for training.

[0009] In some embodiments, the machine learning model is trained based on the gap between the prediction result of the machine learning model on the user interaction index of non-new products and the actual user interaction index of the non-new products.

[0010] In some embodiments, the feature data includes attribute features of the corresponding item and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item.

[0011] In some embodiments, the user interaction index includes at least one of click volume, page view volume, favorite volume, and order volume.

[0012] According to a second aspect of some embodiments of the present invention, there is provided a training method for data processing, comprising: inputting feature data of a new product for training into a feature extraction network, obtaining intermediate reference features of the new product for training output by multiple intermediate layers of the feature extraction network, and obtaining output reference features of the new product for training output by the last layer of the feature extraction network; inputting feature data of products associated with the new product into the feature extraction network, obtaining intermediate reference features of the associated products output by multiple intermediate layers of the feature extraction network, and obtaining output reference features of the associated products output by the last layer of the feature extraction network; inputting the output reference features of the new product and the output reference features of the associated products into a discriminant network, and obtaining a discriminant result of the discriminant network on the source of the input reference features; determining a loss value of the feature extraction network based on the gap between the intermediate reference features of the new product for training and the intermediate reference features of the associated products, and the discriminant result of the discriminant network; and adjusting the parameters of the feature extraction network based on the loss value.

[0013] In some embodiments, the feature data includes attribute features of the corresponding item and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item.

[0014] According to a third aspect of some embodiments of the present invention, there is provided a data processing device, comprising: a feature extraction module, configured to input feature data of a new product into a pre-trained feature extraction network, and obtain output reference features of the new product outputted by the last layer of the feature extraction network, wherein the feature extraction network is trained based on a feature extraction loss function, and the feature extraction loss function is constructed based on maximum mean difference matching results of output data of multiple layers of the feature extraction network and a discrimination result of a discriminant network for distinguishing the source of the output reference features; a user interaction index acquisition module, configured to input the output reference features into a pre-trained machine learning model, and obtain the user interaction index of the new product outputted by the machine learning model.

[0015] According to a fourth aspect of some embodiments of the present invention, there is provided a training device for data processing, comprising: a first reference feature extraction module, configured to input feature data of a new product for training into a feature extraction network, obtain intermediate reference features of the new product for training output by multiple intermediate layers of the feature extraction network, and obtain output reference features of the new product for training output by the last layer of the feature extraction network; a second reference feature extraction module, input feature data of products associated with the new product into the feature extraction network, obtain intermediate reference features of the associated products output by multiple intermediate layers of the feature extraction network, and obtain output reference features of the associated products output by the last layer of the feature extraction network; a discrimination module, configured to input output reference features of the new product and output reference features of the associated products into the discrimination network, and obtain discrimination results of the discrimination network on the sources of the input reference features; a loss value determination module, configured to determine the loss value of the feature extraction network according to the gap between the intermediate reference features of the new product for training and the intermediate reference features of the associated products, and the discrimination results of the discrimination network; an adjustment module, configured to adjust the parameters of the feature extraction network according to the loss value.

[0016] According to a fifth aspect of some embodiments of the present invention, there is provided a data processing system, comprising: the aforementioned data processing device; and the aforementioned training device for data processing.

[0017] According to a sixth aspect of some embodiments of the present invention, there is provided a data processing device, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the aforementioned data processing methods based on instructions stored in the memory.

[0018] According to a seventh aspect of some embodiments of the present invention, there is provided a data processing device, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the aforementioned training methods for data processing based on instructions stored in the memory.

[0019] According to an eighth aspect of some embodiments of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, any one of the aforementioned data processing methods or training methods for data processing is implemented.

[0020] Some embodiments of the above invention have the following advantages or beneficial effects. In the case of a lack of historical data for new products, the method of the embodiment of the present invention can determine the output reference features that are close to the distribution features of the associated products of the new product through the feature extraction network in advance, and further use the output reference features to predict the user interaction index. Therefore, by introducing multi-layer transfer learning in the training of the feature extraction network, the difference between the attribute distribution of new products and the attribute distribution of historical new products is reduced, so that the initial sales data of historical new products can help predict the user interaction index of the current new product, thereby improving the accuracy of information prediction for new products.

[0021] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0023] Figure 1 A schematic flow chart of a data processing method according to some embodiments of the present invention is shown.

[0024] Figure 2 A schematic flow chart of a feature extraction network training method according to some embodiments of the present invention is shown.

[0025] Figure 3 The schematic diagram of the feature extraction network training process is shown exemplarily.

[0026] Figure 4 A schematic structural diagram of a data processing device according to some embodiments of the present invention is shown.

[0027] Figure 5 A schematic diagram of the structure of a training device for data processing according to some embodiments of the present invention is shown.

[0028] Figure 6 A schematic structural diagram of a data processing system according to some embodiments of the present invention is shown.

[0029] Figure 7A schematic structural diagram of a data processing device according to some embodiments of the present invention is shown.

[0030] Figure 8 A schematic structural diagram of a data processing device according to some other embodiments of the present invention is shown. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present invention and its application or use. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0032] The relative arrangement of components and steps, the numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

[0033] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0034] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0035] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0036] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0037] Figure 1 FIG. 2 is a flow chart showing a data processing method according to some embodiments of the present invention. Figure 1 As shown, the data processing method of this embodiment includes steps S102 to S104.

[0038] In step S102, the feature data of the new product is input into a pre-trained feature extraction network to obtain the output reference features of the new product outputted by the last layer of the feature extraction network.

[0039] New products refer to items that have not been launched online, or items that have been launched online for a very short time, resulting in insufficient data collection. The feature data of new and non-new products includes, for example, attribute information of the corresponding items, such as category, brand, price, function, etc.

[0040] In some embodiments, the feature data includes attribute features of the corresponding items and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item. In some embodiments, the user interaction index includes clicks, views, favorites, orders, and the like. For example, if the clicks of a new product of a certain shoe brand are to be predicted, the environmental features of the new shoe product may include, for example, the average user interaction index of all shoe items of the same brand in the most recent preset period, the average user interaction index of all shoe items on the e-commerce platform in the most recent preset period, the average user interaction index of all clothing items on the e-commerce platform in the most recent preset period, and the like. Thus, when predicting, the feature extraction results may be influenced by environmental features, so that the extracted features are more in line with the current user interaction environment.

[0041] The feature extraction network is used to generate output reference features for predicting the user interaction index based on the feature data of the item. The output reference features are also multidimensional data. The feature extraction network is, for example, a multi-layer neural network.

[0042] The feature extraction network is constructed based on the feature extraction loss function, which is constructed based on the maximum mean difference matching results of the output data of multiple layers of the generation network and the discrimination results of the discriminant network used to distinguish the source of the output reference features.

[0043] The maximum mean difference matching result is used to measure the similarity of two different data sets in data distribution. By using it as a constraint in feature extraction network training, more reasonable new product features can be obtained, thereby improving the accuracy of prediction.

[0044] In some embodiments, the maximum mean difference matching result is used to measure the difference between the features extracted by the new product in the feature extraction network and the features extracted by the associated products of the new product in the feature extraction network. The associated products of the new product are items that have been online for a period of time, that is, non-new products. Thus, the output reference features of the new product output by the feature extraction network can be closer to the distribution of the associated products, making the prediction results more accurate.

[0045] The discriminant network is used to receive the features output by the feature extraction network and determine whether the features belong to new products or not. Thus, the feature extraction network and the discriminant network form a "generative-adversarial" model, which makes the features extracted by the feature extraction network more accurate on the one hand, and makes the discrimination of the adversarial network more accurate on the other hand. The two promote each other and improve the overall prediction accuracy.

[0046] In step S104, the output reference features are input into a pre-trained machine learning model to obtain a user interaction index of the new product output by the machine learning model.

[0047] In some embodiments, the machine learning model is trained based on the gap between the prediction result of the machine learning model on the user interaction index of non-new products and the actual user interaction index of the non-new products.

[0048] The machine learning model is, for example, a regression model. Assume that in the training phase, the input data of the machine learning model is x extracted by the feature extraction network. i , the extracted features are the features of the training data, and x i After inputting into the machine learning model, the prediction result r is obtained i The training data has a pre-labeled user interaction index y i , then in some embodiments, the square loss is constructed using formula (1) to train the regression model.

[0049] L=(y i -r i ) 2 (1)

[0050] After adjusting the parameters of the machine learning model multiple times based on the square loss, a fully trained machine learning model can be obtained.

[0051] In the case of a lack of historical data for new products, the method of the above embodiment can determine the output reference features close to the distribution features of the associated products of the new product through the feature extraction network in advance, and further use the output reference features to predict the user interaction index. Thus, by introducing multi-layer transfer learning in the training of the feature extraction network, the difference between the attribute distribution of new products and the attribute distribution of historical new products is reduced, so that the initial sales data of historical new products can help predict the user interaction index of the current new product, thereby improving the accuracy of information prediction of new products.

[0052] Reference below Figure 2 An embodiment of a feature extraction network training method is described.

[0053] Figure 2 FIG. 2 is a flow chart showing a feature extraction network training method according to some embodiments of the present invention. Figure 2 As shown, the training method of this embodiment includes steps S202 to S210.

[0054] In step S202, the feature data of the new product for training is input into the feature extraction network to obtain intermediate reference features of the new product for training output by multiple intermediate layers of the feature extraction network, and output reference features of the new product for training output by the last layer of the feature extraction network are obtained.

[0055] In step S204, the feature data of the new product's associated products are input into the feature extraction network to obtain intermediate reference features of the associated products output by multiple intermediate layers of the feature extraction network, and to obtain output reference features of the associated products output by the last layer of the feature extraction network.

[0056] In some embodiments, the associated products of the new product are items belonging to the same category as the new product. For example, when the new product is a refrigerator of brand A, the associated products are other products under the refrigerator category of the e-commerce platform.

[0057] In step S206, the output reference features of the new product and the output reference features of the related product are input into the discriminant network to obtain the discriminant network's discriminant result on the source of the input reference features.

[0058] In step S208, the loss value of the feature extraction network is determined according to the difference between the intermediate reference features of the new product used for training and the intermediate reference features of the related products, and the discrimination result of the discriminant network.

[0059] In step S210, the parameters of the feature extraction network are adjusted according to the loss value.

[0060] Figure 3 The schematic diagram of the feature extraction network training process is shown as an example. Figure 3 As shown in Figure 2, the feature data of new products and related products are input into the feature extraction network G respectively to obtain the intermediate reference features output by multiple intermediate layers. Figure 3 In the figure, the boxes in network G represent several layers of network G. Then, the output reference features output by the last layer of network G are input into the discriminant network D. Network D identifies which of the input reference features comes from the new product and which comes from the related product, and outputs the discrimination result. The value of the loss function of network G is determined by the maximum mean difference match between the intermediate reference features corresponding to the feature data from different sources and the discrimination result of network D. Therefore, when training, network G can approach the existing data in terms of both the output results and the intermediate results, thereby improving the accuracy of the prediction.

[0061] By predicting the user interaction index, the e-commerce system can be better optimized. In some embodiments, the predicted click volume, page view volume, favorite volume, order volume and other data are used as reference factors when pushing items to users, for example, as a feature of the item to participate in the calculation of the recommendation degree. In some embodiments, by predicting the click volume, page view volume, order volume and other data, the load of the website at a future moment is predicted, and the server or network configuration is adjusted accordingly. By predicting the order volume, the order system and logistics system are adjusted in advance to meet the balance between computing resources and response rate.

[0062] Reference below Figure 4 Embodiments of the data processing apparatus of the present invention are described.

[0063] Figure 4 FIG. 2 shows a schematic diagram of the structure of a data processing device according to some embodiments of the present invention. Figure 4 As shown, the data processing device 400 of this embodiment includes: a feature extraction module 4100, which is configured to input the feature data of a new product into a pre-trained feature extraction network, and obtain the output reference features of the new product output by the last layer of the feature extraction network, wherein the feature extraction network is trained based on a feature extraction loss function, and the feature extraction loss function is constructed based on the maximum mean difference matching results of the output data of multiple layers of the feature extraction network and the discrimination results of the discriminant network for distinguishing the source of the output reference features; a user interaction index acquisition module 4200, which is configured to input the output reference features into a pre-trained machine learning model, and obtain the user interaction index of the new product output by the machine learning model.

[0064] In some embodiments, in the feature extraction loss function, the maximum mean difference matching result of the output data of multiple layers of the generation network is determined based on the difference between the intermediate reference features of the new product used for training and the intermediate reference features of the related products of the new product used for training. The intermediate reference features are the features output by the intermediate layer of the feature extraction network after the feature data is input into the feature extraction network.

[0065] In some embodiments, in the feature extraction loss function, the discrimination result of the discriminant network of the generating network is determined based on the correctness of the source discrimination of the discriminant network on the output reference features of the new product used for training and the output reference features of the related products used for training.

[0066] In some embodiments, the machine learning model is trained based on the gap between the prediction result of the machine learning model on the user interaction index of non-new products and the actual user interaction index of the non-new products.

[0067] In some embodiments, the feature data includes attribute features of the corresponding item and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item.

[0068] In some embodiments, the user interaction index includes at least one of click volume, page view volume, favorite volume, and order volume.

[0069] Reference below Figure 5 An embodiment of a training device for data processing according to the present invention is described.

[0070] Figure 5 FIG. 2 shows a schematic diagram of a training device for data processing according to some embodiments of the present invention. Figure 5 As shown, the training device 500 for data processing of this embodiment includes: a first reference feature extraction module 5100, which is configured to input feature data of a new product for training into a feature extraction network, obtain intermediate reference features of the new product for training output by multiple intermediate layers of the feature extraction network, and obtain output reference features of the new product for training output by the last layer of the feature extraction network; a second reference feature extraction module 5200, which inputs feature data of products associated with the new product into the feature extraction network, obtains intermediate reference features of products associated with the new product output by multiple intermediate layers of the feature extraction network, And obtain the output reference features of the last layer output of the feature extraction network and the output reference features of the related products; the discrimination module 5300 is configured to input the output reference features of the new product and the output reference features of the related products into the discrimination network, and obtain the discrimination result of the discrimination network on the source of the input reference features; the loss value determination module 5400 is configured to determine the loss value of the feature extraction network according to the gap between the intermediate reference features of the new product used for training and the intermediate reference features of the related products, and the discrimination result of the discrimination network; the adjustment module 5500 is configured to adjust the parameters of the feature extraction network according to the loss value.

[0071] In some embodiments, the feature data includes attribute features of the corresponding item and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item.

[0072] Reference below Figure 6 An embodiment of a data processing system of the present invention is described.

[0073] Figure 6 FIG. 2 shows a schematic diagram of a data processing system according to some embodiments of the present invention. Figure 6 As shown, the data processing system 60 of this embodiment includes a data processing device 400; and the aforementioned training device 500 for data processing.

[0074] Figure 7FIG. 2 shows a schematic diagram of a data processing device or a training device for data processing according to some embodiments of the present invention. Figure 7 As shown, the data processing device 70 of this embodiment includes: a memory 710 and a processor 720 coupled to the memory 710 , and the processor 720 is configured to execute the data processing method in any one of the aforementioned embodiments based on instructions stored in the memory 710 .

[0075] The memory 710 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application program, a boot loader, and other programs.

[0076] Figure 8 FIG. 2 shows a schematic diagram of a data processing device or a training device for data processing according to some other embodiments of the present invention. Figure 8 As shown, the data processing device 80 of this embodiment includes: a memory 810 and a processor 820, and may also include an input / output interface 830, a network interface 840, a storage interface 850, etc. These interfaces 830, 840, 850 and the memory 810 and the processor 820 may be connected, for example, via a bus 860. Among them, the input / output interface 830 provides a connection interface for input / output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 840 provides a connection interface for various networked devices. The storage interface 850 provides a connection interface for external storage devices such as SD cards and USB flash drives.

[0077] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements any one of the aforementioned data processing methods when executed by a processor.

[0078] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0079] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0080] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0081] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0082] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A data processing method, include: Inputting feature data of a new product into a pre-trained feature extraction network to obtain output reference features of the new product outputted by the last layer of the feature extraction network, wherein the feature extraction network is trained based on a feature extraction loss function, the feature extraction loss function is constructed based on maximum mean difference matching results of output data of multiple layers of the feature extraction network, and a discrimination result of a discriminant network for distinguishing the source of the output reference features, and the maximum mean difference matching result is used to measure the difference between features extracted from the new product in the feature extraction network and features extracted from related products of the new product in the feature extraction network; The output reference feature is input into a pre-trained machine learning model to obtain a user interaction index of the new product output by the machine learning model.

2. The data processing method according to claim 1, in, In the feature extraction loss function, the maximum mean difference matching result of the output data of multiple layers of the feature extraction network is determined based on the gap between the intermediate reference features of the new product used for training and the intermediate reference features of the related products of the new product used for training. The intermediate reference features are the features output by the intermediate layer of the feature extraction network after the feature data is input into the feature extraction network.

3. The data processing method according to claim 1, in, In the feature extraction loss function, the discrimination result of the discriminant network of the feature extraction network is determined based on the correctness of the source discrimination of the discriminant network on the output reference features of the new product used for training and the output reference features of the related products of the new product used for training.

4. The data processing method according to claim 1, in, The machine learning model is trained based on the gap between the prediction result of the machine learning model on the user interaction index of non-new products and the actual user interaction index of the non-new products.

5. The data processing method according to any one of claims 1 to 4, in, The feature data includes attribute features of the corresponding item and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item.

6. The data processing method according to any one of claims 1 to 4, in, The user interaction index includes at least one of click volume, page view volume, favorite volume, and order volume.

7. A training method for data processing, include: Inputting feature data of a new product for training into a feature extraction network, obtaining intermediate reference features of the new product for training output by multiple intermediate layers of the feature extraction network, and obtaining output reference features of the new product for training output by the last layer of the feature extraction network; Inputting feature data of new product-related products into a feature extraction network, obtaining intermediate reference features of the related products output by multiple intermediate layers of the feature extraction network, and obtaining output reference features of the related products output by the last layer of the feature extraction network; Inputting the output reference features of the new product and the output reference features of the related product into a discriminant network, and obtaining a discriminant result of the discriminant network on the source of the input reference features; Determining the loss value of the feature extraction network according to the gap between the intermediate reference features of the new product used for training and the intermediate reference features of the related product, and the discrimination result of the discriminant network; The parameters of the feature extraction network are adjusted according to the loss value.

8. The training method according to claim 7, in, The feature data includes attribute features of the corresponding item and environmental features, and the environmental features include a user interaction index corresponding to a set of one or more associated items of the item.

9. A data processing device, include: A feature extraction module is configured to input feature data of a new product into a pre-trained feature extraction network to obtain an output reference feature of the new product outputted by the last layer of the feature extraction network, wherein the feature extraction network is trained based on a feature extraction loss function, the feature extraction loss function is constructed based on a maximum mean difference matching result of output data of multiple layers of the feature extraction network and a discrimination result of a discriminant network for distinguishing a source of the output reference feature, and the maximum mean difference matching result is used to measure the difference between features extracted from the feature extraction network for the new product and features extracted from the feature extraction network for associated products of the new product; The user interaction index acquisition module is configured to input the output reference features into a pre-trained machine learning model to obtain the user interaction index of the new product output by the machine learning model.

10. A training device for data processing, include: A first reference feature extraction module is configured to input feature data of a new product for training into a feature extraction network, obtain intermediate reference features of the new product for training output by multiple intermediate layers of the feature extraction network, and obtain output reference features of the new product for training output by the last layer of the feature extraction network; A second reference feature extraction module inputs feature data of the new product's associated products into a feature extraction network, obtains intermediate reference features of the associated products output by multiple intermediate layers of the feature extraction network, and obtains output reference features of the associated products output by the last layer of the feature extraction network; A discrimination module is configured to input the output reference features of the new product and the output reference features of the related product into a discrimination network, and obtain a discrimination result of the discrimination network on the source of the input reference features; A loss value determination module, configured to determine the loss value of the feature extraction network according to the gap between the intermediate reference features of the new product for training and the intermediate reference features of the related product, and the discrimination result of the discriminant network; An adjustment module is configured to adjust parameters of the feature extraction network according to the loss value.

11. A data processing system, include: The data processing device according to claim 9; as well as The training device for data processing as claimed in claim 10.

12. A data processing device, include: Memory; as well as A processor coupled to the memory, the processor being configured to execute the data processing method according to any one of claims 1 to 6 based on instructions stored in the memory.

13. A training device for data processing, include: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the training method for data processing according to claim 7 or 8 based on instructions stored in the memory.

14. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the data processing method described in any one of claims 1 to 6, or the training method for data processing described in claim 7 or 8.

Images