A method, device and storage medium for information display

Abstract

This application discloses a method, apparatus, and storage medium for information display, specifically including: receiving a first information display request message; retrieving matching first information to be displayed from a database based on a first user ID; obtaining a first probability and a second probability of user activation using a first machine learning model; using the difference between the first probability and the second probability as the increment value of the user activation probability; calculating the first value information corresponding to the first information to be displayed based on the increment value of the user activation probability and returning a first information display response message. By applying the embodiments of this application, since the increment value of the user activation probability is used to calculate the first value information corresponding to the first information to be displayed, the first value information can be accurately estimated, maximizing the role of the first value information and reducing the cost of pushing the first information to be displayed.

Description

Technical Field

[0001] This application relates to the field of Internet technology, and in particular to a method for displaying information, an apparatus for displaying information, a computer-readable storage medium, and a computer program product. Background Technology

[0002] More and more businesses are now pushing or displaying business information to users online via mobile phones or tablets. Users browse various types of information online according to their daily habits, such as news and other media platforms. To better activate users and keep them active on the platform, businesses want to appropriately push information to users while they are browsing online, thereby activating users through information display. However, information display usually incurs costs, and inappropriate information push display will result in significant costs. Summary of the Invention

[0003] In view of the above-mentioned prior art, this invention discloses an information display method that can overcome the defects of inappropriately pushed and displayed information, achieve the purpose of accurately estimating and maximizing the value of the information to be displayed, and reduce costs.

[0004] In view of this, embodiments of this application propose a method for displaying information, the method comprising:

[0005] A first information display request message is received, the first information display request message including a first user ID;

[0006] Based on the first user ID, the first information to be displayed is retrieved from the database to match the first user ID. The first information to be displayed is the information to be displayed to the first user corresponding to the first user ID.

[0007] The first user feature recorded in advance is used as the first feature, and the feature that the first user obtains the first information to be displayed is used as the second feature. The first feature and the second feature are used as parameters and input into the trained first machine learning model to obtain the first probability of the user being activated. The first machine learning model is a model used to calculate the probability of the user being activated. The user being activated means that the user is in an active state within a preset time.

[0008] The first user feature, which was recorded in advance, is used again as the first feature, and the feature that the first user has not obtained the first information to be displayed is used as the third feature. The first feature and the third feature are used as parameters and input into the trained first machine learning model to obtain the second probability that the user is activated.

[0009] The difference between the first probability and the second probability is used as the increment of the user's activation probability.

[0010] The first value information corresponding to the first information to be displayed is calculated based on the incremental value of the user activation probability.

[0011] The system returns a first information display reply message, which includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information has been evaluated and approved.

[0012] Furthermore,

[0013] Before the step of receiving the first information display request message, the method further includes: training the first machine learning model;

[0014] The steps for training the first machine learning model include:

[0015] The obtained user set is randomly divided into experimental buckets and control buckets according to the user device ID. The user set in the experimental bucket is the second user set used for training, and the user set in the control bucket is the third user set used for testing.

[0016] For each user in the second user set, a second information to be displayed filtering process, a second value information determination process, and a second value information evaluation process are executed respectively. The second information to be displayed filtering process refers to the process of filtering out the second information to be displayed that matches the user from the database. The second value information determination process refers to the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process refers to the process of participating the second value information in the evaluation, and the second information to be displayed is obtained after the second value information is evaluated and passed.

[0017] For each user in the third user set, a third information to be displayed filtering process and a third value information determination process are executed respectively. The third information to be displayed filtering process refers to the process of filtering out the third information to be displayed that matches the user from the database. The third value information determination process refers to the process of determining the third value information corresponding to the third information to be displayed.

[0018] In the second user set, the users who obtained the second information to be displayed will form a fourth user set;

[0019] The first machine learning model after initialization is trained using the fourth user set and the third user set to obtain the trained first machine learning model.

[0020] Furthermore,

[0021] The step of training the initialized first machine learning model using the fourth user set and the third user set to obtain the trained first machine learning model includes:

[0022] In the second user set, users who have not obtained the second information to be displayed form the fifth user set;

[0023] Users in the fourth user set are used as positive samples to train the second machine learning model, and users in the fifth user set are used as negative samples to train the second machine learning model. The second machine learning model is trained to obtain the trained second machine learning model, which is used to calculate the probability of obtaining the second information to be displayed.

[0024] The probability of each user in the fourth user set obtaining the second information to be displayed is calculated using the second machine learning model, and the proportion of users in each probability interval of the fourth user set is determined according to the set probability interval.

[0025] The probability of each user in the third user set obtaining the third information to be displayed is calculated using the second machine learning model, and users are extracted from the third user set according to the proportion of users in each probability interval of the fourth user set. The extracted users form the sixth user set.

[0026] The fourth user set and the sixth user set are used as training samples and input into the initialized first machine learning model for training to obtain the trained first machine learning model. The activated users in the fourth user set and the sixth user set are used as positive samples for training the first machine learning model, and the inactive users in the fourth user set and the sixth user set are used as negative samples for training the first machine learning model.

[0027] Furthermore,

[0028] The step of retrieving the first matching information to be displayed from the database based on the first user ID includes:

[0029] The first user characteristics are retrieved from the database based on the first user ID. The first user characteristics include basic user characteristics and user in-site behavior characteristics. The basic user characteristics include the user's download channel and the time since registration. The user's in-site behavior characteristics include the number of times the user logs in, the number of times the user searches, and the number of times the user performs pre-transaction processes. The user's download channel refers to the channel through which the user downloads the application. The application is the target application associated with the first information to be displayed. The time since registration is the time since the user registered in the application. The number of times the user logs in is the number of times the user logs in to the application within a preset time period. The number of times the user searches is the number of times the user performs search behavior within a preset time period. The number of times the user performs pre-transaction processes is the number of times the user performs pre-transaction process behavior within a preset time period.

[0030] Based on the first user characteristic, candidate display information is filtered from the database;

[0031] The information is sorted according to a pre-set sorting strategy, and the candidate information ranked first is selected as the first information to be displayed.

[0032] In view of the above-mentioned prior art, embodiments of the present invention also disclose an information display device, which can overcome the defects of inappropriately pushed and displayed information, achieve the purpose of accurately estimating and maximizing the value of the information to be displayed, and reducing costs.

[0033] In view of this, embodiments of this application propose an information display device, which includes: a receiving module, a filtering module, a value calculation module, and a sending module;

[0034] The receiving module is configured to receive a first information display request message, wherein the first information display request message includes a first user ID;

[0035] The filtering module is used to obtain matching first information to be displayed from the database based on the first user ID. The first information to be displayed is information to be displayed to the first user corresponding to the first user ID.

[0036] The value calculation module takes a pre-recorded first user feature as a first feature, and the feature that the first user has obtained the first information to be displayed as a second feature. The first and second features are input as parameters into a trained first machine learning model to obtain a first probability of user activation. The first machine learning model is used to calculate the user activation probability, where user activation indicates that the user is active within a preset time period. The module then takes the pre-recorded first user feature as the first feature again, and the feature that the first user has not obtained the first information to be displayed as a third feature. The first and third features are input as parameters into the trained first machine learning model to obtain a second probability of user activation. The difference between the first and second probabilities is used as the increment of the user activation probability. Based on the increment of the user activation probability, the module calculates the corresponding first value information of the first information to be displayed.

[0037] The sending module is used to return a first information display reply message, which includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information has been evaluated and approved.

[0038] The device further includes a first training module;

[0039] The first training module is used to train the first machine learning model, including randomly dividing the obtained user set into experimental buckets and control buckets according to user device IDs. The user set in the experimental bucket is the second user set participating in training, and the user set in the control bucket is the third user set used for testing. For each user in the second user set, a second information to be displayed filtering process, a second value information determination process, and a second value information evaluation process are executed respectively. The second information to be displayed filtering process refers to the process of filtering out the second information to be displayed that matches the user from the database. The second value information determination process refers to the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process refers to the process of evaluating the second value information. The evaluation process involves value information, and the second information to be displayed is obtained after the second value information is evaluated and approved. For each user in the third user set, a third information to be displayed filtering process and a third value information determination process are executed. The third information to be displayed filtering process refers to the process of filtering out the third information to be displayed that matches the user from the database, and the third value information determination process refers to the process of determining the third value information corresponding to the third information to be displayed. In the second user set, the users who obtain the second information to be displayed form a fourth user set. The first machine learning model after initialization is trained using the fourth user set and the third user set to obtain the trained first machine learning model.

[0040] Furthermore,

[0041] When the first training module trains the initialized first machine learning model using the fourth user set and the third user set to obtain the trained first machine learning model, the process includes: in the second user set, users who have not obtained the second information to be displayed form a fifth user set; using users in the fourth user set as positive samples for training the second machine learning model, and using users in the fifth user set as negative samples for training the second machine learning model, the second machine learning model is trained to obtain a trained second machine learning model, which is used to calculate the probability of obtaining the second information to be displayed; using the second machine learning model to calculate the probability of each user in the fourth user set obtaining the information to be displayed, and according to a given... The probability intervals are set to determine the proportion of users in each probability interval of the fourth user set; the probability of each user in the third user set obtaining the information to be displayed is calculated using the second machine learning model, and users are extracted from the third user set according to the proportion of users in each probability interval of the fourth user set to form a sixth user set; the fourth user set and the sixth user set are used as training samples and input into the initialized first machine learning model for training to obtain the trained first machine learning model, and the activated users in the fourth user set and the sixth user set are used as positive samples for training the first machine learning model, and the inactive users in the fourth user set and the sixth user set are used as negative samples for training the first machine learning model.

[0042] Furthermore,

[0043] When the filtering module retrieves the first matching information to be displayed from the database based on the first user ID, it includes: retrieving the corresponding first user feature from the database based on the first user ID; the first user feature includes basic user features and user in-site behavior features; the basic user features include the user's download channel and registration date; the user's in-site behavior features include the number of logins, the number of searches, and the number of pre-transaction process executions; the user's download channel refers to the channel through which the user downloaded the application; the application is the target application associated with the first information to be displayed; the registration date is the time since the user registered in the application; the number of logins is the number of times the user logged into the application within a preset time period; the number of searches is the number of times the user performed search behavior within the preset time period; and the number of pre-transaction process executions is the number of times the user performed pre-transaction process behavior within the preset time period. The module then filters candidate information to be displayed from the database based on the first user feature; sorts the candidate information according to a pre-set sorting strategy, and selects the candidate information ranked first as the first information to be displayed.

[0044] This application also provides a computer-readable storage medium storing computer instructions thereon, characterized in that the instructions, when executed by a processor, can realize the above-described method of information display.

[0045] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the information display method as described above.

[0046] In summary, the embodiments of this application do not directly provide the value corresponding to the first information to be displayed, but calculate the incremental value of the user activation probability, and use the incremental value of the user activation probability to calculate the first value information corresponding to the first information to be displayed. This can accurately predict the first value information, maximize the role of the first value information, and reduce the cost of pushing the first information to be displayed. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of an embodiment of the method for displaying information in this application.

[0049] Figure 2 This is a flowchart of the method for training the first machine learning model in Embodiment 2 of this application.

[0050] Figure 3 This is a flowchart of the method for training the first machine learning model in Embodiment 3 of this application.

[0051] Figure 4 This is an example diagram illustrating the process of determining training samples in Embodiment 4 of this application.

[0052] Figure 5 This is a schematic diagram of the structure of an embodiment of the device for displaying information in this application.

[0053] Figure 6 This is a schematic diagram of the structure of Embodiment 2 of the device for displaying information in this application. Detailed Implementation

[0054] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0055] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0056] The technical solution of the present invention will be described in detail below with reference to specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0057] This application does not directly provide value information for the information to be displayed. Instead, it considers the user activation probability and uses the increment of the user activation probability as a factor in calculating the value information, thus accurately predicting the value of the information to be displayed. The value information in this application can be understood as the cost of information display. Accurately predicting the value information can fully leverage the role of the information to be displayed and minimize its cost.

[0058] Figure 1 This is a flowchart of a first embodiment of the method for displaying information according to this application. Step 101 is the step of receiving a first information display request message; steps 102 to 106 are steps of estimating the information to be displayed and calculating its value; and step 107 is the step of returning a first information display reply message. Figure 1 As shown, the method specifically includes:

[0059] Step 101: Receive a first information display request message, which includes a first user ID.

[0060] In practical applications, users typically use devices such as mobile phones or tablets to browse various information online, such as news and other media platforms. To push information to users, it can be displayed opportunely during their browsing process. Assume the party owning the information is the company, and the party pushing the information is the media outlet. The media outlet will send an information display request message to the company, indicating that it can push the company's information to users browsing on the media platform at this moment. To distinguish it from subsequent naming conventions, this information display request message is referred to as the "first information display request message," which can be understood as the company receiving the first information display request message from the media outlet.

[0061] Additionally, regardless of whether a user is browsing a media platform or logging into a company's application, a user ID can be recorded to identify the user. To distinguish it from other names used later, it will be referred to here as the "first user ID".

[0062] Step 102: Retrieve the first matching information to be displayed from the database based on the first user ID. The first information to be displayed is the information to be displayed to the first user corresponding to the first user ID.

[0063] This step involves filtering the information to be displayed to users. Different users can be matched with different information to meet their personalized needs. To distinguish it from other names used later, the information to be displayed will be referred to as "First Information to be Displayed," and the user will be referred to as "First User."

[0064] Steps 103 and 104 below involve inputting different features into the same machine learning model to obtain two different user activation probabilities. Their purpose is to calculate the probability increment between the two probabilities in step 105. To distinguish it from other names used later, the machine learning model in this embodiment is referred to as the "first machine learning model," which is used to calculate the activation probability. The input to the first machine learning model is divided into two categories: user features and whether or not the first piece of information to be displayed is obtained. The output is the probability of user activation. To distinguish it from other names, the user features here are called "first user features," which are parameters used to describe user characteristics, such as basic user characteristics and user in-site behavior characteristics.

[0065] Step 103: Take the first user feature recorded in advance as the first feature, take the feature of the first user obtaining the first information to be displayed as the second feature, and input the first feature and the second feature as parameters into the trained first machine learning model to obtain the first probability of the user being activated. The first machine learning model is a model used to calculate the probability of the user being activated. The user being activated means that the user is in an active state within a preset time.

[0066] This step is the first time that features are input as parameters into the first machine learning model. The input parameters include the first user feature and the feature indicating whether the first user has obtained the first piece of information to be displayed. Assuming the first user feature is an N-dimensional feature (first feature), and whether the first user has obtained the first piece of information to be displayed is a 1-dimensional feature (second feature, e.g., 1 represents obtaining), then the first and second features should be N+1 dimensional features. These N+1 dimensional features are then input as parameters into the first machine learning model to obtain the first probability of the output.

[0067] Step 104: Again, use the first user feature recorded in advance as the first feature, and the feature that the first user has not obtained the first information to be displayed as the third feature. Input the first feature and the third feature as parameters into the trained first machine learning model to obtain the second probability that the user is activated.

[0068] This step is the second process of inputting features as parameters into the first machine learning model. The input parameters include the first user feature and the feature indicating that the first user did not receive the first piece of information to be displayed. Assuming the first user feature is an N-dimensional feature (first feature), and whether the first user received the first piece of information to be displayed is a 1-dimensional feature (second feature, e.g., 0 indicates not received), then the first and second features should be N+1 dimensional features. These N+1 dimensional features are then input as parameters into the first machine learning model to obtain the second probability of the output.

[0069] The difference between the input parameters in steps 103 and 104 lies in the feature of whether the first user has obtained the first information to be displayed. The first input parameter assumes that the first user has obtained the first information to be displayed, while the second input parameter assumes that the first user has not obtained the first information to be displayed. It should be noted that steps 103 and 104 merely input the assumption that the first user has obtained or has not obtained the first information to be displayed as features into the first machine learning model for calculation. Whether the first user actually obtains the first information to be displayed in the future still needs to be determined based on the subsequent evaluation results.

[0070] Step 105: Use the difference between the first probability and the second probability as the increment of the user's activation probability.

[0071] The first probability calculated in step 103 above represents the probability that the first user is activated when the first information to be displayed is obtained, and the first probability calculated in step 104 above represents the probability that the first user is activated when the first information to be displayed is not obtained. The difference is the increment of the user's activation probability.

[0072] Step 106: Calculate the first value information corresponding to the first information to be displayed based on the incremental value of the user activation probability.

[0073] The embodiments of this application calculate the first value information based on the incremental value of the user activation probability, with the aim of maximizing the role of the first value information and controlling the cost of pushing the first information to be displayed.

[0074] If the increment value of the user activation probability is large, it means that the factors that activate the first user are closely related to the first information to be displayed. This allows for the calculation of higher first-value information, which is easier to evaluate and thus enables the first user to successfully obtain the first information to be displayed.

[0075] Conversely, if the incremental value of the user activation probability is small, it means that the factors that activate the first user are not closely related to the first information to be displayed, and a lower first value information can be calculated, thereby reliably controlling costs.

[0076] Step 107: Return to the first information display reply message. The first information display reply message includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information has been evaluated and approved.

[0077] In practical applications, assuming the company returns the first set of information to be displayed and the first set of valuable information to the media outlet, the media outlet, after evaluation, will push the first set of information to be displayed to the first user. The first user can then access this information while browsing the media outlet's platform using a mobile phone or tablet. Subsequently, if the first user logs in to the company by clicking on the first set of information to be displayed, or logs in manually, that user will be active within the company's platform.

[0078] Furthermore, step 102 above, which involves retrieving the first matching information to be displayed from the database based on the first user ID, can be implemented as follows: Retrieve the corresponding first user characteristics from the database based on the first user ID; filter candidate display information from the database based on the first user characteristics; sort according to a pre-set sorting strategy, and use the candidate display information ranked first as the first information to be displayed. In practical applications, enterprises typically store user IDs and their corresponding user characteristics in the database. User characteristics can include basic user characteristics and user in-site behavior characteristics. Basic user characteristics include the user's download channel and registration date, while user in-site behavior characteristics include the number of logins, searches, and pre-transaction executions. The user's download channel refers to the channel through which the user downloaded the application; the application is the target application associated with the first information to be displayed; the registration date is the time elapsed since the user registered in the application; the number of logins is the number of times the user logged into the application within a preset time period; the number of searches is the number of times the user performed searches within a preset time period; and the number of pre-transaction executions is the number of times the user performed pre-transaction actions within a preset time period. Taking the real estate transaction sector as an example, the "first user" mentioned here can be considered a real estate transaction user who has registered for a real estate transaction-related application. The download channel indicates how the user downloaded the application on their phone or tablet. The registration date is the difference between the user's registration date and the current date. The number of logins within the application indicates the number of times the user has logged into the application. The number of searches within the application indicates the number of times the user has searched for property information within the application. The number of pre-transaction process executions indicates the number of pre-transaction actions performed by the user before the transaction, such as showing properties. Utilizing these user characteristics to filter the first set of information to be displayed from the database allows for personalized information to be displayed to each user, better meeting their needs.

[0079] The solution applied in this application calculates the first value information corresponding to the first information to be displayed based on the incremental value of the user activation probability, which can maximize the role of the first value information and reduce the cost of pushing the first information to be displayed.

[0080] The above-described method embodiment involves a first machine learning model, which requires further training in practical applications. This application embodiment can train the first machine learning model using the following experimental method. Figure 2 This is a flowchart of the method for training the first machine learning model in Embodiment 2 of this application. Figure 2 As shown, the method includes:

[0081] Step 201: Randomly divide the obtained user set into experimental buckets and control buckets according to the user device ID. The user set in the experimental bucket is the second user set used for training, and the user set in the control bucket is the third user set used for testing.

[0082] In practical applications, companies typically store user-related information. The user set mentioned here can be obtained from a database, i.e., a collection of users. There are many reasons why a user might be activated. To accurately pinpoint the direct causal relationship between receiving displayed information and user activation, it's necessary to exclude user characteristics other than whether or not information was received. Therefore, experimental and control buckets are set up during training. To distinguish them from other names used later, the user set assigned to the experimental bucket is called the second user set, and the user set assigned to the control bucket is called the third user set.

[0083] Step 202: For each user in the second user set, execute the second information to be displayed screening process, the second value information determination process, and the second value information evaluation process respectively. The second information to be displayed screening process means the process of filtering out the second information to be displayed that matches the user from the database. The second value information determination process means the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process means the process of participating the second value information in the evaluation, and the second information to be displayed is obtained after the second value information is evaluated.

[0084] During the second selection process, information that meets the requirements for display can be filtered from the database based on each user's characteristics. If multiple second selections are found, they can be sorted according to a preset strategy, with the highest-ranked selection being the final second selection.

[0085] During the evaluation process for the second value information, the company returns the second information to be displayed and the second value information to the media. The media evaluates the second value information; if the evaluation passes, the user will receive the second information to be displayed. Conversely, if the evaluation fails, the user will not receive the second information to be displayed. Therefore, after all users in the second user set have completed the second information to be displayed screening process, the second value information determination process, and the second value information evaluation process, some users will receive the second information to be displayed, while others will not.

[0086] Step 203: For each user in the third user set, execute the third information to be displayed filtering process and the third value information determination process respectively. The third information to be displayed filtering process means the process of filtering out the third information to be displayed that matches the user from the database. The third value information determination process means the process of determining the third value information corresponding to the third information to be displayed.

[0087] This step is similar to step 202, but for ease of distinction, it is referred to here as the third information to be displayed filtering process and the third value information determination process. Similarly, during the third information to be displayed filtering process, information that meets the requirements for display can be filtered from the database based on each user's user characteristics; this is the third information to be displayed. If multiple third information items are filtered out, they can be sorted according to a preset strategy, and the one ranked first will be the final filtered third information to be displayed.

[0088] Unlike step 202, step 203 does not execute the third value information evaluation process. That is, for users in the second user set, the result of executing the second value information evaluation process includes either obtaining the feature of the second information to be displayed or not obtaining this feature. For users in the third user set, this mainly serves as a control, and the third value information evaluation process is unnecessary, thus reducing the cost of training the first machine learning model.

[0089] Step 204: In the second user set, the users who have obtained the second information to be displayed are grouped into a fourth user set.

[0090] Step 205: Train the initialized first machine learning model using the fourth user set and the third user set to obtain the trained first machine learning model.

[0091] Since some users in the second user set received the second piece of information to be displayed, it can be assumed that the activation of these users is related to receiving the second piece of information to be displayed. Therefore, they can be used as samples in the fourth user set for training the first machine learning model. Furthermore, the users in the third user set have also undergone the third process of filtering the information to be displayed and determining the third value information, which is essentially the same as the fourth user set. Therefore, they can also be used as samples for training the first machine learning model, thus obtaining a well-trained first machine learning model.

[0092] In step 203 of the above embodiment, the third value information evaluation process is not performed, which results in the user samples in the third user set not being completely homogeneous with the fourth user set. To further improve the accuracy of the first machine learning model and further improve the training samples, this application embodiment provides another detailed training method for step 203 without increasing training costs.

[0093] Steps 301 to 304 below are the process of extracting users homogeneous with the users extracted from the third user set and the fourth user set. Here, homogeneity means having the same characteristics. In Method Embodiment Two, since the users in the third user set in the control bucket did not perform the third value information evaluation process, and therefore lacked the characteristics to obtain the third information to be displayed, or did not obtain the third information to be displayed, the users in the third user set and the users in the fourth user set are not completely homogeneous. To overcome this deficiency, Method Embodiment Three uses a second machine learning model to calculate the probability of obtaining the information to be displayed and extracts homogeneous users according to the probability ratio.

[0094] Figure 3 This is a flowchart of the method for training the first machine learning model in Embodiment 3 of this application, and a specific implementation of step 205 in Embodiment 2. Figure 3 As shown, the method includes:

[0095] Step 301: In the second user set, the users who have not received the second information to be displayed are grouped into the fifth user set.

[0096] Step 302: Use users in the fourth user set as positive samples to train the second machine learning model, and use users in the fifth user set as negative samples to train the second machine learning model. Train the second machine learning model to obtain the trained second machine learning model. The second machine learning model is used to calculate the probability of obtaining the second information to be displayed.

[0097] Steps 301-302 above constitute the process of training the second machine learning model. As mentioned earlier, since the users in the second user set performed the second value information evaluation process, some users obtained the second information to be displayed, while others did not. Therefore, these users can be used as positive and negative samples to input into the second machine learning model for training. In practical applications, when inputting samples into the second machine learning model for training, it is also necessary to input the user characteristics of the sample users as parameters. The user characteristics described here are the same as the first user characteristics described in step 102 of the first embodiment of the method, and may include basic user characteristics and user in-site behavior characteristics, etc., which will not be elaborated here.

[0098] Step 303: Calculate the probability of each user in the fourth user set obtaining the second information to be displayed using the second machine learning model, and determine the proportion of users in each probability interval of the fourth user set according to the set probability interval.

[0099] This step calculates and determines the distribution across different probability intervals. Assuming the probability interval is 0.1, the number of users in the fourth user set who have the probability of receiving the second piece of information to be displayed is shown in Table 1:

[0100] [1，0.9) [0.9,0.8) [0.8,0.7) [0.7,0.6) [0.6,0.5) 0.5 or less 2000 4000 8000 4000 2000 1000

[0101] Table 1

[0102] In other words, the ratio of the number of users in each probability interval of the fourth user set is 2:4:8:4:2:1.

[0103] Step 304: Calculate the probability of each user in the third user set obtaining the third information to be displayed using the second machine learning model, and extract users from the third user set according to the proportion of users in each probability interval of the fourth user set. The extracted users form the sixth user set.

[0104] Although the users in the third user set did not participate in the evaluation process for the third value information, and they either lacked the characteristics to obtain or did not possess the characteristics to obtain the third information to be displayed, the probability of each user obtaining the third information to be displayed can be calculated using the second machine learning model, thus achieving the effect of participating in the evaluation process for the third value information. Then, users are proportionally drawn from the third user set according to the probability distribution ratio in Table 1 above. The drawn users can be considered to be homogeneous with the users in the fourth user set.

[0105] Step 305: Use the fourth user set and the sixth user set as training samples, input them into the initialized first machine learning model for training, and obtain the trained first machine learning model. Activated users in the fourth user set and the sixth user set are used as positive samples for training the first machine learning model, and inactive users in the fourth user set and the sixth user set are used as negative samples for training the first machine learning model.

[0106] Unlike step 205 in the second embodiment of the method described above, this embodiment uses the fourth user set and the sixth user set as samples to train the first machine learning model. On one hand, the users in the fourth user set are those who have obtained the second information to be displayed, while the users in the sixth user set are homogeneous users who did not obtain the third information to be displayed because they did not perform the third value information evaluation process. Therefore, the probability of a user being activated can be attributed to whether or not the value information evaluation process was performed, thus making the trained first machine learning model more reliable. Similarly, when inputting samples into the first machine learning model for training, the user characteristics of the sample users also need to be input as parameters. The user characteristics described here are the same as the first user characteristics described in step 102 of the first embodiment of the method, and may include basic user characteristics and user in-site behavior characteristics, etc., which will not be elaborated here.

[0107] In Embodiment 4 of this application, it is assumed that 1 million users are obtained from the database. These user information represents the historical processing of information display requests sent by enterprises to media outlets, and can be used as training samples. Figure 4This is an example diagram illustrating the process of determining training samples in Embodiment 4 of this application. Figure 4 As shown:

[0108] Step 401: Randomly divide the 1 million users in the obtained user set into experimental buckets and control buckets according to their user device IDs. The user set in the experimental bucket is the second user set used for training, and the user set in the control bucket is the third user set used for testing.

[0109] This step is carried out according to step 201 of method embodiment two, and it is assumed that 900,000 users are divided into experimental buckets and 100,000 users are divided into control buckets.

[0110] Step 402: For each user in the second user set, execute the second information to be displayed filtering process, the second value information determination process, and the second value information evaluation process respectively. The second information to be displayed filtering process means the process of filtering out the second information to be displayed that matches the user from the database. The second value information determination process means the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process means the process of participating the second value information in the evaluation, and the second information to be displayed is obtained after the second value information is evaluated.

[0111] This step is implemented according to step 202 of method embodiment two, and assumes that 300,000 users have obtained the second information to be displayed. It should be noted that since the second information to be displayed is filtered based on each user's user characteristics, the second information to be displayed obtained by each user is usually different.

[0112] Step 403: For each user in the third user set, execute the third information to be displayed filtering process and the third value information determination process respectively. The third information to be displayed filtering process means the process of filtering out the third information to be displayed that matches the user from the database. The third value information determination process means the process of determining the third value information corresponding to the third information to be displayed.

[0113] This step is performed according to step 203 of method embodiment two.

[0114] Step 404: In the second user set, the users who have obtained the second information to be displayed are grouped into a fourth user set.

[0115] This step is carried out according to step 204 of method embodiment two, and it is assumed that the 300,000 users who obtained the second information to be displayed constitute the fourth user set.

[0116] Step 405: In the second user set, the users who have not received the second information to be displayed are grouped into the fifth user set.

[0117] This step is carried out according to step 301 of method embodiment three, and it is assumed that the 600,000 users who have not obtained the second information to be displayed constitute the fifth user set.

[0118] Step 406: Use users in the fourth user set as positive samples to train the second machine learning model, and use users in the fifth user set as negative samples to train the second machine learning model. Train the second machine learning model to obtain the trained second machine learning model. The second machine learning model is used to calculate the probability of obtaining the second information to be displayed.

[0119] This step is performed according to step 302 of method embodiment three. Assume a second machine learning model is trained and denoted by e(x) = Pr[X], where X represents the user feature vector of the input model sample, Pr[X] represents the second machine learning model, and e(x) represents the probability of the output second information to be displayed. In practical applications, the second machine learning model can be a deep learning model, such as XGBoost or a DNN XGBoost LR model, etc.

[0120] Step 407: Calculate the probability of each user in the fourth user set obtaining the second information to be displayed using the second machine learning model, and determine the proportion of users in each probability interval of the fourth user set according to the set probability interval.

[0121] This step is carried out according to step 303 of the method embodiment three, and it is assumed that the ratio of the number of users in each probability interval is 3:5:9:5:2:1.

[0122] Step 408: Calculate the probability of each user in the third user set obtaining the third information to be displayed using the second machine learning model, and extract users from the third user set according to the proportion of users in each probability interval of the fourth user set. The extracted users form the sixth user set.

[0123] This step is carried out according to step 304 of method embodiment three, and it is assumed that one user is drawn from the 100,000 users in the third user set in a ratio of 3:5:9:5:2:1, and the 10,000 users drawn are used to form the sixth user set.

[0124] Step 409: Use the fourth user set and the sixth user set as training samples, input them into the initialized first machine learning model for training, and obtain the trained first machine learning model. Activated users in the fourth user set and the sixth user set are used as positive samples for training the first machine learning model, and inactive users in the fourth user set and the sixth user set are used as negative samples for training the first machine learning model.

[0125] This step is implemented according to step 305 of method embodiment three, that is, the 300,000 users in the fourth user set and the 10,000 users in the sixth user set are used as samples to input into the first machine learning model. Assume the first machine learning model is represented by p = G(Y|X,T), where G(Y|X,T) represents the first machine learning model, X represents the user feature vector of the input sample, T represents whether the feature to be displayed is obtained, and p represents the output user activation probability. In practical applications, the first machine learning model can also be a deep learning model, such as XGBoost or a DNN XGBoost LR model, etc.

[0126] At this point, the first machine learning model G(Y|X,T) has been obtained after the entire training process.

[0127] Suppose the enterprise receives a new information display request message from the media outlet. The analysis can be performed according to the method described in Example 1 above, specifically: 1) The enterprise receives the first information display request message from the media outlet, which includes a first user ID; 2) The enterprise retrieves matching first information to be displayed from the database based on the first user ID. This first information to be displayed is the information to be displayed to the first user corresponding to the first user ID; 3) The enterprise uses the pre-recorded first user features as the first feature X, and the feature that the first user obtains the first information to be displayed as the second feature (T=1). The first and second features are input as parameters into the trained first machine learning model G(Y|X,T) to obtain the first probability p of user activation. 1; 4) Again, take the first user feature recorded in advance as the first feature X, take the feature that the first user has not obtained the first information to be displayed as the third feature (T=0), and input the first feature and the third feature as parameters into the trained first machine learning model to obtain the second probability p2 of the user being activated; 5) Take the difference between the first probability and the second probability as the incremental value p1-p2 of the user activation probability (also known as incremental MAU); 6) Calculate the first value information Z corresponding to the first information to be displayed based on the incremental value MAU of the user activation probability; 7) Return the first information display reply message, which includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information is evaluated and approved.

[0128] In practical applications, the calculation of the first value information Z in step 6) above can be performed as follows: T = Activation Rate S * Activation Value N * Adjustment Coefficient M * (Incremental MAU / Average Incremental MAU), where the activation rate S represents the probability of a user being activated, the activation value N represents the value generated after user activation, the adjustment coefficient M is a coefficient adjusted according to the actual situation, the incremental MAU is the incremental value of the user activation probability calculated above, and the average incremental MAU represents the historical average of incremental MAU. The activation rate S, activation value N, adjustment coefficient M, and average incremental MAU can be provided by the enterprise's system, while incremental MAU becomes an important factor in calculating the first value information Z.

[0129] By applying the solution of this application embodiment, since the incremental MAU is used as an important factor in calculating the first value information Z, the value of the first information to be displayed can be accurately estimated, thereby maximizing the role of the first value information Z and reducing the cost of pushing the first information to be displayed.

[0130] This application also provides an embodiment of an information display device. Figure 5 This is a schematic diagram of the structure of an embodiment of the device for displaying information according to this application. Figure 5 As shown, the device includes: a receiving module 501, a filtering module 502, a value calculation module 503, and a sending module 504. Wherein:

[0131] The receiving module 501 is used to receive a first information display request message, wherein the first information display request message includes a first user ID.

[0132] The filtering module 502 is used to obtain matching first information to be displayed from the database based on the first user ID. The first information to be displayed is information to be displayed to the first user corresponding to the first user ID.

[0133] The value calculation module 503 takes a pre-recorded first user feature as a first feature, and the feature that the first user has obtained the first information to be displayed as a second feature. The first and second features are input as parameters into a trained first machine learning model to obtain a first probability of user activation. The first machine learning model is used to calculate the user activation probability, where user activation indicates that the user is active within a preset time period. The module then takes the pre-recorded first user feature again as the first feature, and the feature that the first user has not obtained the first information to be displayed as a third feature. The first and third features are input as parameters into the trained first machine learning model to obtain a second probability of user activation. The difference between the first and second probabilities is used as the increment value of the user activation probability. Based on the increment value of the user activation probability, the module calculates the corresponding first value information of the first information to be displayed.

[0134] The sending module 504 is used to return a first information display reply message, which includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information has been evaluated and approved.

[0135] In other words, when the receiving module 501 receives the first information display request message, the filtering module 502 retrieves the matching first information to be displayed from the database based on the first user ID. The value calculation module 503 obtains the first probability and the second probability of user activation, uses the difference between the first and second probabilities as the increment value of the user activation probability, and calculates the first value information corresponding to the first information to be displayed; the sending module 504 returns the first information display reply message.

[0136] The solution applied in this application calculates the first value information corresponding to the first information to be displayed based on the incremental value of the user activation probability, which can maximize the role of the first value information and reduce the cost of pushing the first information to be displayed.

[0137] Figure 6 This is a schematic diagram of the structure of Embodiment 2 of the device for displaying information in this application. Figure 6 As shown, the device includes: a receiving module 501, a filtering module 502, a value calculation module 503, and a sending module 504, and also includes a first training module 505. The receiving module 501, filtering module 502, value calculation module 503, and sending module 504 are the same as those in the first embodiment of the device described above, and will not be repeated here.

[0138] The first training module 505 is used to train the first machine learning model. This includes randomly dividing the obtained user set into experimental and control buckets based on user device IDs. The user set in the experimental bucket is the second user set participating in training, and the user set in the control bucket is the third user set used for testing. For each user in the second user set, a second information to be displayed filtering process, a second value information determination process, and a second value information evaluation process are executed. The second information to be displayed filtering process refers to the process of filtering out the second information to be displayed that matches the user from the database. The second value information determination process refers to the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process is also executed. The process involves incorporating second value information into the evaluation, and obtaining second information to be displayed after the second value information passes the evaluation. For each user in the third user set, a third information to be displayed filtering process and a third value information determination process are executed. The third information to be displayed filtering process refers to the process of filtering out the third information to be displayed that matches the user from the database, and the third value information determination process refers to the process of determining the third value information corresponding to the third information to be displayed. In the second user set, the users who obtain the second information to be displayed form a fourth user set. The first machine learning model after initialization is trained using the fourth user set and the third user set to obtain the trained first machine learning model.

[0139] After training, the first machine learning model can be used by the value calculation module 503.

[0140] The first training module 505, when training the initialized first machine learning model using the fourth user set and the third user set to obtain a trained first machine learning model, includes the following methods: In the second user set, users who have not obtained the second information to be displayed form a fifth user set; users in the fourth user set are used as positive samples for training the second machine learning model, and users in the fifth user set are used as negative samples for training the second machine learning model, to obtain a trained second machine learning model. The second machine learning model is used to calculate the probability of obtaining the second information to be displayed; the second machine learning model is then used to calculate the probability of each user in the fourth user set obtaining the information to be displayed. The probability of obtaining the information to be displayed for each user in the third user set is calculated using the second machine learning model. Users are then selected from the third user set based on the proportion of users in each probability interval of the fourth user set. These selected users form the sixth user set. The fourth and sixth user sets are used as training samples and input into the initialized first machine learning model for training. The activated users in the fourth and sixth user sets are used as positive samples for training the first machine learning model, while the inactive users in the fourth and sixth user sets are used as negative samples for training the first machine learning model.

[0141] In addition, when the filtering module 502 retrieves the first matching information to be displayed from the database based on the first user ID, its method includes: retrieving the corresponding first user characteristics from the database based on the first user ID. The first user characteristics include basic user characteristics and user in-site behavior characteristics. The basic user characteristics include the user's download channel and registration time. The user's in-site behavior characteristics include the number of logins, the number of searches, and the number of pre-transaction process executions. The user's download channel refers to the channel through which the user downloads the application. The application is the target application associated with the first information to be displayed. The registration time is the time since the user registered in the application. The number of logins is the number of times the user logs into the application within a preset time period. The number of searches is the number of times the user performs search behavior within a preset time period. The number of pre-transaction process executions is the number of times the user performs pre-transaction process behavior within a preset time period. The candidate information to be displayed is filtered from the database based on the first user characteristics. The candidate information to be displayed is sorted according to a pre-set sorting strategy, and the candidate information to be displayed with the highest sorting is selected as the first information to be displayed.

[0142] In this embodiment, the first machine learning model is trained using the fourth and sixth user sets as samples. The users in the fourth user set are those who have obtained the second information to be displayed, while the users in the sixth user set are homogeneous users who have not obtained the third information to be displayed. Therefore, the probability of a user being activated can be attributed to whether or not the value information evaluation process has been performed, thus making the trained first machine learning model more reliable.

[0143] This application also provides a computer-readable medium storing instructions that, when executed by a processor, can perform the information display method described above. In practical applications, the computer-readable medium may be included in the device / apparatus / system described in the above embodiments, or it may exist independently without being assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, and when one or more programs are executed, the information display method described in the above embodiments can be implemented. According to the embodiments disclosed in this application, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof, but this is not intended to limit the scope of protection of this application. In the embodiments disclosed in this application, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0144] This application also provides a computer program product, which includes computer instructions that, when executed by a processor, implement the information display method as described in any of the above embodiments.

[0145] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments disclosed in this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings. For example, two blocks shown connectedly may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0146] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this disclosure can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, without departing from the spirit and teachings of this application, the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways, and all such combinations and / or combinations fall within the scope of this application.

[0147] This document uses specific embodiments to illustrate the principles and implementation methods of the present invention. The descriptions of these embodiments are merely illustrative of the method and core concepts of the present invention and are not intended to limit this application. Those skilled in the art can make changes to the specific implementation methods and application scope based on the ideas, spirit, and principles of the present invention. Any modifications, equivalent substitutions, or improvements made should be included within the scope of protection of this application.

Claims

1. A method for displaying information, characterized in that, The method includes: A first information display request message is received, the first information display request message including a first user ID; Based on the first user ID, the first information to be displayed is retrieved from the database to match the first user ID. The first information to be displayed is the information to be displayed to the first user corresponding to the first user ID. The first user feature recorded in advance is used as the first feature, and the feature that the first user obtains the first information to be displayed is used as the second feature. The first feature and the second feature are used as parameters and input into the trained first machine learning model to obtain the first probability of the user being activated. The first machine learning model is a model used to calculate the probability of the user being activated. The user being activated means that the user is in an active state within a preset time. The first user feature, which was recorded in advance, is used again as the first feature, and the feature that the first user has not obtained the first information to be displayed is used as the third feature. The first feature and the third feature are used as parameters and input into the trained first machine learning model to obtain the second probability that the user is activated. The difference between the first probability and the second probability is used as the increment of the user's activation probability. The first value information corresponding to the first information to be displayed is calculated based on the incremental value of the user activation probability. The system returns a first information display reply message, which includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information has been evaluated and approved.

2. The method according to claim 1, characterized in that, Before the step of receiving the first information display request message, the method further includes: training the first machine learning model; The steps for training the first machine learning model include: The obtained user set is randomly divided into experimental buckets and control buckets according to the user device ID. The user set in the experimental bucket is the second user set used for training, and the user set in the control bucket is the third user set used for testing. For each user in the second user set, a second information to be displayed filtering process, a second value information determination process, and a second value information evaluation process are executed respectively. The second information to be displayed filtering process refers to the process of filtering out the second information to be displayed that matches the user from the database. The second value information determination process refers to the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process refers to the process of participating the second value information in the evaluation, and the second information to be displayed is obtained after the second value information is evaluated and passed. For each user in the third user set, a third information to be displayed filtering process and a third value information determination process are executed respectively. The third information to be displayed filtering process refers to the process of filtering out the third information to be displayed that matches the user from the database. The third value information determination process refers to the process of determining the third value information corresponding to the third information to be displayed. In the second user set, the users who obtained the second information to be displayed will form a fourth user set; The first machine learning model after initialization is trained using the fourth user set and the third user set to obtain the trained first machine learning model.

3. The method according to claim 2, characterized in that, The step of training the initialized first machine learning model using the fourth user set and the third user set to obtain the trained first machine learning model includes: In the second user set, users who have not obtained the second information to be displayed form the fifth user set; Users in the fourth user set are used as positive samples to train the second machine learning model, and users in the fifth user set are used as negative samples to train the second machine learning model. The second machine learning model is trained to obtain the trained second machine learning model, which is used to calculate the probability of obtaining the second information to be displayed. The probability of each user in the fourth user set obtaining the second information to be displayed is calculated using the second machine learning model, and the proportion of users in each probability interval of the fourth user set is determined according to the set probability interval. The probability of each user in the third user set obtaining the third information to be displayed is calculated using the second machine learning model, and users are extracted from the third user set according to the proportion of users in each probability interval of the fourth user set. The extracted users form the sixth user set. The fourth user set and the sixth user set are used as training samples and input into the initialized first machine learning model for training to obtain the trained first machine learning model. The activated users in the fourth user set and the sixth user set are used as positive samples for training the first machine learning model, and the inactive users in the fourth user set and the sixth user set are used as negative samples for training the first machine learning model.

4. The method according to any one of claims 1 to 3, characterized in that, The step of retrieving the first matching information to be displayed from the database based on the first user ID includes: The first user characteristics are retrieved from the database based on the first user ID. The first user characteristics include basic user characteristics and user in-site behavior characteristics. The basic user characteristics include the user's download channel and the time since registration. The user's in-site behavior characteristics include the number of times the user logs in, the number of times the user searches, and the number of times the user performs pre-transaction processes. The user's download channel refers to the channel through which the user downloads the application. The application is the target application associated with the first information to be displayed. The time since registration is the time since the user registered in the application. The number of times the user logs in is the number of times the user logs in to the application within a preset time period. The number of times the user searches is the number of times the user performs search behavior within a preset time period. The number of times the user performs pre-transaction processes is the number of times the user performs pre-transaction process behavior within a preset time period. Based on the first user characteristic, candidate display information is filtered from the database; The information is sorted according to a pre-set sorting strategy, and the candidate information ranked first is selected as the first information to be displayed.

5. An information display device, characterized in that, The device includes: a receiving module, a filtering module, a value calculation module, and a sending module; The receiving module is configured to receive a first information display request message, wherein the first information display request message includes a first user ID; The filtering module is used to obtain matching first information to be displayed from the database based on the first user ID. The first information to be displayed is information to be displayed to the first user corresponding to the first user ID. The value calculation module takes a pre-recorded first user feature as a first feature, and the feature that the first user has obtained the first information to be displayed as a second feature. The first and second features are input as parameters into a trained first machine learning model to obtain a first probability of user activation. The first machine learning model is used to calculate the user activation probability, where user activation indicates that the user is active within a preset time period. The module then takes the pre-recorded first user feature as the first feature again, and the feature that the first user has not obtained the first information to be displayed as a third feature. The first and third features are input as parameters into the trained first machine learning model to obtain a second probability of user activation. The difference between the first and second probabilities is used as the increment of the user activation probability. Based on the increment of the user activation probability, the module calculates the corresponding first value information of the first information to be displayed. The sending module is used to return a first information display reply message, which includes the first information to be displayed and the first value information. The first information to be displayed is displayed to the first user after the first value information has been evaluated and approved.

6. The apparatus according to claim 5, characterized in that, The device further includes a first training module; The first training module is used to train the first machine learning model, including randomly dividing the obtained user set into experimental buckets and control buckets according to user device IDs. The user set in the experimental bucket is the second user set participating in training, and the user set in the control bucket is the third user set used for testing. For each user in the second user set, a second information to be displayed screening process, a second value information determination process, and a second value information evaluation process are executed respectively. The second information to be displayed screening process refers to the process of screening out the second information to be displayed that matches the user from the database. The second value information determination process refers to the process of determining the second value information corresponding to the second information to be displayed. The second value information evaluation process refers to the process of participating the second value information in the evaluation, and the second information to be displayed is obtained after the second value information is evaluated. For each user in the third user set, a third information to be displayed filtering process and a third value information determination process are executed respectively. The third information to be displayed filtering process refers to the process of filtering out the third information to be displayed that matches the user from the database. The third value information determination process refers to the process of determining the third value information corresponding to the third information to be displayed. In the second user set, the users who obtained the second information to be displayed form a fourth user set; the first machine learning model after initialization is trained using the fourth user set and the third user set to obtain the trained first machine learning model.

7. The apparatus according to claim 6, characterized in that, When the first training module trains the initialized first machine learning model using the fourth user set and the third user set to obtain the trained first machine learning model, the process includes: in the second user set, users who have not obtained the second information to be displayed form a fifth user set; using users in the fourth user set as positive samples for training the second machine learning model, and using users in the fifth user set as negative samples for training the second machine learning model, the second machine learning model is trained to obtain a trained second machine learning model, which is used to calculate the probability of obtaining the second information to be displayed; using the second machine learning model to calculate the probability of each user in the fourth user set obtaining the information to be displayed, and according to a given... The probability intervals are set to determine the proportion of users in each probability interval of the fourth user set; the probability of each user in the third user set obtaining the information to be displayed is calculated using the second machine learning model, and users are extracted from the third user set according to the proportion of users in each probability interval of the fourth user set to form a sixth user set; the fourth user set and the sixth user set are used as training samples and input into the initialized first machine learning model for training to obtain the trained first machine learning model, and the activated users in the fourth user set and the sixth user set are used as positive samples for training the first machine learning model, and the inactive users in the fourth user set and the sixth user set are used as negative samples for training the first machine learning model.

8. The apparatus according to any one of claims 5 to 7, characterized in that, When the filtering module retrieves the first matching information to be displayed from the database based on the first user ID, it includes: retrieving the corresponding first user feature from the database based on the first user ID; the first user feature includes basic user features and user in-site behavior features; the basic user features include the user's download channel and registration date; the user's in-site behavior features include the number of logins, the number of searches, and the number of pre-transaction process executions; the user's download channel refers to the channel through which the user downloaded the application; the application is the target application associated with the first information to be displayed; the registration date is the time since the user registered in the application; the number of logins is the number of times the user logged into the application within a preset time period; the number of searches is the number of times the user performed search behavior within the preset time period; and the number of pre-transaction process executions is the number of times the user performed pre-transaction process behavior within the preset time period. The module then filters candidate information to be displayed from the database based on the first user feature; sorts the candidate information according to a pre-set sorting strategy, and selects the candidate information ranked first as the first information to be displayed.

9. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the instruction is executed by the processor, it can implement the information display method according to any one of claims 1 to 4.

10. A computer program product comprising computer instructions that, when executed by a processor, implement the information display method as described in any one of claims 1 to 4.

Images