APP intelligent push engine and method based on prediction model

Through the APP intelligent push engine based on the prediction model, the push time period and user volume are dynamically adjusted, which solves the problem of excessive system load and conflicts between promotion and stability in traditional push technology, and achieves more efficient and accurate message push.

CN114004630BActive Publication Date: 2024-11-05FUJIAN FUNO MOBILE COMM TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202111191523.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-10-13
Publication Date
2024-11-05
Estimated Expiration
2041-10-13

AI Technical Summary

Technical Problem

The message push technology of traditional mobile APP pushes messages to all users in a fixed period of time, resulting in excessive system load and lack of dynamic adjustments, and the balance between activity promotion and stable system operation cannot be taken into account.

Method used

Using the APP intelligent push engine based on prediction model, through push result collection, activity participation recording, push performance statistics and prediction model construction, the push volume ratio and number of users in each time period are calculated, and the push time period and user volume are dynamically adjusted to realize push Intelligent distribution of the engine.

Benefits of technology

It reduces system pressure, improves push efficiency, reduces user harassment, enhances promotion results, adapts to changes in system operation status and user activity, and reduces fault complaints.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114004630B_ABST
    Figure CN114004630B_ABST
Patent Text Reader

Abstract

The present invention proposes an APP intelligent push engine and method based on a prediction model, including: a push result collection module, used to obtain the time when the push reaches the user and the time when the user responds to the push information; an activity participation record module, used to obtain the time when the user starts to participate in the activity and the time when the result is returned; a push effectiveness statistics module, used to obtain the delay parameters representing the user's activeness and system load according to the time information collected by the push result collection module and the activity participation record module; a prediction model construction module, used to construct a push prediction model based on the balance weight between the promotion effect and the system load according to the delay parameter; an intelligent push allocation module, used to allocate the push time period and the push user quantity according to the push prediction model based on the daily estimated weight distribution of each time period to perform specific push operations. This solves the problem that the existing push technology fails to dynamically arrange the time period, and the conflict between the activity promotion and the stable operation of the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of message push technology, and in particular to an APP intelligent push engine and method based on a prediction model. Background Art

[0002] Traditional mobile app marketing mainly pushes various event messages to registered users to promote new businesses and new activities, while increasing user activity and their stickiness to the platform. On the one hand, it enhances the attractiveness to new users, and on the other hand, it improves the retention of existing users.

[0003] In order to increase the exposure of event information and improve the reach of message push, the management backend of the mobile app usually initiates push notifications to all registered users, striving to ensure that every user can receive push messages. The push time is generally selected to push all messages during a certain time period, such as 9:00-10:00 in the morning and 15:00-16:00 in the afternoon. After receiving the message, the mobile app will display the event description in the notification bar. The advantage of choosing to push during a fixed time period is that it is convenient to prepare the event content in advance, and then the management backend will trigger the push operation at a fixed time. In addition, according to past experience, users will be more active during these two time periods and are relatively easy to participate in event acceptance.

[0004] While traditional push notification technology is simple and easy to implement, it can push activity messages to all users within a fixed time period. However, this method has the following technical drawbacks:

[0005] 1. Pushing messages to all users at once will cause excessive system load during that time period, affecting normal business processing;

[0006] 2. Setting push time periods based on human experience, lacking actual operational data support, and failing to arrange time more rationally;

[0007] 3. The system load before and after the system upgrade changes dynamically, and the user's active time period in different seasons is also constantly adjusting. Fixed-time push does not take these fluctuations into account. Summary of the Invention

[0008] To reduce system pressure, improve push efficiency, and minimize user nuisance, a more accurate and efficient campaign push mechanism is needed. The key is to automatically analyze past campaign push results and the system's response at the time, predict daily user push distribution based on this, and intelligently schedule the number of campaign pushes per time period, achieving a balance between campaign promotion and system stability.

[0009] To address the shortcomings and deficiencies of existing technologies, the present invention proposes an intelligent APP push engine and method based on a prediction model, aiming to solve the problems of existing push technologies that fail to dynamically schedule time periods, conflict between activity promotion and stable system operation.

[0010] The present invention specifically includes the following contents:

[0011] An APP intelligent push engine based on a prediction model, characterized by including:

[0012] The push result collection module is used to obtain the time when the push reaches the user and the time when the user responds to the push information;

[0013] Activity participation record module, used to obtain the time when the user starts participating in the activity and the time when the result is returned;

[0014] A push effectiveness statistics module, configured to obtain delay parameters representing user activity and system load respectively based on the time information collected by the push result collection module and the activity participation recording module;

[0015] The prediction model building module is used to build a push prediction model based on the balance weight between the promotion effect and the system load according to the delay parameter;

[0016] The intelligent push allocation module is used to allocate push time periods and push user quantities to perform specific push operations based on the daily estimated weight distribution of each time period according to the push prediction model.

[0017] Furthermore, notification response delay = notification click time - push arrival time, which is used to characterize promotion effectiveness; activity participation delay = result return time - participation start time, which is used to characterize system load.

[0018] Furthermore, in the push prediction model, the promotion effectiveness of a certain time period is associated with the inverse of the delay parameter to construct a balance weight parameter for obtaining the promotion effectiveness; and then the balance weight parameter is used to estimate the proportion of push volume allocated to each time period and the number of push users allocated to each time period based on different time periods.

[0019] Furthermore, the delay parameter is defined as:

[0020]

[0021] Where D represents the delay parameter of user activity or system load, Xi represents the delay size of each record, and n represents the total number of records;

[0022] The balance weight parameter is defined as:

[0023]

[0024] Where Wr represents the actual balance weight of a time period, Da represents the notification response delay parameter of the time period, Db represents the activity participation delay parameter of the time period, and 100 is a fixed constant factor.

[0025] Furthermore, the estimation model of the balance weight parameter is:

[0026]

[0027] Where Wf represents the estimated balance weight for a certain period of time, Wi represents the actual daily balance weight for the previous 7 days, and i represents the day number in these 7 days. Date factor for adjusting weight;

[0028] Calculate the push volume ratio allocated to each time period = estimated weight value of the time period / total estimated weight value of all time periods:

[0029]

[0030] Here, Ri represents the push volume ratio in a certain time period, Wi' represents the estimated weight value of the time period, and m represents the number of hours available for push on that day.

[0031] Furthermore, after obtaining the push volume ratio for each time period, multiply it by the total number of users to be pushed to obtain the number of push users allocated for each time period:

[0032]

[0033] Among them, Ni represents the number of push users allocated in a certain time period, and S represents the total number of users to be pushed.

[0034] And, a method for intelligently pushing apps based on a prediction model, characterized by comprising the following steps:

[0035] Step S1: Based on previous push results and the current system pressure, a push prediction model is established in the recent spatiotemporal context;

[0036] Step S2: Calculate the balance weight between promotion effectiveness and system load for each push time period based on the delay in users clicking on the notification and the delay in accepting the activity;

[0037] Step S3: Based on the daily estimated weight distribution of each time period, the prediction model allocates the push time period and the number of push users for the day, and then the push engine performs the specific push operation.

[0038] And, an electronic device, including a memory, a processor and a computer program stored in the memory and runnable on the processor, characterized in that when the processor executes the program, the steps of the APP intelligent push method based on the prediction model as described above are implemented.

[0039] And, a non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the above-mentioned method for intelligent APP push based on the prediction model are implemented.

[0040] Compared with existing push technologies, the present invention and its preferred solution more flexibly address the shortcomings of traditional push technologies' centralized notifications. The predictive model accurately predicts the appropriate push volume for each time period, thus avoiding the short-term surge in system pressure caused by blind centralized push notifications. The predictive model can timely estimate the balance weights for each time period based on recent push results and system pressure, dynamically adjusting the daily push volume distribution. This allows the push engine to intelligently and gradually adapt to environmental changes regardless of changes in system operating conditions or user activity, reducing unnecessary fault complaints and improving the effectiveness of new campaign promotions, thus possessing high practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0042] Figure 1 Schematic diagram of the system framework and workflow of an embodiment of the present invention;

[0043] Figure 2 This is an example diagram of a weight curve according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] To make the features and advantages of this patent more clearly understood, the following embodiments are specifically described in detail with reference to the accompanying drawings:

[0045] like Figure 1 As shown, the solution of the APP intelligent push engine based on the prediction model provided in this embodiment mainly includes a push result collection module, an activity participation record module, a push effectiveness statistics module, a prediction model construction module, and an intelligent allocation push module.

[0046] The specific process of building and executing the engine includes:

[0047] Step 1: After the activity message is pushed to the mobile phone notification bar, the user clicks the notification to open the APP activity page. At this time, the WEB server records the user's notification click result.

[0048] Step 2: The user participates in an activity on the APP page (such as a lottery, application, group formation, etc.). At this time, the WEB server records the user's activity participation results.

[0049] Step 3: Based on the notification click result in Step 1, the push arrival time and notification click time can be calculated as: Notification Response Latency = Notification Click Time - Push Arrival Time. Based on the activity participation result in Step 2, the activity participation delay = Result Return Time - Start Participation Time can be calculated as: Activity Participation Latency = Result Return Time - Start Participation Time.

[0050] According to the solution constructed in this embodiment, the notification response delay reflects the user's activity level during that period. A larger value indicates a less active user, and push effectiveness will decrease. The activity participation delay, on the other hand, reflects the system load during that period. A larger value indicates a busier system, and large amounts of message push should be avoided.

[0051] To express the effectiveness of push notifications over a certain period of time, the notification response delay value is derived from the arithmetic mean of all click record delays over that period, and the activity participation delay value is derived from the arithmetic mean of all participation record delays over that period.

[0052]

[0053] However, considering that the linear delay span is large, it is not conducive to subsequent statistical analysis. Therefore, the arithmetic mean can be squared and its square root can be taken as the delay parameter for subsequent modeling.

[0054]

[0055] Where D represents the delay parameter used for modeling, Xi represents the delay of each record, and n represents the total number of records.

[0056] Step 4: Since notification response latency reflects user activity during that time period, and activity participation latency reflects system load during that time period, and both values ​​are inversely proportional to the effectiveness of the activity promotion, the promotion effectiveness of a given time period is correlated with the inverse of the latency parameter. Since promotion effectiveness reflects the impact of both notification response latency and activity participation latency, the promotion effectiveness balance weight used in this embodiment is the geometric mean of the two latency parameters.

[0057]

[0058] Where Wr represents the actual balance weight of a time period, Da represents the notification response delay parameter of the time period, Db represents the activity participation delay parameter of the time period, and 100 is a fixed constant factor.

[0059] The balance weight calculated above is the value calculated based on the actual push results. By plotting the daily balance weight of each time period on the coordinate system, a gently fluctuating weight curve can be obtained, such as Figure 2 shown.

[0060] The solid line portion of the curve represents the actual weight distribution in the past, and the dotted line portion of the curve represents the estimated weight distribution to be predicted.

[0061] Now that the forecasting model has been initially established, we can use it to estimate the push volume for each time period of the day. To ensure that the estimated results better reflect recent promotion effectiveness, only the balance weights for the last seven days are used in the calculation, with more recent dates receiving higher weights to reflect the latest situation. To achieve this, the balance weight for each date is multiplied by a date factor, with the factor increasing for more recent dates.

[0062]

[0063] Where Wf represents the estimated balance weight for a certain period of time, Wi represents the actual daily balance weight for the previous 7 days, and i represents the day number in these 7 days. Date factor for adjusting weight.

[0064] Step 5: Estimate the balance weight for each time period based on the prediction model. Calculate the proportion of push volume allocated to each time period = estimated weight value for that time period / (sum of estimated weights for all time periods).

[0065]

[0066] Among them, Ri represents the push volume ratio in a certain time period, Wi represents the estimated weight value of the time period, and n represents the number of hours that can be pushed on that day.

[0067] After obtaining the push volume ratio for each time period, multiply it by the total number of users to be pushed to obtain the number of push users allocated to each time period.

[0068]

[0069] Among them, Ni represents the number of push users allocated in a certain time period, and S represents the total number of users to be pushed.

[0070] Then, when the start time of each time period arrives, the push engine pushes activity messages to the number of users assigned to the time period in turn, thereby achieving the goal of intelligently pushing activity notifications based on the prediction model.

[0071] The implementation of the above scheme can also be summarized into the following process:

[0072] 1. Based on past push results and the current system pressure, establish a push volume prediction model in the recent time and space context;

[0073] 2. Calculate the balance between promotion effectiveness and system load for each push time period based on the delay in clicking on the notification and the delay in accepting the activity;

[0074] 3. Based on the daily estimated weight distribution of each time period, the prediction model allocates the push time period and the number of users to be pushed on that day, and then the push engine executes the specific push operation.

[0075] The above solution provided in this embodiment can be stored in a computer-readable storage medium in a coded form and implemented in the form of a computer program, and the basic parameter information required for calculation is input through computer hardware, and the calculation results are output.

[0076] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatuses, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0077] The present invention is described with reference to flowcharts of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process in the flowcharts, as well as combinations of processes in the flowcharts, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts. Figure 1 A device that specifies functions in a process or multiple processes.

[0078] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple functions specified in a flowchart.

[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1The steps of a specified function in a process or multiple processes.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

[0081] This patent is not limited to the above-mentioned best implementation method. Anyone can derive various other forms of APP intelligent push engines and methods based on prediction models under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of this invention should be covered by this patent.

Claims

1. An APP intelligent push engine based on a prediction model, characterized in that: include: The push result collection module is used to obtain the time when the push message reaches the user and the time when the user responds to the push message; Activity participation record module, used to obtain the time when the user starts participating in the activity and the time when the result is returned; A push result statistics module, used to obtain delay parameters representing user activity and system load respectively according to the time information collected by the push result collection module and the activity participation recording module; A prediction model building module is used to build a push prediction model based on the balance weight between the promotion effect and the system load according to the delay parameter; An intelligent push allocation module is used to allocate push time periods and push user quantities to perform specific push operations based on the push prediction model and the daily estimated weight distribution of each time period; Notification response delay = notification click time - push arrival time, used to characterize promotion effectiveness; activity participation delay = result return time - start participation time, used to characterize system load; In the push prediction model, the promotion effect of a certain time period is associated with the inverse of the delay parameter to construct a balance weight parameter for obtaining the promotion effect; and then the balance weight parameter is used to estimate the proportion of push volume allocated to each time period and the number of push users allocated to each time period. The delay parameter is defined as: Where D represents the delay parameter of user activity or system load, Xi represents the delay size of each record, and n represents the total number of records; The balance weight parameter is defined as: Wherein, Wr represents the actual balance weight of a certain time period, Da represents the notification response delay parameter of the time period, Db represents the activity participation delay parameter of the time period, and 100 is a fixed constant factor.

2. The APP intelligent push engine based on the prediction model according to claim 1 is characterized by: The estimation model of the balance weight parameter is: Among them, Wf represents the estimated balance weight of a certain period of time, Wi represents the actual daily balance weight of the previous 7 days, and i represents the day of the 7 days. Date factor for adjusting weight; Calculate the push volume ratio allocated to each time period = estimated weight value of the time period / total estimated weight value of all time periods: Among them, Ri represents the push volume ratio in a certain period of time, W i ′ It represents the estimated weight value of the time period, and m represents the number of hours that can be pushed on that day.

3. The APP intelligent push engine based on the prediction model according to claim 2 is characterized by: After obtaining the push volume ratio for each time period, multiply it by the total number of users to be pushed to get the number of push users allocated to each time period: N i =S*R i Among them, Ni represents the number of push users allocated in a certain time period, and S represents the total number of users to be pushed.

4. A method for intelligently pushing apps based on a prediction model, characterized in that: The following steps are involved: Step S1: Based on the previous push results and the current system pressure, a push prediction model in the recent time and space context is established; Step S2: Calculate the balance weight between promotion effectiveness and system load for each push time period according to the delay of the user clicking the notification and the delay of accepting the activity; Step S3: Based on the daily estimated weight distribution of each time period, the prediction model allocates the push time period and the number of push users for the day, and then the push engine performs specific push operations; Notification response delay = notification click time - push arrival time, used to characterize promotion effectiveness; activity participation delay = result return time - start participation time, used to characterize system load; The push prediction model associates the promotion effect of a certain time period with the inverse of the delay parameter to construct a balance weight parameter for obtaining the promotion effect; and then estimates the push volume ratio allocated to each time period and the number of push users allocated to each time period by using the balance weight parameter. The delay parameter is defined as: Where D represents the delay parameter of user activity or system load, Xi represents the delay size of each record, and n represents the total number of records; The balance weight parameter is defined as: Where Wr represents the actual balance weight of a time period, Da represents the notification response delay parameter of the time period, Db represents the activity participation delay parameter of the time period, and 100 is a fixed constant factor; The estimation model of the balance weight parameter is: Among them, Wf represents the estimated balance weight of a certain period of time, Wi represents the actual daily balance weight of the previous 7 days, and i represents the day of the 7 days. Date factor for adjusting weight; Calculate the push volume ratio allocated to each time period = estimated weight value of the time period / total estimated weight value of all time periods: Among them, Ri represents the push volume ratio in a certain time period, Wi' represents the estimated weight value of the time period, and m represents the number of hours that can be pushed on that day.