Sample generation method and device, electronic equipment and computer readable medium
By obtaining and utilizing the shallow behavior data of the target user and generating virtual transaction events and user value, the accuracy and effectiveness of the promotion information delivery model are solved, and the accuracy and adaptability of the promotion information delivery are improved.
Patent Information
- Application Number
- CN202410951917.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-15
AI Technical Summary
In the prior art, the promotion information delivery model has low accuracy and poor effectiveness, mainly due to the small number of historical transaction events and the single value, and it is unable to adapt to user changes.
By obtaining shallow behavior data of the target user in the target application, input it into the pre-trained user value prediction model, generating virtual transaction events and user value, and building sample data for retraining the promotion information delivery model.
It improves the accuracy and effectiveness of the promotion information delivery model, expands transaction events, enriches the value distribution, enables the model to learn and adapt to changes in user behavior faster, and improves the accuracy of promotion information delivery.
Smart Images

Figure CN118964971A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present application relate to the field of computer technology, and in particular to a sample generation method, device, electronic device, and computer-readable medium. Background Art
[0002] With the development of computer technology, it is usually possible to use automation technology and machine learning algorithms to train promotion information delivery models, for example, AC 3.0 (Application Campaign 3.0, the third edition of the application promotion model), through which the promotion information of the application is delivered to potential users.
[0003] In the prior art, the sample data for training the promotion information delivery model usually comes from the historical transaction events and their values of historical users in the application. However, on the one hand, the number of historical transaction events is small and the value is relatively single, which makes the sample size small, resulting in low accuracy of the promotion information delivery model. On the other hand, the promotion information delivery model obtained by training with historical data cannot adapt to changes in users, resulting in poor effectiveness. Summary of the invention
[0004] The embodiments of the present application propose a sample generation method, device, electronic device and computer-readable medium to solve the technical problems in the prior art that the promotion information delivery model has low accuracy and poor effectiveness.
[0005] In a first aspect, an embodiment of the present application provides a sample generation method, the method comprising: obtaining shallow behavior data of a target user in a target application, the shallow behavior data being behavior data generated within a preset time period after the target user downloads the target application and having a probability of occurrence greater than a first threshold; inputting the shallow behavior data into a pre-trained user value prediction model to obtain a user value of the target user, the user value being the transaction value of the target user in the target application within a preset number of days; constructing a virtual transaction event when the user value is greater than a second threshold, and generating training samples based on the virtual transaction event and the user value, the training samples being used to retrain a pre-trained promotion information delivery model.
[0006] In a second aspect, an embodiment of the present application provides a sample generation device, which includes: an acquisition unit, used to acquire shallow behavior data of a target user in a target application, wherein the shallow behavior data is behavior data generated within a preset time period after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold; a prediction unit, used to input the shallow behavior data into a pre-trained user value prediction model to obtain the user value of the target user, wherein the user value is the transaction value of the target user in the target application within a preset number of days; a generation unit, used to construct a virtual transaction event when the user value is greater than a second threshold, and generate a training sample based on the virtual transaction event and the user value, wherein the training sample is used to retrain the pre-trained promotion information delivery model.
[0007] In a third aspect, an embodiment of the present application provides 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 embodiment of the first aspect.
[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any embodiment of the first aspect.
[0009] The sample generation method, device, electronic device and computer-readable medium provided in the embodiments of the present application obtain shallow behavior data of the target user in the target application, where the shallow behavior data is the behavior data generated within a preset time after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold value; then the shallow behavior data is input into a pre-trained user value prediction model to obtain the user value of the target user, where the user value is the transaction value of the target user in the target application within a preset number of days; then, when the user value is greater than the second threshold value, a virtual transaction event is constructed, thereby generating sample data for retraining the pre-trained promotion information delivery model based on the virtual transaction event and the value. On the one hand, this implementation simulates the user's transaction behavior by constructing a virtual transaction event and predicts its value, thereby expanding the transaction events and enriching the value distribution, solving the problem of sparse user transaction events and single value, and increasing the sample size of the training promotion information delivery model, thereby improving the accuracy of the promotion information delivery model. On the other hand, this implementation method uses shallow behavior data as the data basis for user value prediction, and can transmit sample data back at an earlier time, so that the promotion information delivery model can learn and adapt to changes in user behavior more quickly, thereby improving the effectiveness of the promotion information delivery model. The promotion information delivery model can be used to deliver promotion information, which can improve the accuracy of promotion information delivery. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:
[0011] Figure 1 is a flow chart of an embodiment of a sample generation method of the present application;
[0012] Figure 2 is a schematic diagram of the interactive process of the sample generation method of the present application;
[0013] Figure 3 is a schematic structural diagram of an embodiment of a sample generating device of the present application;
[0014] Figure 4 It is a schematic diagram of the structure of an electronic device used to implement an embodiment of the present application. DETAILED DESCRIPTION
[0015] All actions to obtain signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where they are located, and with the authorization given by the owner of the corresponding device.
[0016] The present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It should also be noted that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.
[0017] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0018] Please refer to Figure 1 , which shows a process 100 of an embodiment of a sample generation method according to the present application. The sample generation method can be applied to various electronic devices with data processing functions. The execution subject of the sample generation method can be a processor in the above electronic device.
[0019] The sample generation method comprises the following steps:
[0020] Step 101, obtaining shallow behavior data of a target user in a target application.
[0021] In this embodiment, the target application can be any specified application, for example, it can be an audio and video application, a reading application, etc. The target user can be a new user of the target application. The above-mentioned new user can be a user who has downloaded the target application for less than the specified time. The above-mentioned specified time can be set as needed, for example, it can be set to 1 day, 3 days, 7 days, etc. It should be noted that the target users can be further limited as needed. For example, when the sample generation method is applied to the overseas promotion information delivery scenario, the target users can be limited to overseas users.
[0022] In this embodiment, shallow behavior data refers to behavior data generated within a preset time after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold. The preset time and the first threshold can be set as needed. For example, if the preset time is 3 days and the first threshold is 60%, then shallow behavior data refers to behavior data generated within 3 days after the target user downloads the target application and the probability of the behavior occurring is greater than 60%. In practice, the correspondence between behavior and its probability of occurrence can be obtained in advance by statistically analyzing the behavior data of a large number of users, which is not specifically limited here.
[0023] Step 102, input the shallow behavior data into a pre-trained user value prediction model to obtain the user value of the target user.
[0024] In this embodiment, a trained user value prediction model may be pre-stored. The user value prediction model may be pre-trained by a machine learning algorithm. The above user value prediction model may be used to predict user value based on user behavior data. User value is the transaction value of the target user in the target application within a preset number of days, that is, the revenue obtained from the target user in the future. In actual applications, the above revenue may include the amount paid by the user for subscribing to membership in the target application. Here, shallow behavior data may be input into the pre-trained user value prediction model to obtain the user value of the target user.
[0025] Step 103: when the user value is greater than the second threshold, construct a virtual transaction event, and generate a training sample based on the virtual transaction event and the user value.
[0026] In this embodiment, when the user value is greater than the second threshold, the target user can be considered a high-value user, and the target user will perform a transaction behavior later. At this time, a virtual transaction event can be constructed. Since the target user has not yet performed a real transaction behavior, the above virtual transaction event is not a real transaction event, but a transaction event pre-constructed based on the shallow behavior data of the target user to infer that the target user will perform a real transaction behavior later. The virtual transaction event is used to indicate that the target user will perform a real transaction behavior later. In practice, a virtual transaction event can be constructed by extracting and recording an identifier used to indicate a virtual transaction event. The above identifier can be a pre-set string.
[0027] In this embodiment, training samples can be generated based on the above-mentioned virtual transaction events and user values. The above-mentioned training samples can be used to retrain the pre-trained promotion information delivery model. Among them, the pre-trained promotion information delivery model can be a model such as AC 3.0 (Application Campaign 3.0, the third edition of the application promotion model). The above-mentioned promotion information delivery model can couple various types of data together and perform user matching based on the above-mentioned data through an intelligent algorithm. The above-mentioned data may include but is not limited to application data, user data, etc. The above-mentioned application data may include but is not limited to labels generated by the application, such as labels corresponding to user transaction behavior events. The above-mentioned user data may include but is not limited to user tags, user portraits, user behavior data, user device information, etc. The user behavior data may include but is not limited to the user's behavior data in the target website. The above-mentioned target website may be a website for the promotion information of the application to be delivered.
[0028] In practical applications, the above user value can be used as the value of the above virtual transaction event, and the identifier and value of the virtual transaction event can be transmitted back to the maintenance end of the above promotion information delivery model through a preset interface. The maintenance end of the promotion information delivery model can obtain the user data of the target user, and thus generate training samples based on the user data, virtual transaction events and their values. Among them, the maintenance end of the promotion information delivery model can be a third-party service end. By transmitting the virtual purchase event and its value back to the maintenance end of the promotion information delivery model, the limitation of the maintenance end of the promotion information delivery model on data utilization is broken, so that the maintenance end of the promotion information delivery model can provide a more effective delivery strategy, which helps to improve the efficiency and effectiveness of the promotion information delivery.
[0029] The method provided by the above embodiment of the present application obtains the shallow behavior data of the target user in the target application, and the shallow behavior data is the behavior data generated within a preset time after the target user downloads the target application and the probability of the behavior occurring is greater than the first threshold; then the shallow behavior data is input into the pre-trained user value prediction model to obtain the user value of the target user, and the user value is the transaction value of the target user in the target application within a preset number of days; then, when the user value is greater than the second threshold, a virtual transaction event is constructed, so as to generate sample data for retraining the pre-trained promotion information delivery model based on the virtual transaction event and the value. On the one hand, this implementation method simulates the user's transaction behavior by constructing a virtual transaction event and predicts its value, expands the transaction event and enriches the value distribution, solves the problem of sparse user transaction events and single value, and increases the sample size of the training promotion information delivery model, thereby improving the accuracy of the promotion information delivery model. On the other hand, this implementation method uses shallow behavior data as the data basis for user value prediction, and can transmit sample data back at an earlier time, so that the promotion information delivery model can learn and adapt to changes in user behavior more quickly, and improves the effectiveness of the promotion information delivery model. By using the promotion information delivery model to deliver promotion information, the accuracy of the promotion information delivery can be improved.
[0030] In some optional embodiments, the target application may include an audio and video application. The shallow behavior data includes at least one of the following: playback behavior data, login behavior data, and interactive behavior data. The playback behavior data may be generated when a user plays audio or video in an audio and video application. The login behavior data may be generated when a user logs into an audio and video application. The interactive behavior data may be generated when a user makes a comment or sends a bullet screen in an audio and video application.
[0031] Since the above-mentioned shallow behavior data is the behavior data generated within a preset time after the target user downloads the target application and the probability of the behavior occurring is greater than the first threshold, the shallow behavior data is used as the data basis for user value prediction, and the sample data can be fed back at an earlier time, so that the promotion information delivery model can learn and adapt to changes in user behavior more quickly, thereby optimizing the promotion information delivery strategy in real time, improving the effectiveness of promotion information delivery and the accuracy of promotion information delivery.
[0032] In some optional embodiments, the user value prediction model in the above step 102 can be pre-trained by the following steps:
[0033] The first step is to obtain historical shallow behavior data and first historical transaction data of multiple first historical users in the above target application.
[0034] The first historical user may be an old user who has downloaded the target application for more than a specified period of time.
[0035] The above-mentioned historical shallow behavior data is behavior data generated within a preset time period after the first historical user downloads the target application and the probability of the behavior occurring is greater than a first threshold.
[0036] The first historical transaction data may include but are not limited to the transaction value of the first historical user in the target application within a preset number of days (e.g., 7 days) after the first historical user downloads the target application. For example, the target application is an audio and video application, and the transaction value may include but is not limited to: monthly membership transaction value, quarterly membership transaction value, annual membership transaction value, etc. If a first historical user activates an annual membership within 7 days of downloading the target user, the corresponding first historical transaction data may include the annual membership transaction value.
[0037] The second step is to determine the real value of each first historical user based on the first historical transaction data of each first historical user. Here, for each first historical user, the transaction value in the first historical transaction data can be used as the real value of the first historical user. That is, the real value is the transaction value of the first historical user in the target application within a preset number of days after downloading the target application.
[0038] The third step is to train the user value prediction model based on the historical shallow behavior data and true value of each first historical user.
[0039] Here, the historical shallow behavior data and true value of each first historical user can be used as a training sample. Based on the historical shallow behavior data and true values of multiple first historical users, multiple samples can be obtained. Multiple samples can constitute a sample set. The above-mentioned execution subject can train the above-mentioned user value prediction model through the sample set. Specifically, the training process can be seen as follows:
[0040] First, samples are extracted from the sample set. The method and number of samples to be extracted are not limited in this application. For example, samples may be extracted randomly or in a specified order from the sample set.
[0041] Then, the extracted sample is input into the basic model to obtain the prediction result output by the basic model, which is the prediction value of the first historical user corresponding to the sample.
[0042] Afterwards, the loss value of the above basic model can be determined based on the predicted results and the true value in the input sample. The above loss value is the value of the loss function, which is a non-negative real-valued function that can be used to characterize the difference between the predicted value and the true value. Generally speaking, the smaller the loss value, the better the robustness of the model. The loss function can be set according to actual needs.
[0043] Afterwards, the loss value can be used to update the parameters of the basic model. Thus, each time a sample is input, the parameters of the basic model can be updated based on the loss value corresponding to the sample until the training is completed.
[0044] In practice, whether training is completed can be determined in a variety of ways. For example, when the difference between the predicted value and the true value of the basic output is less than a preset value (for example, less than 0.5), the training can be determined to be completed. For another example, if the number of training times of the basic model is equal to the preset number of times, the training can be determined to be completed. For another example, when the loss value of the basic model converges, the training can be determined to be completed. Here, if the basic model training is completed, the trained basic model can be determined as the user value prediction model.
[0045] Since the user value prediction model is trained based on the historical shallow behavior data and real value of multiple first historical users in the above-mentioned target application, this prediction method improves the accuracy of the user value prediction results, that is, the accuracy of the value of the virtual transaction event, so that the promotion information delivery model can be trained based on more real and reliable data, further improving the effect of promotion information delivery. In addition, since the behavioral data input into the user value prediction model is shallow behavioral data, the immediacy of user value acquisition can be improved. On this basis, the sample data can be fed back at an earlier time, so that the promotion information delivery model can learn and adapt to changes in user behavior more quickly, thereby optimizing the promotion information delivery strategy in real time, and improving the effectiveness and accuracy of promotion information delivery.
[0046] In some optional embodiments, in the above step 104, the process of generating training samples can be performed according to the following steps: First, the user data of the above target user is obtained. The user data may include but is not limited to user tags, user portraits, user behavior data, user device information, etc. Then, the above user value is used as the value of the above virtual transaction event, and based on the above virtual transaction event, the value of the above virtual transaction event and the user data of the above target user, sample data for retraining the promotion information delivery model is generated. In practice, the above user data, the tags of the virtual transaction event, and the value of the virtual transaction event can be summarized in the target format to obtain sample data. By constructing a virtual transaction event to simulate the user's transaction behavior and predicting its value, the problem of sparse user transaction events and single value is solved, the sample size of the training promotion information delivery model is increased, and the accuracy of the promotion information delivery model is improved. The promotion information delivery model can be used to deliver promotion information, which can improve the accuracy of the promotion information delivery. In addition, using shallow behavioral data as the data basis for user value prediction can feed back sample data at an earlier time, allowing the promotional information delivery model to learn and adapt to changes in user behavior more quickly, thereby optimizing the promotional information delivery strategy in real time, improving the effectiveness of promotional information delivery, and further improving the accuracy of promotional information delivery.
[0047] In some optional embodiments, the above-mentioned pre-trained promotion information delivery model can be trained through the following steps:
[0048] The first step is to obtain user data of multiple second historical users and second historical transaction data of multiple second historical users in the target application.
[0049] The plurality of second historical users may be users who have performed transaction behaviors in the target application. The plurality of second historical users may be the same as, different from, or partially the same as the plurality of first historical users, which is not specifically limited here.
[0050] The second historical transaction data may include, but are not limited to, the transaction value of the second historical user in the target application. For example, the target application is an audio and video application, and the transaction value may include, but is not limited to: monthly membership transaction value, quarterly membership transaction value, annual membership transaction value, etc. If a second historical user activates an annual membership during the use of the target application, the corresponding second historical transaction data may include the annual membership transaction value.
[0051] The second step is to determine the value of the historical transaction event of each second historical user based on the second historical transaction data of each second historical user. Here, for each second historical user, since the user has performed transaction behavior in the target application, it can be determined that the user has a historical transaction event. The transaction value in the second historical transaction data can be used as the value of the historical transaction event.
[0052] The third step is to generate historical training samples based on the user data of each second historical user, historical transaction events and the value of historical transaction events. This step is similar to the above step of generating training samples and will not be repeated here.
[0053] The fourth step is to train the promotion information delivery model through historical training samples. Here, the historical training samples can be input into the promotion information delivery model to be trained, so that the promotion information delivery model to be trained learns the characteristics of the data in the above historical training samples, and then matches potential users based on the characteristics. Potential users are users whose similarity with the user characteristics of the target user is greater than the third threshold.
[0054] In some optional embodiments, before executing step 101, the following steps may also be executed: delivering promotional information of the target application to the user through the promotional information delivery model; when it is detected that the user downloads the target application by clicking on the promotional information, determining the user as a target user.
[0055] Since the target user downloads the target application by clicking on the promotional information delivered to him by the promotional information delivery model, when the target user is a high-value user, that is, when the user value of the target user is greater than the second threshold, using the user data of the target user to retrain the promotional information delivery model can help the promotional information delivery model find potential users whose characteristics are closer to those of the target user. Since the potential user is closer to the characteristics of the target user, it is easier for him to adopt the same method, that is, to download the target application by clicking on the promotional information. As a result, the promotional information delivery model can deliver more accurately and reduce interference to users. Furthermore, since the delivery of promotional information is more accurate, the profit margin of the promotional information delivery is improved, thereby improving the overall operational efficiency.
[0056] In some optional embodiments, after executing step 104, the following steps may be further performed: identifying potential users of the target application through a retrained model obtained by retraining the above-mentioned promotion information delivery model, and delivering promotion information of the target application to the potential users. The potential users are users whose user characteristics are similar to those of the target users and are greater than a third threshold.
[0057] For example, after the virtual transaction events and user values of 10,000 marked high-value users are transmitted back to the maintenance end of the promotion information delivery model, the maintenance end of the promotion information delivery model can obtain the user data of these high-value users, and form training samples with the user data, labels of virtual transaction events, and values, and retrain the promotion information delivery model based on the training samples to obtain a retrained model. Through the retraining model, potential users with user characteristics similar to those of these high-value users can be identified, so as to deliver the promotion information of the above-mentioned target application to these potential users. In actual applications, the retraining model can extract the user characteristics of high-value users including the target user and generate a user portrait. And through the user portrait, the similarity calculation is performed with the user portraits of other users to obtain a similarity calculation result. Users among other users whose similarity calculation results are greater than the third threshold are determined as potential users, so as to push the promotion information of the target application to the potential users.
[0058] Since the target user downloads the target app by clicking on the promotional information delivered to him by the delivery model, when the user is a high-value user, using the user's data and value to retrain the delivery model can help the model find closer users. Closer users are more likely to download the target app by clicking on the promotional information, thus enabling more accurate delivery and reducing interference with users. Since the delivery is more accurate, the profit margin of each delivery is increased, thereby improving overall operational efficiency.
[0059] For some examples, see Figure 2 , which shows a schematic diagram of the interactive process of the sample generation method. First, the server B that maintains the promotion information delivery model delivers the promotion information of the target application to the target user. The target user clicks on the promotion information and downloads the target application. Next, the server A that maintains the user value prediction model obtains the shallow behavior data of the target user in the target application, inputs the shallow behavior data into the user value prediction model, obtains the user value of the target user, and constructs a virtual transaction event when the user value is greater than the second threshold, and generates a training sample based on the virtual transaction event and the user value. Next, the server A sends the training sample to the server B through the interface. The server B retrains the pre-trained promotion information delivery model through the training sample to obtain the retrained model, and identifies the potential users of the target application through the retrained model, and delivers the promotion information of the target application to the potential users.
[0060] The target application may be an overseas user. The target application may be a video application. Since overseas promotion information delivery channels are generally not as open as domestic platforms and do not provide RTA (real-time API) and RTB (real-time bidding) capabilities, it is not easy to deliver promotion information directly. Usually, the AC 3.0 model provided by Google can be used to deliver promotion information. In addition, overseas promotion information delivery needs to be delivered by region, and transaction behaviors in video applications are usually membership purchase behaviors, which are deep behaviors with very low conversion rates and similar values, so the sample size is very small.
[0061] In the above example, after the sample is generated locally, it can be sent to the interface provided by the Google server that provides promotion services, so that the AC 3.0 model can be retrained through the sample on the Google server to identify potential users of the video application and deliver promotion information of the target application to the potential users.
[0062] The method provided by the above-mentioned embodiment of the present application simulates the transaction behavior of users by constructing virtual transaction events and predicting their values, thereby expanding transaction events and enriching the value distribution, solving the problem of sparse user transaction events and single value, and increasing the sample size of the training promotion information delivery model, thereby improving the accuracy of the promotion information delivery model. In addition, by using shallow behavior data as the data basis for user value prediction, the sample data can be fed back at an earlier time, so that the promotion information delivery model can learn and adapt to changes in user behavior more quickly, thereby improving the effectiveness of the promotion information delivery model. The promotion information delivery model can be used to deliver promotion information, which can improve the accuracy of the promotion information delivery.
[0063] Further references Figure 3 As an implementation of the methods shown in the above figures, the present application provides an embodiment of a sample generation device, which is similar to Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0064] like Figure 3As shown, the sample generation device 300 of this embodiment includes: an acquisition unit 301, used to acquire shallow behavior data of a target user in a target application, wherein the shallow behavior data is behavior data generated within a preset time period after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold; a prediction unit 302, used to input the shallow behavior data into a pre-trained user value prediction model to obtain the user value of the target user, wherein the user value is the transaction value of the target user in the target application within a preset number of days; a generation unit 303, used to construct a virtual transaction event when the user value is greater than a second threshold, and generate a training sample based on the virtual transaction event and the user value, wherein the training sample is used to retrain the pre-trained promotion information delivery model.
[0065] In some optional implementations of this embodiment, the device also includes: a sending unit, used to deliver promotional information of the target application to the user through the promotional information delivery model; and a determination unit, used to determine the user as a target user when it is detected that the user downloads the target application by clicking on the promotional information.
[0066] In some optional implementations of this embodiment, the device also includes: an identification unit, used to identify potential users of the target application through a retrained model obtained by retraining the promotion information delivery model, and deliver promotional information of the target application to the potential users, wherein the potential users are users whose similarity in user characteristics with the target users is greater than a third threshold.
[0067] In some optional implementations of the present embodiment, the user value prediction model is pre-trained through the following steps: obtaining historical shallow behavior data and first historical transaction data of multiple first historical users in the target application; determining the true value of each first historical user based on the first historical transaction data of each first historical user, the true value being the transaction value of the first historical user in the target application within a preset number of days after downloading the target application; training the user value prediction model based on the historical shallow behavior data and true value of each first historical user.
[0068] In some optional implementations of the present embodiment, the promotion information delivery model is pre-trained through the following steps: obtaining user data of multiple second historical users and second historical transaction data of the multiple second historical users in the target application; determining the value of the historical transaction event of each second historical user based on the second historical transaction data of each second historical user, the value of the historical transaction event being the transaction value of the second historical user in the target application; generating historical training samples based on the user data of each second historical user, historical transaction events and the value of historical transaction events; and training the promotion information delivery model through the historical training samples.
[0069] In some optional implementations of this embodiment, the generation unit 303 is further used to: obtain user data of the target user; use the user value as the value of the virtual transaction event, and generate sample data for retraining the promotion information delivery model based on the virtual transaction event, the value of the virtual transaction event and the user data of the target user.
[0070] In some optional implementations of this embodiment, the target application includes an audio and video application; the shallow behavior data includes at least one of the following: playback behavior data, login behavior data, and interactive behavior data.
[0071] The device provided by the above embodiment of the present application obtains the shallow behavior data of the target user in the target application, and the shallow behavior data is the behavior data generated within a preset time after the target user downloads the target application and the probability of the behavior occurring is greater than the first threshold; then the shallow behavior data is input into the pre-trained user value prediction model to obtain the user value of the target user, and the user value is the transaction value of the target user in the target application within a preset number of days; then, when the user value is greater than the second threshold, a virtual transaction event is constructed, so as to generate sample data for retraining the pre-trained promotion information delivery model based on the virtual transaction event and the value. On the one hand, this implementation simulates the user's transaction behavior by constructing a virtual transaction event and predicts its value, expands the transaction event and enriches the value distribution, solves the problem of sparse user transaction events and single value, and increases the sample size of the training promotion information delivery model, thereby improving the accuracy of the promotion information delivery model. On the other hand, this implementation uses shallow behavior data as the data basis for user value prediction, and can transmit sample data back at an earlier time, so that the promotion information delivery model can learn and adapt to changes in user behavior more quickly, and improves the effectiveness of the promotion information delivery model. By using the promotion information delivery model to deliver promotion information, the accuracy of the promotion information delivery can be improved.
[0072] Reference below Figure 4, which shows a schematic structural diagram of an electronic device for implementing some embodiments of the present application. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.
[0073] like Figure 4 As shown, the electronic device 400 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage device 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 400 are also stored. The processing device 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.
[0074] Typically, the following devices may be connected to the I / O interface 405: an input device 406 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 407 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 408 including, for example, a disk, a hard disk, etc.; and a communication device 409. The communication device 409 may allow the electronic device 400 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 4 The electronic device 400 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 4 Each block shown in the figure may represent one device, or may represent multiple devices as required.
[0075] In particular, according to some embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present application 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 the network through the communication device 409, or installed from the storage device 408, or installed from the ROM 402. When the computer program is executed by the processing device 401, the above-mentioned functions defined in the method of some embodiments of the present application are executed.
[0076] It should be noted that the computer-readable medium described in some embodiments of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above 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 application, a 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 application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. 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.
[0077] 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.
[0078] The computer-readable medium may be included in the electronic device; or it may exist independently without being installed in 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 shallow behavior data of the target user in the target application, and the shallow behavior data is the behavior data generated within a preset time after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold; inputs the shallow behavior data into a pre-trained user value prediction model to obtain the user value of the target user, and the user value is the transaction value of the target user in the target application within a preset number of days; when the user value is greater than the second threshold, constructs a virtual transaction event, and generates training samples based on the virtual transaction event and the value, and the sample data is used to retrain the pre-trained promotion information delivery model.
[0079] Computer program code for performing the operations of some embodiments of the present application may be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++; and also conventional procedural programming languages such as "C" language 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 the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, or may be connected to an external computer (e.g., through the Internet using an Internet service provider), including a local area network (LAN) or a wide area network (WAN).
[0080] 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 application. 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 alternative implementations, 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.
[0081] The units described in some embodiments of the present application may be implemented by software or hardware. The units described may also be set in a processor, for example, it may be described as: a processor includes a first determination unit, a second determination unit, a selection unit, and a third determination unit. The names of these units do not, in some cases, constitute limitations on the units themselves.
[0082] 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.
[0083] The above description is only some preferred embodiments of the present application 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 application is not limited to the technical solutions formed by the specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with (but not limited to) the technical features with similar functions disclosed in the embodiments of the present application to form a technical solution.
Claims
1. A sample generation method, characterized in that: The method comprises: Obtaining shallow behavior data of a target user in a target application, wherein the shallow behavior data is behavior data generated within a preset time period after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold; Input the shallow behavior data into a pre-trained user value prediction model to obtain the user value of the target user, where the user value is the transaction value of the target user in the target application within a preset number of days; When the user value is greater than a second threshold, a virtual transaction event is constructed, and a training sample is generated based on the virtual transaction event and the user value. The training sample is used to retrain a pre-trained promotion information delivery model.
2. The method according to claim 1, characterized in that Before obtaining the shallow behavior data of the target user in the target application, the method further includes: Delivering promotional information of the target application to users through the promotional information delivery model; When it is detected that the user downloads the target application by clicking on the promotion information, the user is determined as a target user.
3. The method according to claim 1, characterized in that After generating the training sample, the method further includes: The retrained model obtained by retraining the promotion information delivery model is used to identify potential users of the target application, and the promotion information of the target application is delivered to the potential users. The potential users are users whose user characteristics are similar to those of the target users and are greater than a third threshold.
4. The method according to claim 1, characterized in that: The user value prediction model is pre-trained through the following steps: Acquire historical shallow behavior data and first historical transaction data of a plurality of first historical users in the target application; Determine the real value of each first historical user based on the first historical transaction data of each first historical user, where the real value is the transaction value of the first historical user in the target application within the preset number of days after the first historical user downloads the target application; The user value prediction model is trained based on the historical shallow behavior data and true value of each first historical user.
5. The method according to claim 1, characterized in that The promotion information delivery model is pre-trained through the following steps: Acquire user data of a plurality of second historical users and second historical transaction data of the plurality of second historical users in the target application; Determine the value of a historical transaction event of each second historical user based on the second historical transaction data of each second historical user, where the value of the historical transaction event is the transaction value of the second historical user in the target application; Generate historical training samples based on the user data of each second historical user, the historical transaction events, and the value of the historical transaction events; The promotion information delivery model is trained through the historical training samples.
6. The method according to claim 1, characterized in that The generating, based on the virtual transaction event and the user value, sample data for retraining the promotion information delivery model comprises: Acquire user data of the target user; The user value is used as the value of the virtual transaction event, and sample data for retraining a promotion information delivery model is generated based on the virtual transaction event, the value of the virtual transaction event and the user data of the target user.
7. The method according to any one of claims 1 to 6, characterized in that: The target application includes an audio and video application; the shallow behavior data includes at least one of the following: playback behavior data, login behavior data, and interactive behavior data.
8. A sample generating device, characterized in that: The device comprises: An acquisition unit, configured to acquire shallow behavior data of a target user in a target application, wherein the shallow behavior data is behavior data generated within a preset time period after the target user downloads the target application and the probability of the behavior occurring is greater than a first threshold; A prediction unit, configured to input the shallow behavior data into a pre-trained user value prediction model to obtain a user value of the target user, wherein the user value is a transaction value of the target user in the target application within a preset number of days; A generating unit is used to construct a virtual transaction event when the user value is greater than a second threshold value, and to generate a training sample based on the virtual transaction event and the user value, wherein the training sample is used to retrain a pre-trained promotion information delivery model.
9. An electronic device, characterized in that: include: 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.
10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Data return method and device based on double-model joint estimation, equipment and medium
CN122534280A
