Model training method and device, article recommendation method and device, equipment and medium

By obtaining and processing value conversion data, determining the waiting time window, generating a delay data set, and training based on cross-domain learning model, the sample delay and delay feedback problems in value conversion information generation model training are solved, and high-quality value conversion information is achieved Generate and accurate item recommendations.

CN118940025APending Publication Date: 2024-11-12JINGDONG TECH HLDG CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410868060.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-28
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing value conversion information generation model does not consider the sample conversion delay problem during the training process, which leads to the training sample set being inaccurate enough and the delay feedback problem is not considered, resulting in the generated value conversion information being inaccurate enough.

Method used

By obtaining user feature information, value-converted item feature information, cross feature information and domain feature information, feature engineering processing is carried out, the waiting time window for scene information is determined, the delay data set is generated, and training is carried out based on the initial model of cross-domain learning, and generating Value conversion information generation model.

Benefits of technology

Accurately generate value conversion information that characterizes the user's value conversion operation on the target item under different scenario information, which improves the accuracy and accuracy of the information, reduces system resource overhead and the cost of model development and deployment, and brings business Value enhancement.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940025A_ABST
    Figure CN118940025A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a model training method and device, an article recommendation method and device, equipment and a medium. A specific embodiment of the method comprises the steps of obtaining a value conversion data set; performing feature engineering processing on each piece of value conversion data in the value conversion data set to generate a processed data set; for each piece of scene information, determining a waiting time window according to value conversion proportion information corresponding to the target value conversion data subset; according to each waiting time window, delay data corresponding to each piece of processed data in the processed data set is generated, and a delay data set is obtained; and according to the delay data set, performing model training on the initial value conversion information generation model based on cross-domain learning to generate a value conversion information generation model. The implementation mode is related to artificial intelligence, the value conversion information generation model is utilized, and the value conversion information representing that the user executes the value conversion operation on the target article can be accurately generated under different scene information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a model training method, an item recommendation method, an apparatus, a device, and a medium. Background Art

[0002] At present, with the rapid development of network multimedia, more and more scenarios are improving conversion rates by distributing value coupons to enhance user activity and obtain higher value benefits. For the generation of value conversion information for value coupons, the method usually adopted is: first, obtain the user feature information set and the item feature information corresponding to the value coupon. Then, input the user feature information set and the item feature information corresponding to the value coupon into the value conversion information generation model (for example, the convolution classification model) to generate value conversion information.

[0003] However, the inventors have found that when the above method is used to generate value conversion information, the following technical problems often occur:

[0004] The training of the value conversion information generation model did not take into account the problem of sample conversion delay, resulting in inaccurate training sample sets. In addition, when building the model, the delayed feedback problem was not taken into account, resulting in inaccurate value conversion information generated by the generated value conversion information generation model.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0006] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0007] Some embodiments of the present disclosure propose model training methods, item recommendation methods, devices, equipment and media to solve the technical problems mentioned in the above background technology section.

[0008] In a first aspect, some embodiments of the present disclosure provide a model training method, including: obtaining a value conversion data set, wherein each value conversion data includes: user feature information, value conversion item feature information, cross-feature information and domain feature information under corresponding scenario information; performing feature engineering processing on each value conversion data in the above value conversion data set to generate a processed data set; for each scene information corresponding to each scenario information, according to the value conversion ratio information corresponding to a target value conversion data subset, determining a waiting time window corresponding to the above scene information, wherein the above target value conversion data subset is a data set in the above value conversion data set that corresponds to the above scenario information; according to the obtained waiting time windows, generating delay data corresponding to each processed data in the above processed data set to obtain a delay data set; according to the above delay data set, performing model training on an initial value conversion information generation model based on cross-domain learning to generate a value conversion information generation model.

[0009] Optionally, the above-mentioned determination of the waiting time window corresponding to the above-mentioned scenario information based on the value conversion ratio information corresponding to the target value conversion data subset includes: determining the end conversion time point in response to determining that the value conversion ratio information corresponding to the above-mentioned target value conversion data subset reaches the target value, and / or each value conversion voucher corresponding to the above-mentioned target value conversion data subset reaches the conversion date; obtaining a predetermined start conversion time point; and generating a waiting time window corresponding to the above-mentioned end conversion time point and the above-mentioned start conversion time point.

[0010] Optionally, the processed data also includes: a value conversion label; and the delay data corresponding to each processed data in the processed data set is generated according to the obtained waiting time windows to obtain the delay data set, including: for each of the above-mentioned scene information, the following generation steps are performed: determining the waiting time window corresponding to the above-mentioned scene information as the target waiting time window; determining the processed data subset in the processed data set corresponding to the above-mentioned scene information as the target processed data subset; according to the above-mentioned target waiting time window, updating the value conversion label corresponding to each processed data in the above-mentioned target processed data subset to generate updated data as delay data.

[0011] Optionally, the above-mentioned initial value conversion information generation model based on cross-domain learning is trained according to the above-mentioned delayed data set to generate a value conversion information generation model, including: according to a predetermined historical time period sequence, the above-mentioned delayed data set is divided into data to generate a delayed data subset sequence, wherein each delayed data subset has a corresponding historical time period; each delayed data subset in the above-mentioned delayed data subset sequence is respectively determined as a training data set, a verification data set and a test data set; according to the above-mentioned training data set, the above-mentioned verification data set and the above-mentioned test data set, the above-mentioned initial value conversion information generation model is trained to generate the above-mentioned value conversion information generation model.

[0012] Optionally, the above-mentioned feature engineering processing is performed on each value conversion data in the above-mentioned value conversion data set to generate a processed data set, including: feature screening of each data in the above-mentioned value conversion data set to generate a screened data set; data preprocessing of each data in the above-mentioned screened data set to generate the above-mentioned processed data set.

[0013] In a second aspect, some embodiments of the present disclosure provide a model training device, comprising: an acquisition unit, configured to acquire a value conversion data set, wherein each value conversion data includes: user feature information, value conversion item feature information, cross feature information and domain feature information under corresponding scenario information; a processing unit, configured to perform feature engineering processing on each value conversion data in the above value conversion data set to generate a processed data set; a determination unit, configured to determine, for each scenario information in the corresponding scenario information, a waiting time window corresponding to the above scenario information according to value conversion ratio information corresponding to a target value conversion data subset, wherein the above target value conversion data subset is a dataset in the above value conversion data set corresponding to the above scenario information; a generation unit, configured to generate delay data corresponding to each processed data in the above processed data set according to the obtained waiting time windows, to obtain a delay data set; a training unit, configured to perform model training on an initial value conversion information generation model based on cross-domain learning according to the above delay data set to generate a value conversion information generation model.

[0014] Optionally, the determination unit can be configured to: determine the end conversion time point in response to determining that the value conversion ratio information corresponding to the above-mentioned target value conversion data subset has reached the target value, and / or that each value conversion voucher corresponding to the above-mentioned target value conversion data subset has reached the conversion date; obtain a predetermined start conversion time point; and generate a waiting time window corresponding to the above-mentioned end conversion time point and the above-mentioned start conversion time point.

[0015] Optionally, the processed data also includes: a value conversion label; and the generation unit can be configured to: for each of the above-mentioned scenario information, perform the following generation steps: determine the waiting time window corresponding to the above-mentioned scenario information as the target waiting time window; determine the processed data subset in the above-mentioned processed data set corresponding to the above-mentioned scenario information as the target processed data subset; according to the above-mentioned target waiting time window, update the value conversion label corresponding to each processed data in the above-mentioned target processed data subset to generate updated data as delayed data.

[0016] Optionally, the training unit can be configured to: divide the above-mentioned delay data set according to a predetermined historical time period sequence to generate a delay data subset sequence, wherein each delay data subset has a corresponding historical time period; determine each delay data subset in the above-mentioned delay data subset sequence as a training data set, a verification data set and a test data set respectively; and perform model training on the above-mentioned initial value conversion information generation model according to the above-mentioned training data set, the above-mentioned verification data set and the above-mentioned test data set to generate the above-mentioned value conversion information generation model.

[0017] Optionally, the processing unit may be configured to: perform feature screening on each data in the above-mentioned value conversion data set to generate a screened data set; perform data preprocessing on each data in the above-mentioned screened data set to generate the above-mentioned processed data set.

[0018] On the third aspect, some embodiments of the present disclosure provide an item recommendation method, including: for each scene information in each scene information, performing the following recommendation steps: determining a value conversion item information set corresponding to the above scene information; using a pre-trained value conversion information generation model to generate item conversion information corresponding to each value conversion item information in the above value conversion item information set, to obtain an item conversion information set, wherein the above value conversion information generation model is generated based on a model training method; according to the above item conversion information set, setting recommendation information corresponding to each value conversion item information in the above value conversion item information set, to obtain a recommendation information set; according to the above recommendation information set, performing item recommendation for at least one value conversion item information, wherein the above at least one value conversion item information is a subset of item information in the target value conversion item information set.

[0019] Optionally, the above-mentioned use of a pre-trained value conversion information generation model to generate item conversion information corresponding to each value conversion item information in the above-mentioned value conversion item information set includes: using the above-mentioned value conversion information generation model to determine the value conversion information corresponding to each user in the target user set to obtain the value conversion information set, wherein the above-mentioned target user set is a user set corresponding to the above-mentioned value conversion item information; generating the above-mentioned item conversion information based on the above-mentioned value conversion information set.

[0020] In a fourth aspect, some embodiments of the present disclosure provide an item recommendation device, comprising: an execution unit, configured to perform the following recommendation steps for each scene information in each scene information: determine a value conversion item information set corresponding to the above scene information; use a pre-trained value conversion information generation model to generate item conversion information corresponding to each value conversion item information in the above value conversion item information set, and obtain an item conversion information set, wherein the above value conversion information generation model is generated based on the method mentioned above in one of claims 1-5; according to the above item conversion information set, set recommendation information corresponding to each value conversion item information in the above value conversion item information set, and obtain a recommendation information set; according to the above recommendation information set, perform item recommendation for at least one value conversion item information, wherein the above at least one value conversion item information is a subset of item information in the target value conversion item information set.

[0021] Optionally, the execution unit can be configured to: utilize the above-mentioned value conversion information generation model to determine the value conversion information corresponding to each user in the target user set to obtain a value conversion information set, wherein the above-mentioned target user set is a user set corresponding to the above-mentioned value conversion item information; and generate the above-mentioned item conversion information based on the above-mentioned value conversion information set.

[0022] In a fifth aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the implementation methods in the first and third aspects.

[0023] In a sixth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any one of the implementation modes of the first aspect and the third aspect is implemented.

[0024] In a seventh aspect, some embodiments of the present disclosure provide a computer program product, including a computer program, which, when executed by a processor, implements the method described in any one of the implementation modes of the first and third aspects above.

[0025] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, value conversion information characterizing the value conversion operation performed by the user on the target item can be accurately generated under different scene information. Specifically, the reason for the inaccuracy of the relevant value conversion information is that the training of the value conversion information generation model does not take into account the problem of sample conversion delay, resulting in the inaccuracy of the training sample set. In addition, when modeling the model, the delayed feedback problem is not taken into account, resulting in the value conversion information generated by the generated value conversion information generation model being inaccurate. Based on this, the model training method of some embodiments of the present disclosure first obtains a value conversion data set, wherein each value conversion data includes: user feature information and value conversion item feature information, cross feature information and domain feature information under corresponding scene information. Here, the acquired value conversion data set is used for feature learning of subsequent models. Cross feature information allows subsequent models to learn more interactive features between users and value conversion items, so that the subsequent model outputs more accurate value conversion information. In addition, through the domain feature information corresponding to each different scene through the value conversion data set, the subsequent model learns more feature information. Then, feature engineering is performed on each value conversion data in the above value conversion data set to generate a processed data set to further improve the data quality and fully reflect the feature information of the key features. Next, for each scene information in the corresponding scene information, according to the value conversion ratio information corresponding to the target value conversion data subset, the waiting time window corresponding to the above scene information can be accurately determined, wherein the above target value conversion data subset is a data set in the above value conversion data set corresponding to the above scene information. Here, by determining the corresponding waiting time window for each scene, the adaptive adjustment of the waiting time window is carried out according to the characteristics of the scene in a targeted manner, so that the subsequent guarantee can obtain a high-quality delayed data set. Furthermore, according to the obtained waiting time windows, the delayed data corresponding to each processed data in the above processed data set can be accurately generated to obtain high-quality data, so that the subsequent model can fully learn the data features to accurately output the corresponding value conversion information. Finally, according to the above delayed data set, the initial value conversion information generation model based on cross-domain learning is trained to generate a value conversion information generation model. Here, using the initial value conversion information generation model based on cross-domain learning can utilize and learn the differences in different scenarios and user behavior differences between different domains, so as to obtain accurate value conversion information later. In addition, cross-domain learning joint modeling for different scenarios greatly reduces system resource overhead and model development and deployment costs, bringing actual business value improvement.In summary, by generating a waiting time window corresponding to each scene information and using an initial value conversion information generation model based on cross-domain learning to generate value conversion information, a high-quality data set and a high-precision value conversion information generation model can be obtained through corresponding training. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0027] Figure 1 is a schematic diagram of an application scenario of a model training method according to some embodiments of the present disclosure;

[0028] Figure 2 is a flowchart of some embodiments of the model training method according to the present disclosure;

[0029] Figure 3 is a flowchart of other embodiments of the model training method according to the present disclosure;

[0030] Figure 4 is a flow chart of some embodiments of the item recommendation method according to the present disclosure;

[0031] Figure 5 is a schematic diagram of the structure of some embodiments of the model training device according to the present disclosure;

[0032] Figure 6 is a schematic diagram of the structure of some embodiments of the item recommendation device according to the present disclosure;

[0033] Figure 7 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0034] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0035] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0036] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0037] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0038] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0039] With regard to the collection, storage, and use of user personal information (such as user feature information) involved in this disclosure, before performing the corresponding operations, the relevant organizations or individuals shall fulfill their obligations, including conducting personal information security impact assessments, fulfilling the obligation to inform the personal information subject, and obtaining the authorization and consent of the personal information subject in advance.

[0040] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0041] Figure 1 It is a schematic diagram of an application scenario of the model training method according to some embodiments of the present disclosure.

[0042] exist Figure 1In the application scenario, first, the electronic device 101 can obtain a value conversion data set 102. Each value conversion data includes: user feature information, value conversion item feature information, cross feature information and domain feature information under the corresponding scene information. Then, the electronic device 101 can perform feature engineering processing on each value conversion data in the above value conversion data set 102 to generate a processed data set 103. Next, for each scene information in each scene information, the electronic device 101 can determine the waiting time window corresponding to the above scene information according to the value conversion ratio information corresponding to the target value conversion data subset. The target value conversion data subset is a data set in the value conversion data set corresponding to the above scene information. In this application scenario, each scene information includes: a first scene 104, a second scene 105 and a third scene 106. The target value conversion data subset corresponding to the first scene 104 is a value conversion data subset 107. The target value conversion data subset corresponding to the second scene 105 is a value conversion data subset 108. The target value conversion data subset corresponding to the third scene 106 is a value conversion data subset 109. The waiting time window corresponding to the first scenario 104 is the waiting time window 110. The waiting time window corresponding to the second scenario 105 is the waiting time window 111. The waiting time window corresponding to the third scenario 106 is the waiting time window 112. The waiting time window 110 can be "delayed for 3 days". The waiting time window 111 can be "delayed for 4 days". The waiting time window 112 can be "delayed for 4.5 days". Furthermore, the electronic device 101 can generate the delay data corresponding to each processed data in the above-mentioned processed data set 103 according to the obtained waiting time windows, and obtain the delay data set 113. Finally, the electronic device 101 can perform model training on the initial value conversion information generation model 114 based on cross-domain learning according to the above-mentioned delay data set 113 to generate a value conversion information generation model 115.

[0043] It should be noted that the electronic device 101 can be hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules for providing distributed services, for example, or it can be implemented as a single software or software module. No specific limitation is made here.

[0044] It should be understood that Figure 1 The number of electronic devices in the embodiment is only for illustration. Any number of electronic devices may be provided according to implementation requirements.

[0045] Continue to refer Figure 2, shows a process 200 of some embodiments of the model training method according to the present disclosure. The model training method comprises the following steps:

[0046] Step 201, obtaining a value conversion data set.

[0047] In some embodiments, the execution entity of the above model training method (for example Figure 1The electronic device 101 shown can obtain a value conversion data set through a wired connection or a wireless connection. Among them, each value conversion data includes: user feature information, value conversion item feature information, cross feature information and domain feature information under the corresponding scenario information. The value conversion data can be data that supports value conversion operations. In practice, the value conversion data can be data that has undergone value conversion or data that has not undergone value conversion. The value conversion can be that the value voucher of the corresponding item has been used. In practice, for the e-commerce field, the value conversion operation can be a voucher or coupon use operation. The value conversion can be that the voucher or coupon has been used. The value conversion cannot be performed can be that the voucher or coupon has not been used. The value voucher can be a voucher or coupon. The value conversion data can be data of the voucher or coupon of the corresponding product that has been used. The user feature information can be the feature information of the target user. The target user can be the executor of the value conversion operation. In practice, the target user can be the user who is given a voucher or coupon. The user characteristic information may include but is not limited to at least one of the following: user name, user age, user gender, user behavior characteristic information, user delivery address, user portrait. The user behavior characteristic information may be characteristic information of user behavior. The user behavior may be but is not limited to at least one of the following: consumption preference behavior, consumption level characteristics. The value conversion item characteristic information may be characteristic information of the item as the object of value conversion operation. The value conversion item characteristic information may include but is not limited to at least one of the following: the time of receiving the value voucher of the item, the time of conversion of each value voucher corresponding to the item, the expiration time of each value voucher corresponding to the item, and the conversion rate of each value voucher corresponding to the item. In practice, the item as the object of value conversion operation may be a voucher product or a coupon product. The cross-characteristic information may be the characteristic information of the user performing characteristic statistics on the value voucher corresponding to the value conversion item. For example, the cross-characteristic information may include but is not limited to at least one of the following: the user's collection information of the value voucher corresponding to the value conversion item in the scenario, the user's conversion information of the value voucher corresponding to the value conversion item in the scenario, the expiration information of the value voucher corresponding to the value conversion item in the scenario, and the conversion rate of the value voucher corresponding to the value conversion item in the scenario. The domain feature information corresponding to the scene information may be the domain feature information corresponding to the scene information of the value conversion data. The domain feature information may be scene feature information. Specifically, the domain feature information may include but is not limited to at least one of the following: the collection information of each value voucher corresponding to the value conversion item in the scene, the conversion information of each value voucher corresponding to the value conversion item in the scene, the expiration of each value voucher corresponding to the value conversion item in the scene, and the conversion rate of each value voucher corresponding to the value conversion item in the scene.

[0048] Step 202: Perform feature engineering processing on each value conversion data in the above value conversion data set to generate a processed data set.

[0049] In some embodiments, the execution entity may perform feature engineering processing on each value conversion data in the value conversion data set to generate a processed data set.

[0050] As an example, the above-mentioned execution entity can normalize each type of feature information in each value conversion data to generate a normalized feature information set and obtain normalized value conversion data as processed data.

[0051] In some optional implementations of some embodiments, the above-mentioned performing feature engineering processing on each value conversion data in the above-mentioned value conversion data set to generate a processed data set may include the following steps:

[0052] The first step is to perform feature screening on each data in the above value conversion data set to generate a screened data set.

[0053] As an example, at least one of the following processing steps is performed on the above-mentioned value conversion dataset: remove data with attribute feature data importance of 0, remove data with feature coverage less than 30%, remove data corresponding to any feature in the feature group with feature similarity greater than 90%, and remove data with a correlation between the label and the corresponding feature less than a preset threshold, to obtain a filtered dataset.

[0054] The second step is to perform data preprocessing on each data in the above-screened data set to generate the above-mentioned processed data set.

[0055] As an example, first, the execution subject can remove abnormal data from the screened data set and fill in the default values ​​of the corresponding features to obtain a preliminary processed data set. Then, MD5hash encoding is performed on the data whose corresponding features are discrete features in the preliminary processed data set, and dimension difference normalization is performed on the features whose corresponding features are continuous features to generate a processed data set.

[0056] Step 203: for each scene information in the corresponding scene information, determine the waiting time window corresponding to the scene information according to the value conversion ratio information corresponding to the target value conversion data subset.

[0057] In some embodiments, the execution subject may determine the waiting time window corresponding to the above-mentioned scenario information according to the value conversion ratio information corresponding to the target value conversion data subset for each scenario information in the corresponding scenario information. The value conversion ratio information may be a preset ratio value. For example, the value conversion ratio information may be 95%. The value conversion ratio information corresponding to the target value conversion data subset may be the ratio information of the value conversion occurring in each value voucher corresponding to the target value conversion data subset. The waiting time window may characterize the time window of the waiting time for the corresponding value voucher conversion to occur in the waiting value conversion data. For example, the waiting time window may be a future week after the current time. For another example, the waiting time window may be a 3-day time window after a certain time node. The above-mentioned target value conversion data subset is a data set in the above-mentioned value conversion data set corresponding to the above-mentioned scenario information. That is, each value conversion data in the target value conversion data subset is the conversion data for the corresponding value voucher under the premise of the scenario information corresponding to the scenario.

[0058] As an example, first, in response to determining that the conversion speed of each value conversion voucher corresponding to the target value conversion data subset is less than a preset speed, determine the conversion time when the conversion speed is less than the preset speed. Then, set the conversion duration. Finally, generate a waiting time window according to the conversion duration and the conversion time.

[0059] Step 204 : generating delay data corresponding to each processed data in the processed data set according to each obtained waiting time window, and obtaining a delay data set.

[0060] In some embodiments, the execution subject may generate delay data corresponding to each processed data in the processed data set in various ways according to the obtained waiting time windows to obtain a delay data set. The delay data is data that corresponds to a more accurate value conversion label relative to the processed data. The value conversion label may be a label indicating whether the value voucher corresponding to the value conversion data actually undergoes value conversion. Specifically, the value conversion label may include: a label indicating that the user has performed a conversion operation on the value voucher, and a label indicating that the user has not performed a conversion operation on the value voucher.

[0061] In some optional implementations of some embodiments, the processed data further includes: a value conversion label.

[0062] Optionally, the generating of delay data corresponding to each processed data in the processed data set according to each obtained waiting time window to obtain the delay data set may include the following steps:

[0063] For each of the above-mentioned scene information, the following second generation step is performed:

[0064] Sub-step 1: The execution entity may determine a waiting time window corresponding to the scene information as a target waiting time window.

[0065] In sub-step 2, the execution entity may determine a processed data subset in the processed data set corresponding to the scene information as a target processed data subset.

[0066] Sub-step 3, according to the above-mentioned target waiting time window, the value conversion label corresponding to each processed data in the above-mentioned target processed data subset is updated to generate updated data as delayed data.

[0067] As an example, for each processed data, first, the execution subject may determine whether the processed data undergoes a transformation change corresponding to the value voucher within the target waiting time window. Then, in response to determining that a change has occurred, the latest change result corresponding to the value voucher is determined as the value transformation tag corresponding to the processed data to generate delayed data.

[0068] Step 205 , based on the above-mentioned delayed data set, model training is performed on the initial value conversion information generation model based on cross-domain learning to generate a value conversion information generation model.

[0069] In some embodiments, the execution entity may perform model training on the initial value conversion information generation model based on cross-domain learning according to the delayed data set to generate a value conversion information generation model. The initial value conversion information generation model may be a value conversion information generation model that has not yet been trained. The value conversion information generation model may be a neural network model that generates value conversion information. In practice, the value conversion information may be the probability information of the conversion of the value voucher. In practice, the value conversion information generation model may be a tree model or a deep learning model. For example, the value conversion information generation model may be a LightGBM model or a STAR neural network model.

[0070] It should be noted that the model for generating value conversion information is the STAR neural network model, which uses the Partitioned Normalization (PN) layer to normalize the value conversion data subsets corresponding to each scene information. For different scenarios, the PN layer uses private statistics and learnable parameters to achieve adaptive normalization of value conversion data subsets corresponding to different scenarios, retaining scene difference information for more refined modeling. At the same time, the auxiliary network is used to take the domain feature information set as input to capture domain differences in a direct and simple way. Finally, the shared parameters of the center and the scene-private parameters of a specific domain are updated through a star topology network structure. Among them, the model parameters can be optimized by minimizing the cross entropy loss.

[0071] In some optional implementations of some embodiments, the above-mentioned model training of the initial value conversion information generation model based on cross-domain learning according to the above-mentioned delayed data set to generate the value conversion information generation model may include the following steps:

[0072] In the first step, the delay data set is divided according to a predetermined historical time period sequence to generate a delay data subset sequence. Each delay data subset has a corresponding historical time period. Each historical time period in the historical time period sequence is continuous. For example, the historical time period sequence may include: the historical time period corresponding to [T-40, T-6], the historical time period corresponding to [T-5], and the historical time period corresponding to [T-4]. T may be the current time. T-40 may represent the 40th day before the current time. T-6 may represent the 6th day before the current time.

[0073] In the second step, each delayed data subset in the delayed data subset sequence is determined as a training data set, a verification data set and a test data set.

[0074] As an example, the above-mentioned execution entity can determine the delay data subset corresponding to the historical time period farthest from the current time as the training data set, and the delay data subset corresponding to the historical time period closest to the current time as the test data set. The remaining delay data subsets are determined as verification data sets. For example, the delay data subset in the historical time period corresponding to [T-40, T-6] is determined as the training data set. The delay data subset in the historical time period corresponding to [T-5] is determined as the verification data set. The delay data subset in the historical time period corresponding to [T-4] is determined as the test data set.

[0075] The third step is to perform model training on the initial value conversion information generation model based on the training data set, the verification data set and the test data set to generate the value conversion information generation model.

[0076] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the model training method of some embodiments of the present disclosure, value conversion information characterizing the value conversion operation performed by the user on the target item can be accurately generated under different scene information. Specifically, the reason for the inaccuracy of the relevant value conversion information is that the training of the value conversion information generation model does not take into account the problem of sample conversion delay, resulting in the inaccuracy of the training sample set. In addition, when modeling the model, the delayed feedback problem is not taken into account, resulting in the value conversion information generated by the generated value conversion information generation model being inaccurate. Based on this, the model training method of some embodiments of the present disclosure first obtains a value conversion data set, wherein each value conversion data includes: user feature information and value conversion item feature information, cross feature information and domain feature information under corresponding scene information. Here, the acquired value conversion data set is used for feature learning of subsequent models. Cross feature information allows subsequent models to learn more interactive features between users and value conversion items, so that the subsequent model outputs more accurate value conversion information. In addition, through the domain feature information corresponding to each different scene through the value conversion data set, the subsequent model learns more feature information. Then, feature engineering is performed on each value conversion data in the above value conversion data set to generate a processed data set to further improve the data quality and fully reflect the feature information of the key features. Next, for each scene information in the corresponding scene information, according to the value conversion ratio information corresponding to the target value conversion data subset, the waiting time window corresponding to the above scene information can be accurately determined, wherein the target value conversion data subset is a data set in the above value conversion data set corresponding to the above scene information. Here, by determining the corresponding waiting time window for each scene, the adaptive adjustment of the waiting time window is carried out in a targeted manner according to the characteristics of the scene, so that a high-quality delayed data set can be obtained in the subsequent guarantee. Furthermore, according to the obtained waiting time windows, the delayed data corresponding to each processed data in the above processed data set can be accurately generated to obtain high-quality data, so that the subsequent model can fully learn the data features to accurately output the corresponding value conversion information. Finally, according to the above delayed data set, the initial value conversion information generation model based on cross-domain learning is trained to generate a value conversion information generation model. Here, using the initial value conversion information generation model based on cross-domain learning can utilize and learn the differences in different scenarios and user behavior differences between different domains, so as to obtain accurate value conversion information later. In addition, cross-domain learning joint modeling for different scenarios greatly reduces system resource overhead and model development and deployment costs, bringing actual business value improvement.In summary, by generating a waiting time window corresponding to each scene information and using an initial value conversion information generation model based on cross-domain learning to generate value conversion information, a high-quality data set and a high-precision value conversion information generation model can be obtained through corresponding training.

[0077] Further references Figure 3 , shows a process 300 of other embodiments of the model training method according to the present disclosure. The model training method comprises the following steps:

[0078] Step 301, obtaining a value conversion data set.

[0079] Step 302: Perform feature engineering processing on each value conversion data in the above value conversion data set to generate a processed data set.

[0080] Step 303: For each scene information in the corresponding scene information, the following first generation step is performed:

[0081] Step 3031, in response to determining that the value conversion ratio information corresponding to the above target value conversion data subset reaches the target value, and / or each value conversion voucher corresponding to the above target value conversion data subset reaches the conversion date, determine the end conversion time point.

[0082] In some embodiments, in response to determining that the value conversion ratio information corresponding to the target value conversion data subset reaches the target value, and / or each value conversion voucher corresponding to the target value conversion data subset reaches the conversion date, the execution entity (e.g. Figure 1 The electronic device shown in the figure can determine the end conversion time point. The target value can be a preset value. For example, the target value can be 95%. The conversion date can be the final convertible date corresponding to the value voucher. The end conversion time point can be the last time point in the waiting time window. The end conversion time point can be the time point under the condition that each value conversion voucher reaches the conversion date.

[0083] Step 3032, obtaining a predetermined starting conversion time point.

[0084] In some embodiments, the execution entity may obtain a predetermined start conversion time point, wherein the start conversion time point may be a start time point in the waiting time window.

[0085] Step 3033: Generate a waiting time window corresponding to the above-mentioned end conversion time point and the above-mentioned start conversion time point.

[0086] In some embodiments, the execution entity may generate a waiting time window corresponding to the end conversion time point and the start conversion time point.

[0087] As an example, the execution entity may use the start conversion time point as the left boundary time point of the waiting time window and the end conversion time point as the right boundary time point of the waiting time window to generate the waiting time window.

[0088] Step 304: generating delay data corresponding to each processed data in the processed data set according to each obtained waiting time window, to obtain a delay data set.

[0089] Step 305 , based on the above-mentioned delayed data set, model training is performed on the initial value conversion information generation model based on cross-domain learning to generate a value conversion information generation model.

[0090] In some embodiments, the specific implementation of steps 301-302, 304-305 and the technical effects thereof can be referred to in Figure 2 The corresponding steps 201-202 and 204-205 in the embodiment are not described in detail here.

[0091] from Figure 3 It can be seen that Figure 2 Compared with the description of some corresponding embodiments, Figure 3 In the process 300 of the model training method in some corresponding embodiments, when it is determined that the value conversion ratio information corresponding to the target value conversion data subset reaches the target value, and / or each value conversion voucher corresponding to the target value conversion data subset reaches the conversion date, the end conversion time point is accurately generated, so as to combine the start conversion time point to generate a waiting time window corresponding to the accurate scene information. Based on the accurate waiting time windows, a value conversion information generation model with higher output accuracy can be obtained.

[0092] Continue to refer Figure 4 , shows a process 400 of some embodiments of the item recommendation method according to the present disclosure. The item recommendation method comprises the following steps:

[0093] Step 401: for each scene information in each scene information, perform the following recommended steps:

[0094] Step 4011, determine the value conversion item information set corresponding to the above scene information.

[0095] In some embodiments, the execution subject (e.g., electronic device) of the above-mentioned item recommendation method may determine the value conversion item information set corresponding to the above-mentioned scene information. The value conversion item information may be item information of items that support value conversion. Specifically, the value conversion item set corresponding to the value conversion item information set may be each item recommended under the scene information.

[0096] Step 4012, using the pre-trained value conversion information generation model, generates item conversion information corresponding to each value conversion item information in the above value conversion item information set to obtain the item conversion information set.

[0097] In some embodiments, the execution subject may use a pre-trained value conversion information generation model to generate item conversion information corresponding to each value conversion item information in the value conversion item information set to obtain an item conversion information set. The value conversion information generation model is generated based on a model training method. The item conversion information may be probability information of the value conversion of the item corresponding to the value conversion item information. In practice, for e-commerce scenarios, the item conversion information may be the probability of using the corresponding coupon of the item to purchase the item.

[0098] As an example, first, determine the user group characteristic information corresponding to each value conversion item information. Then, based on the above user group characteristic information, generate the corresponding user characteristic information, the corresponding value conversion item characteristic information, the corresponding cross-characteristic information and the domain characteristic information under the corresponding scene information as the first user characteristic information, the first value conversion item characteristic information, the first cross-characteristic information and the first domain characteristic information. Specifically, the first user characteristic information and the first value conversion item characteristic information can be generated by a related generative neural network model. The first cross-characteristic information and the first domain characteristic information are determined by a time series neural network model. Finally, the first user characteristic information, the first value conversion item characteristic information, the first cross-characteristic information and the first domain characteristic information are input into the value conversion information generation model to generate the item conversion information corresponding to the value conversion item information.

[0099] In some optional implementations of some embodiments, the above-mentioned use of the pre-trained value conversion information generation model to generate the item conversion information corresponding to each value conversion item information in the above-mentioned value conversion item information set may include the following steps:

[0100] The first step is to use the above-mentioned value conversion information generation model to determine the value conversion information corresponding to each user in the target user set, and obtain a value conversion information set. Among them, the above-mentioned target user set is a user set corresponding to the above-mentioned value conversion item information. The above-mentioned target user set is a user set that has a value conversion relationship with the value conversion item information. In practice, the above-mentioned target user set can be a user set that has performed value circulation on the items corresponding to the value conversion item information, and can also be a beneficiary group of the items corresponding to the value conversion item information. In practice, the target user set can be a user set that has purchased the items corresponding to the value conversion item information. In addition, the target user set can also be a user set with preset user characteristics. For example, the preset user characteristics can be but are not limited to at least one of the following: long-term login to the value application, and the number of value circulations is greater than the predetermined number.

[0101] As an example, first, the user characteristic information, value conversion item characteristic information, cross characteristic information and domain characteristic information under the corresponding scenario information corresponding to the above target user are obtained as target user characteristic information, target value conversion item characteristic information, target cross characteristic information and target domain characteristic information respectively. Then, the above target user characteristic information, target value conversion item characteristic information, target cross characteristic information and target domain characteristic information are input into the value conversion information generation model to generate the value conversion information corresponding to the target user.

[0102] The second step is to generate the above-mentioned item conversion information based on the above-mentioned value conversion information set.

[0103] As an example, the execution entity may add up the conversion values ​​corresponding to the individual value conversion information in the value conversion information set to generate added information as item conversion information.

[0104] As another example, the execution entity may average each value conversion information in the value conversion information set to generate average information as item conversion information.

[0105] Step 4013, according to the above-mentioned item conversion information set, set the recommended information corresponding to each value conversion item information in the above-mentioned value conversion item information set to obtain the recommended information set.

[0106] In some embodiments, the execution subject may set the recommendation information corresponding to each value conversion item information in the value conversion item information set according to the item conversion information set to obtain the recommendation information set. In practice, the recommendation information may be the recommendation weight of the item corresponding to the value conversion item information. For example, the recommendation information may be the rule weight of the recommendation rule corresponding to the value conversion item information.

[0107] As an example, for each value conversion item information in the value conversion item information set, first, the value conversion item information corresponding to the value conversion item information is determined as the target value conversion item information. Then, the numerical interval corresponding to the above target item conversion information is determined as the target numerical interval. Next, the recommendation weight corresponding to the target numerical interval is determined as the recommendation information corresponding to the value conversion item information.

[0108] Step 4014: Execute item recommendation for at least one value conversion item information according to the above recommendation information set.

[0109] In some embodiments, the execution subject may perform item recommendation for at least one value conversion item information according to the recommendation information set, wherein the at least one value conversion item information is an item information subset in the target value conversion item information set.

[0110] As an example, for each value conversion item in the at least one value conversion item information, first, the execution subject may determine the recommendation information corresponding to the value conversion item as the target recommendation information, and then, perform the item recommendation of the value conversion item according to the target recommendation information.

[0111] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: the item recommendation method of some embodiments of the present disclosure can accurately and efficiently implement the recommendation of items.

[0112] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a model training device, and these device embodiments are Figure 2 Corresponding to the method embodiments shown, the model training device can be specifically applied to various electronic devices.

[0113] like Figure 5As shown, a model training device 500 includes: an acquisition unit 501, a processing unit 502, a determination unit 503, a generation unit 504 and a training unit 505. Among them, the acquisition unit 501 is configured to acquire a value conversion data set, wherein each value conversion data includes: user feature information, value conversion item feature information, cross feature information and domain feature information under the corresponding scene information; the processing unit 502 is configured to perform feature engineering processing on each value conversion data in the above value conversion data set to generate a processed data set; the determination unit 503 is configured to determine the waiting time window corresponding to the above scene information according to the value conversion ratio information corresponding to the target value conversion data subset for each scene information corresponding to each scene information, wherein the above target value conversion data subset is the data set in the above value conversion data set that corresponds to the above scene information; the generation unit 504 is configured to generate delay data corresponding to each processed data in the above processed data set according to the obtained waiting time windows to obtain a delay data set; the training unit 505 is configured to perform model training on the initial value conversion information generation model based on cross-domain learning according to the above delay data set to generate a value conversion information generation model.

[0114] In some optional implementations of some embodiments, the determination unit 503 can be further configured to: for each scenario information in the corresponding scenario information, perform the following first generation step: in response to determining that the value conversion ratio information corresponding to the above-mentioned target value conversion data subset reaches the target value, and / or each value conversion voucher corresponding to the above-mentioned target value conversion data subset reaches the conversion date, determine the end conversion time point; obtain a predetermined start conversion time point; generate a waiting time window corresponding to the above-mentioned end conversion time point and the above-mentioned start conversion time point.

[0115] In some optional implementations of some embodiments, the processed data further includes: a value conversion tag; and the generation unit 504 may be further configured to: for each of the above-mentioned scene information, perform the following second generation step: determine the waiting time window corresponding to the above-mentioned scene information as the target waiting time window; determine the processed data subset in the above-mentioned processed data set corresponding to the above-mentioned scene information as the target processed data subset. According to the above-mentioned target waiting time window, the value conversion tag corresponding to each processed data in the above-mentioned target processed data subset is updated to generate updated data as delayed data.

[0116] In some optional implementations of some embodiments, the training unit 505 can be further configured to: divide the above-mentioned delay data set according to a predetermined historical time period sequence to generate a delay data subset sequence, wherein each delay data subset has a corresponding historical time period; determine each delay data subset in the above-mentioned delay data subset sequence as a training data set, a verification data set and a test data set respectively; and perform model training on the above-mentioned initial value conversion information generation model according to the above-mentioned training data set, the above-mentioned verification data set and the above-mentioned test data set to generate the above-mentioned value conversion information generation model.

[0117] In some optional implementations of some embodiments, the processing unit 502 may be further configured to: perform feature screening on each data in the above-mentioned value conversion data set to generate a screened data set; perform data preprocessing on each data in the above-mentioned screened data set to generate the above-mentioned processed data set.

[0118] It is understandable that the units recorded in the model training device 500 are similar to the reference Figure 2 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the model training device 500 and the units contained therein, and will not be repeated here.

[0119] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of an item recommendation device. These device embodiments are similar to Figure 4 Corresponding to the method embodiments shown, the item recommendation device can be specifically applied to various electronic devices.

[0120] like Figure 6 As shown, an item recommendation device 600 includes: an execution unit 601. The execution unit 601 is configured to perform the following recommendation steps for each scene information in each scene information: determine the value conversion item information set corresponding to the above scene information; generate item conversion information corresponding to each value conversion item information in the above value conversion item information set using a pre-trained value conversion information generation model to obtain an item conversion information set, wherein the above value conversion information generation model is generated based on a model training method; according to the above item conversion information set, set the recommendation information corresponding to each value conversion item information in the above value conversion item information set to obtain a recommendation information set; according to the above recommendation information set, perform item recommendation for at least one value conversion item information, wherein the at least one value conversion item information is an item information subset in the target value conversion item information set.

[0121] It is understandable that the units recorded in the item recommendation device 600 are similar to those in the reference Figure 4 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the item recommendation device 600 and the units included therein, and will not be described in detail here.

[0122] Reference below Figure 7 , which shows an electronic device (eg, Figure 1 Schematic diagram of the structure of the electronic device 101)700. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0123] like Figure 7 As shown, the electronic device 700 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory 702 or a program loaded from a storage device 708 into a random access memory 703. Various programs and data required for the operation of the electronic device 700 are also stored in the random access memory 703. The processing device 701, the read-only memory 702, and the random access memory 703 are connected to each other via a bus 704. An input / output interface 705 is also connected to the bus 704.

[0124] Typically, the following devices may be connected to the input / output interface 705: an input device 706 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 707 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 708 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 709. The communication device 709 may allow the electronic device 700 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 7 The electronic device 700 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 7 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0125] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from a network through a communication device 709, or installed from a storage device 708, or installed from a read-only memory 702. When the computer program is executed by the processing device 701, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0126] It should be noted that the computer-readable medium in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0127] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0128] The computer-readable medium may be included in the electronic device; or it may exist independently without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: obtains a value conversion data set, wherein each value conversion data includes: user feature information, value conversion item feature information, cross feature information, and domain feature information under corresponding scene information; performs feature engineering processing on each value conversion data in the value conversion data set to generate a processed data set; for each scene information in each corresponding scene information, determines the waiting time window corresponding to the scene information according to the value conversion ratio information corresponding to the target value conversion data subset, wherein the target value conversion data subset is the data set in the value conversion data set corresponding to the scene information; generates delay data corresponding to each processed data in the processed data set according to the obtained waiting time windows, and obtains a delay data set; performs model training on the initial value conversion information generation model based on cross-domain learning according to the delay data set to generate a value conversion information generation model. For each scene information in each scene information, the following recommendation steps are performed: determine the value conversion item information set corresponding to the above scene information; use a pre-trained value conversion information generation model to generate item conversion information corresponding to each value conversion item information in the above value conversion item information set, and obtain the item conversion information set, wherein the above value conversion information generation model is generated based on the model training method; according to the above item conversion information set, set the recommendation information corresponding to each value conversion item information in the above value conversion item information set, and obtain the recommendation information set; according to the above recommendation information set, perform item recommendation for at least one value conversion item information, wherein the at least one value conversion item information is a subset of the item information in the target value conversion item information set.

[0129] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0130] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0131] The units described in some embodiments of the present disclosure may be implemented by software or hardware. The described units may also be provided in a processor, for example, may be described as: a processor includes an acquisition unit, a processing unit, a determination unit, a generation unit, and a training unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the acquisition unit may also be described as a "unit for acquiring a value conversion data set".

[0132] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0133] Some embodiments of the present disclosure also provide a computer program product, including a computer program, which implements any of the above-mentioned model training methods or item recommendation methods when executed by a processor.

[0134] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A model training method, comprising: Acquire a value conversion data set, wherein each value conversion data includes: user feature information, value conversion item feature information, cross feature information, and domain feature information under corresponding scenario information; Performing feature engineering processing on each value conversion data in the value conversion data set to generate a processed data set; For each scene information in each scene information, determine the waiting time window corresponding to the scene information according to the value conversion ratio information corresponding to the target value conversion data subset, wherein the target value conversion data subset is a data set in the value conversion data set corresponding to the scene information; According to each obtained waiting time window, generating delay data corresponding to each processed data in the processed data set to obtain a delay data set; According to the delayed data set, model training is performed on an initial value conversion information generation model based on cross-domain learning to generate a value conversion information generation model.

2. The method according to claim 1, wherein: For each scene information in each scene information, determining the waiting time window corresponding to the scene information according to the value conversion ratio information corresponding to the target value conversion data subset includes: For each scene information in the corresponding scene information, the following first generation step is performed: In response to determining that the value conversion ratio information corresponding to the target value conversion data subset reaches the target value, and / or each value conversion voucher corresponding to the target value conversion data subset reaches the conversion date, determining the end conversion time point; Obtaining a predetermined starting conversion time point; A waiting time window corresponding to the end conversion time point and the start conversion time point is generated.

3. The method according to claim 1, wherein: The processed data also includes: value conversion tags; and The generating, according to each obtained waiting time window, delay data corresponding to each processed data in the processed data set to obtain the delay data set comprises: For each of the scene information, the following second generation step is performed: Determine a waiting time window corresponding to the scene information as a target waiting time window; Determine a processed data subset in the processed data set corresponding to the scene information as a target processed data subset; According to the target waiting time window, a value conversion label corresponding to each processed data in the target processed data subset is updated to generate updated data as delayed data.

4. The method according to claim 1, wherein: The performing model training on the initial value conversion information generation model based on cross-domain learning according to the delayed data set to generate a value conversion information generation model includes: According to a predetermined historical time period sequence, the delay data set is divided into data to generate a delay data subset sequence, wherein each delay data subset has a corresponding historical time period; Determine each delayed data subset in the delayed data subset sequence as a training data set, a verification data set, and a test data set; The initial value conversion information generation model is trained according to the training data set, the verification data set and the test data set to generate the value conversion information generation model.

5. The method according to claim 1, wherein: The performing feature engineering processing on each value conversion data in the value conversion data set to generate a processed data set includes: Performing feature screening on each data in the value conversion data set to generate a screened data set; Data preprocessing is performed on each data in the screened data set to generate the processed data set.

6. An item recommendation method, comprising: For each of the scenario information in each scenario information, perform the following recommended steps: Determine a value conversion item information set corresponding to the scene information; Using a pre-trained value conversion information generation model, generating item conversion information corresponding to each value conversion item information in the value conversion item information set, to obtain an item conversion information set, wherein the value conversion information generation model is generated based on the method according to any one of claims 1 to 5; According to the item conversion information set, setting the recommended information corresponding to each value conversion item information in the value conversion item information set to obtain a recommended information set; According to the recommendation information set, item recommendation is performed for at least one value conversion item information, wherein the at least one value conversion item information is an item information subset in the target value conversion item information set.

7. The method according to claim 6, wherein: The generating of the item conversion information corresponding to each value conversion item information in the value conversion item information set by using the pre-trained value conversion information generating model comprises: Using the value conversion information generation model, determine the value conversion information corresponding to each user in the target user set to obtain a value conversion information set, wherein the target user set is a user set corresponding to the value conversion item information; The item conversion information is generated according to the value conversion information set.

8. A model training device, comprising: An acquisition unit is configured to acquire a value conversion data set, wherein each value conversion data includes: user feature information, value conversion item feature information, cross feature information, and domain feature information under corresponding scene information; A processing unit, configured to perform feature engineering processing on each value conversion data in the value conversion data set to generate a processed data set; a determining unit configured to determine, for each of the corresponding scene information, a waiting time window corresponding to the scene information according to value conversion ratio information corresponding to a target value conversion data subset, wherein the target value conversion data subset is a data set in the value conversion data set corresponding to the scene information; A generating unit is configured to generate delay data corresponding to each processed data in the processed data set according to each obtained waiting time window, so as to obtain a delay data set; The training unit is configured to perform model training on the initial value conversion information generation model based on cross-domain learning according to the delayed data set to generate a value conversion information generation model.

9. An item recommendation device, comprising: The execution unit is configured to execute the following recommendation steps for each of the scene information: determining a value conversion item information set corresponding to the scene information; Utilize a pre-trained value conversion information generation model to generate item conversion information corresponding to each value conversion item information in the value conversion item information set to obtain an item conversion information set, wherein the value conversion information generation model is generated based on the method described in any one of claims 1 to 5; according to the item conversion information set, set recommendation information corresponding to each value conversion item information in the value conversion item information set to obtain a recommendation information set; according to the recommendation information set, perform item recommendations for at least one value conversion item information, wherein the at least one value conversion item information is a subset of item information in the target value conversion item information set.

10. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

11. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

12. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Resource recommendation method and device and electronic equipment

    CN110413877A

  • Data processing method and related device

    CN113869964A

  • Training method and device, equipment, storage medium and computer program product

    CN115204388A

  • Recommendation model training method and device, recommendation method and device and computer readable medium

    CN115917535A

  • Information recommendation method and device, equipment, storage medium and program product

    CN116628349A