Electronic commerce recommendation method and device

Abstract

The present application discloses an electronic commerce recommendation method. The method comprises: acquiring user feature data, wherein the user feature data comprises first sample data and second sample data, and the first sample data and the second sample data are respectively positive and negative sample data for each other; inputting the first sample data to a generative adversarial network to obtain supplementary sample data; on the basis of the first sample data, the second sample data, and the supplementary sample data, generating a transaction conversion model, and constructing a transaction click model; performing model fusion on the transaction conversion model and the transaction click model to obtain a fused model; and recommending a product to a user on the basis of the fused model. Further disclosed is an electronic commerce recommendation device.

Description

E-commerce recommendation method and device

[0001] Related applications

[0002] This application claims priority to Chinese patent application number 2024103799664, filed on March 29, 2024, entitled “E-commerce recommendation method and device,” the entire text of which is hereby incorporated by reference. Technical Field

[0003] The present application relates to the field of computer technology, and more specifically, to an e-commerce recommendation method and device. Background Art

[0004] Recommendation systems typically use multi-objective modeling methods to balance multiple business objectives. Multi-model fusion methods train separate models based on business scenarios, combine the final goal, and fuse the model results by adding or multiplying them. This can enhance the diversity, accuracy, and novelty of the system. In actual recommendation scenarios, multi-model fusion methods are often used to predict CTR (Click-Through Rate) and CVR (Conversion Rate) indicators. In the process of building CTR and CVR fusion models, the CTR model uses data with clicks as positive samples, and the CVR model uses data with transactions as positive samples. In terms of data volume, the positive samples of the CVR model are very sparse compared to the positive samples of the CTR model. In addition, the different data spaces corresponding to the models will cause certain sample deviations, resulting in problems of data sparsity and sample selection bias in multi-model fusion.

[0005] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0006] The embodiments of the present application provide an e-commerce recommendation method and device.

[0007] According to a first aspect of an embodiment of the present application, an e-commerce recommendation method is provided, including: obtaining user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data for each other; inputting the first sample data into a generative adversarial network to obtain supplementary sample data; generating a transaction conversion model based on the first sample data, the second sample data and the supplementary sample data, and constructing a transaction click model; fusing the transaction conversion model and the transaction click model to obtain a fusion model; and recommending products to the user based on the fusion model.

[0008] Optionally, the first sample data is input into a generative adversarial network to obtain supplementary sample data, including: obtaining random data generated by a preset generator; updating the parameters of the discriminator according to the first sample data and the random data, and updating the parameters of the generator according to the first sample data, so that the generator updates the random data until the discriminator cannot distinguish between the first sample data and the updated random data, and uses the updated random data as generated transaction data; obtaining the supplementary sample data according to the generated transaction data and a pre-constructed initial transaction conversion model, wherein the initial transaction conversion model is constructed based on the first sample data.

[0009] Optionally, the supplementary sample data is obtained based on the generated transaction data and the pre-built initial transaction conversion model, including: inputting the generated transaction data into the initial transaction conversion model to obtain multiple prediction results; and using the generated transaction data corresponding to the prediction results greater than the preset probability among the multiple prediction results as the supplementary sample data.

[0010] Optionally, a transaction conversion model is generated based on the first sample data, the second sample data and the supplementary sample data, including: determining the first sample data and the supplementary sample data as a first positive sample data set; determining the second sample data as a first negative sample data set; using the first positive sample data set and the first negative sample data set as training data sets for the transaction conversion model, and training to obtain the transaction conversion model.

[0011] Optionally, constructing a transaction click model includes: obtaining third sample data from the user feature data, the third sample data being used to represent the user's interactive behavior; dividing the third sample data into a second positive sample data set and a second negative sample data set based on whether feedback is received after the recommendation information is issued; and training the transaction click model based on the second positive sample data set and the second negative sample data set.

[0012] Optionally, the transaction conversion model and the transaction click model are fused to obtain a fusion model, including: multiplying the transaction conversion model and the transaction click model to obtain a first model; and adjusting the first model according to preset weights to obtain the fusion model.

[0013] Optionally, recommending products to users based on the fusion model includes: obtaining a probability based on the fusion model, wherein the probability is used to characterize the probability that the user is interested in the recommended product; sorting multiple users according to the probabilities, selecting users with probabilities higher than a preset probability as target users, and recommending the recommended products to the target users.

[0014] According to the second aspect of the embodiment of the present application, an e-commerce recommendation device is also provided, including: an acquisition module for acquiring user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data for each other; a supplementation module for inputting the first sample data into a generative adversarial network to obtain supplementation sample data; a construction module for generating a transaction conversion model based on the first sample data, the second sample data and the supplementation sample data, and constructing a transaction click model; a fusion module for fusing the transaction conversion model with the transaction click model to obtain a fusion model; and a recommendation module for recommending products to users based on the fusion model.

[0015] According to the third aspect of the embodiment of the present application, a computer device is also provided, including: a memory for storing program instructions; a processor, connected to the memory, for executing the program instructions of the following functions: obtaining user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data for each other; inputting the first sample data into a generative adversarial network to obtain supplementary sample data; generating a transaction conversion model based on the first sample data, the second sample data and the supplementary sample data, and constructing a transaction click model; fusing the transaction conversion model and the transaction click model to obtain a fusion model; and recommending products to users based on the fusion model.

[0016] According to a fourth aspect of an embodiment of the present application, a computer program product is further provided, including a computer program, which implements the e-commerce recommendation method when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the conventional technology, the following briefly introduces the drawings required for use in the embodiments or the conventional technology descriptions. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the disclosed drawings without any creative work.

[0018] FIG1 is a hardware structure block diagram of a computer terminal for implementing an e-commerce recommendation method according to some embodiments of the present application;

[0019] FIG2 is a flowchart of an e-commerce recommendation method provided according to some embodiments of the present application;

[0020] FIG3 is a diagram illustrating a training principle of a generative adversarial network according to some embodiments of the present application;

[0021] FIG4 is a schematic diagram of an e-commerce recommendation method according to some embodiments of the present application;

[0022] FIG5 is a flowchart of another e-commerce recommendation method provided according to some embodiments of the present application;

[0023] FIG6 is a structural diagram of an e-commerce recommendation device provided according to some embodiments of the present application. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0025] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] The embodiment of the method for processing a fault work order provided in the embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 shows a hardware structure block diagram of a computer terminal for implementing an e-commerce recommendation method. As shown in Figure 1, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used in the figure to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be used as one of the ports of the I / O interface), a network interface, and a BUS bus. It will be understood by those skilled in the art that the structure shown in Figure 1 is only for illustration and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may also include more or fewer components than those shown in Figure 1, or have a configuration different from that shown in Figure 1.

[0027] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry." The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10. As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0028] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method for handling fault work orders in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the above-mentioned e-commerce recommendation method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0029] The transmission module 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission module 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission module 106 may be a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0030] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0031] It should be noted that, in some alternative embodiments, the computer device shown in FIG. 1 may include hardware components (including circuits), software components (including computer code stored on a computer-readable medium), or a combination of hardware and software components. It should be noted that FIG. 1 is merely an example of a specific embodiment and is intended to illustrate the types of components that may be present in the computer device.

[0032] In the above-mentioned operating environment, an embodiment of the present application provides an embodiment of an e-commerce recommendation method. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0033] FIG2 is a flow chart of an e-commerce recommendation method according to an embodiment of the present application. As shown in FIG2 , the method includes the following steps:

[0034] Step S202: Acquire user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data of each other;

[0035] In the above step S202, the user feature data is the user's actual interactive behavior feature data, wherein click data is used as the training data of the transaction conversion model. The sample data with clicks and transaction data on the same day is positive sample data, otherwise it is negative sample data. The category features of the positive sample data are one-hot encoded, and the category features and numerical features are spliced ​​to obtain the representation vector U1 of the positive sample data. The negative sample data is one-hot encoded to obtain the representation vector U2 of the negative sample data.

[0036] Otherwise it is negative sample data.

[0037] Step S204: input the first sample data into a generative adversarial network to obtain supplementary sample data;

[0038] In the above step S204, since the first sample data is positive sample data and the positive sample data is relatively small, the first sample data is expanded using a generative adversarial network to obtain supplementary sample data.

[0039] Step S206: Generate a transaction conversion model based on the first sample data, the second sample data, and the supplementary sample data, and construct a transaction click model.

[0040] In the above step S206, the transaction conversion model is used to predict the probability of a user making a purchase after clicking on the advertisement of the product to be recommended, and the transaction click model is used to predict the probability of a user clicking on the advertisement after seeing the advertisement of the product to be recommended.

[0041] Step S208: Fusing the transaction conversion model and the transaction click model to obtain a fusion model.

[0042] In the above step S208, by fusing the transaction conversion model and the transaction click model, the accuracy of the recommendation results can be enhanced.

[0043] Step S210: recommend products to the user based on the fusion model.

[0044] Through the e-commerce recommendation method in steps S202 to S210 above, user feature data is obtained, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data for each other; the first sample data is input into a generative adversarial network to obtain supplementary sample data; a transaction conversion model is generated based on the first sample data, the second sample data, and the supplementary sample data, and a transaction click model is constructed; the transaction conversion model and the transaction click model are fused to obtain a fusion model; products are recommended to users based on the fusion model, and the first sample data of the transaction conversion model is expanded using a generative adversarial network to increase sample data; then, a transaction conversion model is generated using the user feature data and the supplementary sample data, and the transaction conversion model and the transaction click model are fused to increase the sample data of the fusion model, thereby solving the problem of sparse data in the fusion of the CTR model and the CVR model, while alleviating sample selection bias, thereby improving the recommendation effect. Detailed description is given below.

[0045] It should be noted that the transaction click model is used to predict the probability of receiving feedback (clicking on the product) after sending the recommendation information, and the transaction conversion model is used to predict the probability of completing the transaction after clicking on the product.

[0046] It can be understood that the transaction click model is a CTR model, and the transaction conversion model is a CVR model.

[0047] In some embodiments of the present application, first sample data is input into a generative adversarial network to obtain supplementary sample data, including: obtaining random data generated by a preset generator; updating the parameters of the discriminator according to the first sample data and the random data, and updating the parameters of the generator according to the first sample data, so that the generator updates the random data until the discriminator cannot distinguish between the first sample data and the updated random data, and using the updated random data as generated transaction data; obtaining supplementary sample data based on the generated transaction data and a pre-constructed initial transaction conversion model, wherein the initial transaction conversion model is constructed based on the first sample data.

[0048] Specifically, as shown in FIG3 , the generative adversarial network includes a generator G and a discriminator D. The generator G can generate random data, which can be a random vector emb fake =(w1,…,w d ), where w d ~N(0,1). The first sample data is the input data of the generative adversarial network, and its representation vector is emb real To express.

[0049] Then the parameter gradient of the discriminator D is updated to judge the consistency between the random data generated by the generator G and the first sample data. The loss function is as follows:

[0050] Where σ(·) is the sigmoid function and F(·) is the function that maps a vector to a scalar.

[0051] Then the parameter gradient of the generator G is updated so that the generated random data is close to the first sample data. Its loss function is as follows:

[0052] Where m represents the number of random vectors, and M represents the number of random vectors.

[0053] Then repeat the above two steps until the following conditions are met:

[0054] Where q is the preset value.

[0055] At this time, the discriminator D cannot distinguish between the random data and the first sample data, and the random data generated by the generator G can be considered as real sample data, and then the random data is used as generated transaction data.

[0056] The categorical features of the generated transaction data are then uniquely encoded, and the categorical features and numerical features are concatenated to generate the generated transaction data representation vector W1. The transaction data representation vector W1 is input into the pre-built initial transaction conversion model to obtain supplementary sample data.

[0057] It should be noted that the initial transaction conversion model was constructed using the GBDT (Gradient Boosting Decision Tree) method.

[0058] In some embodiments, supplementary sample data is obtained based on the generated transaction data and a pre-built initial transaction conversion model, including: inputting the generated transaction data into the initial transaction conversion model to obtain multiple prediction results; and using the generated transaction data corresponding to the prediction results greater than a preset probability among the multiple prediction results as supplementary sample data.

[0059] The transaction data representation vector W1 is input into the pre-built initial transaction conversion model, and the generated transaction data whose prediction results are greater than the preset probability are retained, and the generated transaction data whose prediction results are greater than the preset probability are used as supplementary sample data W2, where the preset probability is between 0-1, for example 0.8.

[0060] In some embodiments of the present application, as shown in Figure 4, a transaction conversion model is generated based on the first sample data, the second sample data and the supplementary sample data, including: determining the first sample data and the supplementary sample data as a first positive sample data set; determining the second sample data as a first negative sample data set; using the first positive sample data set and the first negative sample data set as training data sets for the transaction conversion model, and training to obtain the transaction conversion model.

[0061] Specifically, the first sample data is the positive sample data U1, and the supplementary sample data is generated based on the first sample data, so the supplementary sample data W2 is also the positive sample data. The first positive sample data set add of the transaction conversion model of the first sample data U1 and the supplementary sample data W2 u1 , where add u1 =concat(U1,W2). The second sample data is the negative sample data U2, which constitutes the first negative sample data set U2. Then, based on the DeepFM method, the first positive sample data set is added u1 The CVR model is trained using the first negative sample dataset U2, where DeepFM is obtained by combining the deep learning model and the factorization machine model.

[0062] In some embodiments of the present application, as shown in FIG4 , constructing a transaction click model includes: obtaining third sample data from user feature data, the third sample data being used to represent the user's interactive behavior; dividing the third sample data into a second positive sample data set and a second negative sample data set based on whether feedback is received after the recommendation information is issued; and training the transaction click model based on the second positive sample data set and the second negative sample data set.

[0063] Specifically, user feature data also includes exposure-click behavior, which is used as the third sample data. When an advertisement for a recommended product (i.e., recommendation information) is sent, if a user clicks on it, this data becomes the second positive sample dataset, with a representation vector V1. Otherwise, this data becomes the second negative sample dataset, with a representation vector V2. The second positive and second negative sample datasets are then used to train a CTR model, which is also derived using the DeepFM method.

[0064] In some embodiments, as shown in FIG4 , the transaction conversion model and the transaction click model are fused to obtain a fusion model, including: multiplying the transaction conversion model and the transaction click model to obtain a first model; and adjusting the first model according to preset weights to obtain a fusion model.

[0065] In an optional embodiment, the fusion formula of the transaction conversion model and the transaction click model is as follows: y = a(y1*y2)

[0066] Among them, y is the fusion model, y1 is the transaction conversion model, y2 is the transaction click model, and a is the preset weight.

[0067] It should be noted that the preset weight can be set according to actual conditions, for example, 0.3.

[0068] In some embodiments of the present application, product recommendations are made to users based on a fusion model, including: obtaining a probability based on the fusion model, wherein the probability is used to characterize the probability that the user is interested in the recommended product; sorting multiple users according to the probability, selecting users with a probability higher than a preset probability as target users, and recommending the recommended product to the target user.

[0069] Specifically, the recommended product is input into the fusion model to obtain the probability that the user is interested in the recommended product, multiple users are sorted according to the probability, and users with a probability higher than the preset probability are selected as target users, and the recommended product is recommended to the target user.

[0070] FIG5 shows a flow chart of another e-commerce recommendation method, which includes:

[0071] Step S301: Acquire user characteristic data, wherein the user characteristic data includes first sample data and second sample data.

[0072] S302: Input the first sample data into a generative adversarial network to obtain generated transaction data.

[0073] S303: Filter the generated transaction data using the initial transaction conversion model to obtain supplementary sample data, wherein the initial transaction conversion model is constructed based on the first sample data.

[0074] S304, determining the first sample data and the supplementary sample data as the first positive sample data set of the CVR model, and determining the second sample data as the first negative sample data set of the CVR model, and determining the second positive sample data set and the second negative sample data set of the CTR model according to the user feature data.

[0075] S305 : Training a CVR model using the first positive sample data set and the first negative sample data set, and training a CTR model using the second positive sample data set and the second negative sample data set.

[0076] S306: Fusing the CVR model and the CTR model to obtain a fusion model.

[0077] S307, based on the fusion model, obtain the probability that the user is interested in the recommended product, sort multiple users according to the probability, select the user with a probability higher than the preset probability as the target user, and recommend the recommended product to the target user.

[0078] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0079] Based on the same inventive concept, the present application also provides an e-commerce recommendation device for implementing the aforementioned e-commerce recommendation method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more e-commerce recommendation device embodiments provided below can be found in the above-mentioned limitations of the e-commerce recommendation method and will not be repeated here.

[0080] FIG6 is a structural diagram of an e-commerce recommendation device according to an embodiment of the present application. As shown in FIG6 , the device includes:

[0081] An acquisition module 40 is configured to acquire user characteristic data, wherein the user characteristic data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data of each other;

[0082] A supplementing module 42 is configured to input the first sample data into a generative adversarial network to obtain supplemented sample data;

[0083] A construction module 44 is configured to generate a transaction conversion model based on the first sample data, the second sample data, and the supplementary sample data, and to construct a transaction click model;

[0084] A fusion module 46 is used to fuse the transaction conversion model and the transaction click model to obtain a fusion model;

[0085] The recommendation module 48 is used to recommend products to users based on the fusion model.

[0086] The supplementing module 42 includes: a first acquisition submodule, an update submodule and a supplementing submodule, wherein the first acquisition submodule is used to obtain random data generated by a preset generator; the update submodule is used to update the parameters of the discriminator according to the first sample data and the random data, and update the parameters of the generator according to the first sample data, so that the generator updates the random data until the discriminator cannot distinguish between the first sample data and the updated random data, and uses the updated random data as the generated transaction data; the supplementing submodule is used to obtain supplementary sample data based on the generated transaction data and the pre-constructed initial transaction conversion model, wherein the initial transaction conversion model is constructed based on the first sample data.

[0087] The supplementary submodule includes: an input unit and a determination unit, wherein the input unit is used to input the generated transaction data into the initial transaction conversion model to obtain multiple prediction results; the determination unit is used to use the generated transaction data corresponding to the prediction results greater than the preset probability among the multiple prediction results as supplementary sample data.

[0088] The construction module 44 includes: a first determination submodule, a second determination submodule and a first training submodule, wherein the first determination submodule is used to determine the first sample data and the supplementary sample data as a first positive sample data set; the second determination submodule is used to determine the second sample data as a first negative sample data set; the first training submodule is used to use the first positive sample data set and the first negative sample data set as training data sets for the transaction conversion model to train and obtain the transaction conversion model.

[0089] The construction module 44 includes: a second acquisition submodule, a third determination submodule and a second training submodule, wherein the second acquisition submodule is used to obtain third sample data from user feature data, and the third sample data is used to represent the user's interactive behavior; the third determination submodule is used to divide the third sample data into a second positive sample data set and a second negative sample data set according to whether feedback is received after the recommendation information is issued; the second training submodule is used to train a transaction click model based on the second positive sample data set and the second negative sample data set.

[0090] The fusion module 46 includes: an operation submodule and an adjustment submodule, wherein the operation submodule is used to multiply the transaction conversion model and the transaction click model to obtain a first model; the adjustment submodule is used to adjust the first model according to preset weights to obtain a fusion model.

[0091] The recommendation module 48 includes: a fourth determination submodule and a recommendation submodule, wherein the fourth determination submodule is used to obtain a probability based on a fusion model, wherein the probability is used to characterize the probability that a user is interested in a recommended product; the recommendation submodule is used to sort multiple users according to the probability, select users with a probability higher than a preset probability as target users, and recommend the recommended products to the target users.

[0092] It should be noted that the e-commerce recommendation device shown in Figure 6 is used to execute the e-commerce recommendation method shown in Figure 2. Therefore, the relevant explanations in the above e-commerce recommendation method are also applicable to this type of e-commerce recommendation device and will not be repeated here.

[0093] An embodiment of the present application also provides a computer device, including: a memory for storing program instructions; a processor, connected to the memory, for executing the program instructions of the following functions: obtaining user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data for each other; inputting the first sample data into a generative adversarial network to obtain supplementary sample data; generating a transaction conversion model based on the first sample data, the second sample data and the supplementary sample data, and constructing a transaction click model; fusing the transaction conversion model and the transaction click model to obtain a fusion model; and recommending products to the user based on the fusion model.

[0094] It should be noted that the above-mentioned computer device is used to execute the e-commerce recommendation method shown in Figure 2, so the relevant explanations in the above-mentioned e-commerce recommendation method are also applicable to the computer device and will not be repeated here.

[0095] An embodiment of the present application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the following e-commerce recommendation method by running the computer program: obtaining user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data of each other; inputting the first sample data into a generative adversarial network to obtain supplementary sample data; generating a transaction conversion model based on the first sample data, the second sample data and the supplementary sample data, and constructing a transaction click model; fusing the transaction conversion model and the transaction click model to obtain a fusion model; and recommending products to the user based on the fusion model.

[0096] It should be noted that the above-mentioned non-volatile storage medium is used to execute the e-commerce recommendation method shown in Figure 2, so the relevant explanations in the above-mentioned e-commerce recommendation method are also applicable to the non-volatile storage medium and will not be repeated here.

[0097] The present application also provides a computer program product, including a computer program, which implements the steps of the e-commerce recommendation method when executed by a processor.

[0098] It should be noted that the above-mentioned computer program product is used to execute the e-commerce recommendation method shown in Figure 2, so the relevant explanations in the above-mentioned e-commerce recommendation method are also applicable to the computer program product and will not be repeated here.

[0099] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0100] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0101] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0102] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected to achieve the purpose of the present embodiment according to actual needs.

[0103] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0104] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program code.

[0105] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. An e-commerce recommendation method, comprising: Acquire user characteristic data, wherein the user characteristic data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data of each other; Inputting the first sample data into a generative adversarial network to obtain supplementary sample data; generating a transaction conversion model based on the first sample data, the second sample data, and the supplementary sample data, and constructing a transaction click model; Fusing the transaction conversion model and the transaction click model to obtain a fusion model; Recommend products to users based on the fusion model.

2. The method according to claim 1, wherein Inputting the first sample data into a generative adversarial network to obtain supplementary sample data includes: Get random data generated by the preset built generator; updating parameters of a discriminator according to the first sample data and the random data, and updating parameters of the generator according to the first sample data, so that the generator updates the random data until the discriminator cannot distinguish between the first sample data and the updated random data, and using the updated random data as generated transaction data; The supplementary sample data is obtained based on the generated transaction data and a pre-constructed initial transaction conversion model, wherein the initial transaction conversion model is constructed based on the first sample data.

3. The method according to claim 2, wherein: The supplementary sample data is obtained based on the generated transaction data and the pre-built initial transaction conversion model, including: Inputting the generated transaction data into the initial transaction conversion model to obtain a plurality of prediction results; The generated transaction data corresponding to the prediction results with a probability greater than the preset probability among the multiple prediction results is used as the supplementary sample data.

4. The method according to claim 1, wherein Generating a transaction conversion model according to the first sample data, the second sample data, and the supplementary sample data includes: Determine the first sample data and the supplementary sample data as a first positive sample data set; Determining the second sample data as a first negative sample data set; The first positive sample data set and the first negative sample data set are used as training data sets for the transaction conversion model to obtain the transaction conversion model through training.

5. The method according to claim 1, wherein Build a transaction click model, including: Acquire third sample data from the user characteristic data, where the third sample data is used to represent the user's interactive behavior; dividing the third sample data into a second positive sample data set and a second negative sample data set according to whether feedback is received after the recommendation information is issued; The transaction click model is obtained by training based on the second positive sample data set and the second negative sample data set.

6. The method according to claim 1, wherein The transaction conversion model and the transaction click model are fused to obtain a fusion model, including: Multiplying the transaction conversion model and the transaction click model to obtain a first model; The first model is adjusted according to preset weights to obtain the fusion model.

7. The method according to claim 6, wherein: Recommending products to users based on the fusion model includes: Obtaining a probability based on the fusion model, wherein the probability is used to represent the probability that the user is interested in the recommended product; Sort multiple users according to the probabilities, select users with probabilities higher than a preset probability as target users, and recommend the recommended products to the target users.

8. An e-commerce recommendation device, comprising: An acquisition module, configured to acquire user characteristic data, wherein the user characteristic data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data of each other; A supplementing module, configured to input the first sample data into a generative adversarial network to obtain supplementary sample data; A construction module, configured to generate a transaction conversion model based on the first sample data, the second sample data, and the supplementary sample data, and to construct a transaction click model; A fusion module, configured to fuse the transaction conversion model and the transaction click model to obtain a fusion model; The recommendation module is used to recommend products to users based on the fusion model.

9. A computer device comprising: a memory for storing program instructions; A processor, connected to the memory, is used to execute program instructions for the following functions: obtaining user feature data, wherein the user feature data includes first sample data and second sample data, and the first sample data and the second sample data are positive and negative sample data for each other; inputting the first sample data into a generative adversarial network to obtain supplementary sample data; generating a transaction conversion model based on the first sample data, the second sample data and the supplementary sample data, and constructing a transaction click model; fusing the transaction conversion model with the transaction click model to obtain a fusion model; and recommending products to users based on the fusion model.

10. A computer program product comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 7.

Images