CN111405030B

CN111405030B - Message pushing method and device, electronic equipment and storage medium

Info

Publication number: CN111405030B
Application number: CN202010170226.1A
Authority: CN
Inventors: 刘卓
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-03-12
Filing date: 2020-03-12
Publication date: 2021-08-10
Anticipated expiration: 2040-03-12
Also published as: CN111405030A

Abstract

The application relates to the technical field of internet, in particular to a message pushing method, a message pushing device, electronic equipment and a storage medium, which are used for improving the click rate of a user in low-frequency application, wherein the method comprises the following steps: acquiring behavior data of a target object generated for high-frequency application in a preset time period, wherein the high-frequency application is an application with the use frequency greater than a preset frequency threshold in the preset time period; acquiring behavior habit characteristics of the target object based on the behavior data; determining idle time of the target object capable of using the low-frequency application according to the behavior habit characteristics; and triggering the low-frequency application to send the push message in idle time. According to the embodiment of the application, the user behavior portrait is established through the behavior data of the user to the high-frequency application, the idle time of the user is explored, the low-frequency application is triggered to push messages to the user in the idle time, and the click rate of the user in the low-frequency application is improved.

Description

Message pushing method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of internet technologies, and in particular, to a method and an apparatus for pushing a message, an electronic device, and a storage medium.

Background

With the popularization and application of smart phones in a large number, smart phones can support more and more applications and have more and more powerful functions, and smart phones develop towards diversification and individuation and become indispensable electronic products in user life.

Among them, applications can be divided into two broad categories according to user preferences: frequently used high frequency Application and the infrequently used low frequency Application, the scheme that user duration of use at low frequency APP (Application ) is promoted at present, and the user is attracted through directly regularly pushing screen locking information to low frequency APP mostly.

However, the way of timing push does not consider the behavior habit of the user, and the click desire of the user is low.

Disclosure of Invention

The embodiment of the application provides a message pushing method and device, electronic equipment and a storage medium, which are used for improving the click rate of a user in low-frequency application.

The first message pushing method provided by the embodiment of the application comprises the following steps:

acquiring behavior data of a target object generated for high-frequency application in a preset time period, wherein the high-frequency application is an application with the use frequency greater than a preset frequency threshold in the preset time period;

acquiring behavior habit characteristics of the target object based on the behavior data;

determining idle time of the target object capable of low-frequency application according to the behavior habit characteristics, wherein the low-frequency application is an application with the use frequency not greater than the preset frequency threshold value in the preset time period;

and triggering the low-frequency application to send a push message in the idle time.

The first message pushing apparatus provided in the embodiment of the present application includes:

the data acquisition unit is used for acquiring behavior data of a target object generated for high-frequency application in a preset time period, wherein the high-frequency application is an application with the use frequency greater than a preset frequency threshold in the preset time period;

the characteristic acquisition unit is used for acquiring the behavior habit characteristics of the target object based on the behavior data;

the idle time determining unit is used for determining the idle time of the target object capable of low-frequency application according to the behavior habit characteristics, wherein the low-frequency application is an application with the use frequency not greater than the preset frequency threshold value in the preset time period;

and the triggering unit is used for triggering the low-frequency application to send the push message in the idle time.

Optionally, the feature obtaining unit is specifically configured to determine the behavior habit feature of the target object through any one or more of the following manners:

acquiring the activity of the target object in each unit time of a first statistical time period according to the at least one structured data event, taking the average value of the activity as the long-term activity characteristic of the target object, and determining the behavior habit characteristic of the target object according to the long-term activity characteristic, wherein the first statistical time period is within the range of the preset time period;

acquiring the times of the target object for implementing each operation in a second statistical time period according to the at least one structured data event, taking the acquired times as short-term characteristics of the target object, and determining behavior habit characteristics of the target object according to the short-term characteristics, wherein the second statistical time period is within the preset time period range, and the duration of the second statistical time period is less than that of the first statistical time period;

and acquiring instant state information of the high-frequency application at the current moment according to the at least one structured data event, taking the instant state information as instant characteristics of the target object, and determining behavior habit characteristics of the target object according to the instant characteristics.

Optionally, the behavior habit feature further comprises a feedback feature; the feature acquisition unit is further configured to:

after a push message is displayed on the target object, feedback information used for indicating whether the target object clicks the push message or not is obtained according to the at least one structured data event, and the feedback information is used as a feedback characteristic of the target object; and/or

The behavioral habit features also include static features; the feature acquisition unit is further configured to:

and acquiring identity information input when the target object registers the application according to the at least one structured data event, and taking the identity information as the static feature of the target object.

An electronic device provided in an embodiment of the present application includes a processor and a memory, where the memory stores program codes, and when the program codes are executed by the processor, the processor is caused to execute any one of the steps of the message pushing method.

An embodiment of the present application provides a computer-readable storage medium, which includes program code, when the program product runs on an electronic device, the program code is configured to enable the electronic device to execute the steps of any one of the above-mentioned message pushing methods.

The beneficial effect of this application is as follows:

according to the message pushing method, the message pushing device, the electronic equipment and the storage medium, because the user behavior portrait is established through the behavior data of the user on the high-frequency application, the idle time of the user is explored, and the low-frequency application is triggered to push the message to the user in the idle time, the click rate of the user on the low-frequency application is improved, and the use duration and the activity of the user on the low-frequency application are improved.

Additional features and advantages of the application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the application. The objectives and other advantages of the application may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:

fig. 1 is an alternative schematic diagram of an application scenario in an embodiment of the present application;

fig. 2 is a flowchart of a message pushing method in an embodiment of the present application;

fig. 3 is a schematic diagram of an interface for pushing a message in an embodiment of the present application;

fig. 4A is a schematic diagram of a message pushing method in an embodiment of the present application;

fig. 4B is a schematic diagram of another message pushing method in the embodiment of the present application;

FIG. 5 is a schematic diagram illustrating an alternative interactive implementation timing sequence in the embodiments of the present application;

fig. 6 is a schematic structural diagram of a message pushing apparatus in an embodiment of the present application;

fig. 7 is a block diagram of an electronic device in an embodiment of the present application;

fig. 8 is a schematic diagram of a hardware component of a computing device to which an embodiment of the present invention is applied.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments, but not all embodiments, of the technical solutions of the present application. All other embodiments obtained by a person skilled in the art without any inventive step based on the embodiments described in the present application are within the scope of the protection of the present application.

Some concepts related to the embodiments of the present application are described below.

Target object: the term "object" refers to an object targeted at action or thinking, and in the present embodiment, the object refers to a user targeted at message pushing or a user account.

Structuring data: simply a database. More easily understood when incorporated into typical scenarios, such as Enterprise ERP (Enterprise Resource Planning), financial systems; a medical HIS (Hospital Information System) database; an education all-purpose card; government administration approval; other core databases, etc. The storage schemes of these applications basically include high-speed storage application requirements, data backup requirements, data sharing requirements, and data disaster tolerance requirements. The structured data in the embodiment of the present application refers to behavior data generated when a user uses an application, that is, data generated when the user performs various operations on the application. The structured data event is used for representing an event which occurs when the user implements each operation, and each piece of reported structured data is abstracted into the structured data event which occurs when the user, such as a screen switching event, an event when the user enters the APP, an event consumed by the user in the APP, and the like.

Behavior habit characteristics: the behavior habit of the user using the application is obtained according to the statistics of the behavior data of the user. Specifically, any one or more of long-term activity characteristics, short-term characteristics, and instant characteristics of the user may be included.

Wherein, the long-term activity characteristic: refers to the activity of various operations by the user within the last few days, months, or even years. Such as the activity of using the APP, the activity of consuming within the APP, and so forth. Liveness can be measured in a number of ways, such as the number of days that a user has been active recently, the number of times that a user has been active for the last few days, and so on.

Short-term characteristics: refers to the statistics of the user's various behaviors over the last several seconds, several minutes. For example, for a chat application, the number of times that a dynamic published by a friend in the chat application has been seen in the last 30 seconds, the number of times that hot news has been seen in the last 10 seconds, and the like may be counted.

Instant feature: the method refers to the instant state information of the APP when the user operates currently. For example, for the chat application a, the APP may have information such as the number of red dots in the current message list of the user, whether the current news is at a hotspot, and the like.

In addition, the behavior habit feature in the embodiment of the present application may further include any one or more of a feedback feature and a static feature.

Wherein, the feedback characteristic: the user feedback feature is constructed by judging whether the user clicks the push message after pushing the message to the user. The user feedback characteristics mainly comprise the exposure times, the click times and the click rate of the messages pushed by the user.

Static characteristics: the characteristics refer to the gender and age of the user, and the like, and do not change in a short period. Such features are obtained through material filled in by the user at APP registration. And feature correctness checking and outlier processing are required.

UV (uniform viewer), refers to the number of people who visit a certain site or click on a different IP (Internet Protocol) address of a certain news item. Within the same day, the UV only records visitors with independent IP that entered the web site for the first time, but does not count visiting the web site again within the same day. Individual IP visitors provide statistical indications of the number of different viewers over a period of time without reflecting the overall activity of the web site.

Artificial Intelligence (AI): the method is a theory, method, technology and application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by the digital computer, sensing the environment, acquiring knowledge and obtaining the best result by using the knowledge. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.

The prediction models in the message pushing method provided in the embodiment of the application belong to machine learning models, and the models relate to the technical field of machine learning, and the models can be trained through the technology of machine learning. The prediction model is mainly used for predicting whether a click behavior occurs after the low-frequency APP is triggered to push a message to a user at the current moment after various behavior habit features of the user are collected.

The preferred embodiments of the present application will be described below with reference to the accompanying drawings of the specification, it should be understood that the preferred embodiments described herein are merely for illustrating and explaining the present application, and are not intended to limit the present application, and that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

The following briefly introduces the design concept of the embodiments of the present application:

due to the APP used at the low frequency, the operation behaviors of the user are less, and the behavior picture of the user is difficult to establish, so that the idle time of the user is difficult to discover, and the click rate of the user using the low frequency application is basically improved by adopting a mode of pushing messages at regular time. However, the behavior habit of the user is not considered in the timing pushing process, and the clicking desire of the user is low.

In view of this, the embodiment of the application provides a message pushing method, a device, electronic equipment and a storage medium, based on a record of a user operation behavior when a user uses a high-frequency APP, it is determined that the user is more inclined to use an idle time of a low-frequency APP, and then the low-frequency APP is triggered to push a message to the user in the idle time, so that the use duration, the click rate, the activity and the like of the user at the low-frequency APP are improved, and the user experience is enhanced.

Fig. 1 is a schematic view of an application scenario in the embodiment of the present application. The application scenario diagram includes two terminal devices 110 and a server 130, and the terminal devices 110 can log in the relevant interface 120 for executing the target service. The terminal device 110 and the server 130 can communicate with each other through a communication network.

In an alternative embodiment, the communication network is a wired network or a wireless network.

In this embodiment, the terminal device 110 is an electronic device used by a user, and the electronic device may be a computer device having a certain computing capability and running instant messaging software and a website or social contact software and a website, such as a personal computer, a mobile phone, a tablet computer, a notebook, an e-book reader, and the like. Each terminal device 110 is connected to a server 130 through a wireless network, and the server 130 is a server or a server cluster or a cloud computing center formed by a plurality of servers, or is a virtualization platform.

In this embodiment of the application, the terminal device 110 is configured to obtain behavior data of a target object generated for a high-frequency application in a preset time period, and send the behavior data to the server 130, the server 130 is configured to obtain behavior habit characteristics of the target object based on the behavior data, and further determine, according to the behavior habit characteristics of the target object, idle time during which the target object can use the low-frequency application, that is, time during which the target object tends to use the low-frequency application, by using a prediction model, and further trigger the low-frequency application to send a push message to the target object in the idle time, so as to improve the use duration, click rate, and the like of the low-frequency application by a user.

Referring to fig. 2, an implementation flow chart of a message pushing method provided in the embodiment of the present application is shown, and a specific implementation flow of the method is as follows:

s21: acquiring behavior data of a target object generated for high-frequency application in a preset time period, wherein the high-frequency application is an application with the use frequency greater than a preset frequency threshold in the preset time period;

in the embodiment of the application, behavior data generated when the user uses the high-frequency application can be acquired according to the user behavior log and the like.

When the high-frequency application and the low-frequency application are distinguished according to the use frequency of the application, for example, 10 applications, namely application 1 to application 10, are installed on a mobile terminal of a target object, where the use frequency is higher than a preset frequency threshold, application 1 and application 2 exist, and the use frequencies of the other applications are not higher than the preset frequency threshold, so that the high-frequency application includes application 1 and application 2, and the low-frequency application includes application 3 to application 10, and therefore, behavior data of a user when using application 1 and application 2 need to be collected.

S22: acquiring behavior habit characteristics of the target object based on the behavior data;

in an optional implementation manner, when behavior habit characteristics of a target object are acquired according to behavior data, firstly, structured data in the behavior data need to be constructed into at least one structured data event; and then, feature extraction is carried out according to at least one structured data event to obtain the behavior habit features of the target object.

The structured data events are used for representing events which occur when the target object carries out various operations.

For example, when a user uses high frequency APP, a large amount of behavior occurs. By recording various behavior data of the user and well organizing, each piece of reported structured data is abstracted into an event which occurs to the user, namely a structured data event. Such as a user entering an APP event, a user consuming an event within an APP, a screen switch event, etc.

In the embodiment of the application, for each structured data event, various attributes related to the event can be attached, specifically including a specific attribute for the event and a common attribute which is contained for all events. For example, red point data of the message list, which is a public attribute reported by all events; for another example, for a screen switching event, only the event has the attribute of whether the screen is on or off, i.e. the specific attribute of the screen switching event.

In the embodiment of the application, the client can execute the structured data when the structured data in the behavior data is constructed into at least one structured data event, then the client sends the structured data event to the back-end server, and the server performs feature extraction according to the structured data event to obtain the behavior habit features of the target object.

Specifically, after receiving each reported structured data event, the back-end server calculates some additional user behavior habit characteristics based on the structured data event. Meanwhile, after receiving each structured data event, the back-end server can perform persistent storage, so that operations such as recovery, search and verification of user data can be guaranteed in the future.

In the embodiment of the present application, the behavior habit characteristics specifically include at least some or all of the following:

long-term liveness features, short-term features, immediate features.

These types of behavior habit features are described in detail below:

the long-term activity characteristic refers to the activity of various operations of a user within the last several days, months or even years, and the specific extraction mode is as follows:

the method comprises the steps of obtaining the activity of a target object in each unit time of a first statistical time interval according to at least one structured data event, taking the average value of the activity as the long-term activity characteristic of the target object, and taking the determined long-term activity characteristic as the behavior habit characteristic of the target object, wherein the first statistical time interval is within a preset time interval range.

Wherein, the activity of each operation can be: the activity of the user using the APP, the activity of the user having consumption within the APP, and so on. Furthermore, liveness can be measured in a number of ways, such as the number of days that the user has been active recently, the number of times that the user has been active for the last few days, and so forth.

Taking the number of active times in the last 5 days as an example, assuming that the first statistical period is the last 5 days in the preset period, and the unit time is one day, the main calculation process is as follows:

calculating the activity of each operation of the user in each day by taking the day as a unit, and then accumulating and combining the data of the 5 days. During actual use, the average value of each operation activity of the user in the 5 days is calculated to represent the activity, and the influence of abnormal operation data of the user in a certain day on activity estimation can be greatly reduced in the mode.

It should be noted that the manner of calculating the average value recited in the embodiments of the present application is only an example, and the long-term activity characteristic of the user may also be determined by calculating other statistical manners such as a maximum value, a weighted average value, and a variance.

Second, short-term characteristics refer to the number of times of statistics of various operations of a user in the last several seconds and several minutes, and the specific extraction method is as follows:

and acquiring the times of the target object for implementing various operations in a second statistical time period according to at least one structured data event, taking the acquired times as short-term characteristics of the target object, and taking the determined short-term characteristics as behavior habit characteristics of the target object, wherein the second statistical time period is within a preset time period range, the duration of the second statistical time period is less than that of the first statistical time period, namely the second statistical time period is shorter, and the first statistical time period is longer.

For example, for the chat application a, if the friend trends and the hot news belong to the function modules in the chat application a, the frequency that the user has seen the friend trends in the last 30 seconds can be counted, and the second counting time period is 30 seconds; or counting the number of times the user has seen hot news within the last 10 seconds, where the second counting period is 10 seconds.

It should be noted that, in the embodiment of the present application, when determining the short-term characteristics of the user, a user short-term behavior operation list for storage may be set, and a threshold for storing the list at most is set, so as to avoid that the storage list is too long, which makes backend storage insufficient. In this way, by recording the user short-term behavior operation list, the operation behavior statistics in the latest period of time can be calculated, for example, as shown in table 1, assuming that table 1 can store m records at most, where m is a positive integer.

TABLE 1 user short-term behavior action List in second statistical time period

Operation of	Time (within the second statistical period)
		Click-through application 1	T1
Click-through application 2	T2
		Consumption in application 1	T3
Consumption in application 2	T4
		Consumption in application 2	T5
…	…
		Click-through application 1	Tm

As can be seen from table 1, the operation performed by the user at time T1 is: clicking to enter the application 1, the operation performed by the user at time T2 is: click into application 2, and so on. And the operation and the implementation times of the operation of the user in the second statistical time period can be conveniently counted through the table.

And thirdly, instant characteristics, namely instant state information of the APP when the user operates at the current moment, can be extracted specifically according to the following mode:

and acquiring instant state information of high-frequency application at the current moment according to at least one structured data event, taking the instant state information as the instant characteristic of the target object, and taking the determined instant characteristic as the behavior habit characteristic of the target object.

For example, for the APP of chat application a, statistics may be performed on the number of red dots of the user's current message list, the current news about the hotspots, and other instant status information, where the number of red dots of the user's current message list, and the like, may be determined according to the common attribute or the specific attribute of the structured data event. During actual calculation, all current states of the user can be captured by the client and reported to the back-end server, and the back-end server only needs to extract the features.

In the above embodiment, based on the number of times of implementing each operation of the user, the activity, or the application instant state information, etc. as the behavior habit characteristics of the user, the habit of the user using the application can be effectively described, so as to predict the idle time of the user.

In an alternative embodiment, the behavior habit features of the target object may further include one or more of feedback features and static features.

Wherein, static characteristics refer to characteristics which can not change in a short period, and are determined by the following method:

and acquiring identity information input when the target object registers the application according to at least one structured data event, and taking the identity information as the static characteristic of the target object.

For example, the sex, age, etc. of the user, and the features of the features that will not change in a short period can be obtained by the data filled in by the user during APP registration, that is, the identity information input by the user.

In the embodiment of the present application, after the static features of the user, such as age and gender, are obtained according to the identity information input by the user, the correctness check and the abnormal value processing of the static features are also required.

For example, the age filled by the user when registering the application is 200, and this information about the age is inaccurate, i.e., an abnormal value. In this case, an age threshold may be set, and if the age threshold is 80, an abnormal value may be filtered out based on the age threshold, and the information with the age greater than 80, for example, 200 greater than 80, may be directly removed, so that the abnormal value may be directly deleted.

In the above embodiment, through correctness checking and abnormal value processing, incorrect static features can be effectively removed, so as to ensure the accuracy of static feature extraction.

In the embodiment of the present application, the feedback feature refers to a user feedback feature constructed by whether a user clicks a push message after the message is pushed to the user. The user feedback characteristics mainly comprise the exposure times, the click times and the click rate of the messages pushed by the user. The feedback characteristic may specifically be determined by:

after the push message is displayed on the target object, feedback information used for indicating whether the target object clicks the push message or not is obtained according to at least one structured data event, and the feedback information is used as the feedback characteristic of the target object.

For example, a user push message feedback list is stored in the back end, and each item of the list records the time of the push message, whether the user clicks after the push message, and the like. Based on this list, the number of exposures, clicks, and click rate for each user can be conveniently calculated. In addition, since the storage space on the backend line is limited, the size of the push list needs to be limited, and the latest records are reserved, for example, as shown in table 2, it is assumed that table 2 can store n records at most, where n is a positive integer.

Table 2 user push message feedback list

Number of times of message push	Push message time	Whether to click to push a message
			1 st time	t1	Is that
2 nd time	t2	Whether or not
			3 rd time	t3	Is that
…	…	…
			…	…	…
The nth time	tn	Is that

The records in table 2 include n records of pushing messages to a certain user, including the time of pushing messages specifically, and the feedback of the user after pushing messages to the user, and according to table 2, the exposure times, click rate, and the like of the user can be calculated as feedback characteristics of the user.

In the embodiment, the use habits of the user can be fully associated with the user based on the static characteristics of the user, and the accuracy rate of the prediction of the click probability can be effectively improved based on the feedback information.

It should be noted that the several types of user behavior habit features listed in the embodiment of the present application are only examples, and actually any behavior habit feature determined according to the behavior data of the user is applicable to the embodiment of the present application, and is not limited to the several types of features listed above, and is not specifically limited herein.

S23: determining idle time of a target object capable of using low-frequency application according to the behavior habit characteristics, wherein the low-frequency application is an application of which the use frequency is not more than a preset frequency threshold value in a preset time period;

in an alternative embodiment, the idle time that the user can use the low-frequency application can be determined through a machine learning model, and the specific implementation process is as follows:

inputting the behavior habit characteristics into a trained prediction model, and acquiring the click probability output by the prediction model, wherein the trained prediction model is obtained by training a training sample labeled with the click probability, the training sample contains the behavior habit characteristics, and the click probability is used for expressing the probability that a target object clicks the push message after the push message is displayed to the target object; and determining the idle time according to the click probability.

Wherein, the prediction model can be obtained by training in the following way:

acquiring behavior habit characteristics corresponding to the training samples; inputting the behavior habit characteristics of the training samples into the untrained prediction model to obtain the click probability of the training samples output by the untrained prediction model; and continuously optimizing the parameters of the untrained prediction model until the click probability output by the untrained prediction model and the click probability marked on the training sample are within an allowable difference range, thereby obtaining the trained prediction model.

In the embodiment of the application, the training sample is actually a data set comprising a plurality of training samples, and when a prediction model is trained by using more training samples, the trained prediction model has stronger applicability and the model prediction result is more accurate.

Specifically, the prediction model is obtained by predicting whether a user clicks after pushing a message to the low-frequency APP, so as to model the idle time judgment problem as a binary problem. Therefore, data training and prediction can be performed using models such as DNN (Deep Neural Network), XGBoost (eXtreme Gradient Boosting), LR (Logistic Regression), and the like.

In the embodiment of the present application, the prediction model may be represented as f (X) ═ y, where X represents a feature vector and is obtained by vectorizing and representing the behavior habit features of the user, and y represents the click probability predicted by the model.

And assuming that the preset probability threshold is b, after collecting various behavior habit features of the user, constructing a feature vector X and inputting the feature vector X into the prediction model to obtain the click probability y output by the prediction model. In the embodiment of the present application, when determining whether the target object is in the idle time capable of using the low-frequency application according to the click probability, the method may specifically be divided into the following two determination manners:

the method comprises the steps that in a first judgment mode, if the click probability is larger than a preset probability threshold, the target object is determined to be in idle time; and if the click probability is not greater than a preset probability threshold, determining that the target object is in idle time.

That is, only y > b is needed to determine that the user is in idle time at the current moment, which may trigger the low frequency application to send the push message to the user. And when y is less than or equal to b, determining that the current time of the user is not in idle time, and sending a push message to the user without triggering the low-frequency application.

In the embodiment, all users are fixed thresholds, namely preset probability thresholds, so that the idle time of the users can be simply and efficiently predicted, and the click rate and the click UV of the users in low-frequency application are improved.

It should be noted that, in some cases, the click probability predicted by many users through the prediction model is very small and smaller than the preset probability threshold, so that it is likely that a push message cannot be sent to the users when determining whether the users are in idle time only according to the click probability, but after actually sending the push message to the users, the users are likely to click the push message, and therefore, in order to ensure the coverage of the users, in this case, the determination may be performed by using the method proposed in the determination method two.

Judging the mode II, if the sequence of the click probability in the historical click probability corresponding to the target object is within the preset sequence range, determining that the target object is in idle time; otherwise, determining that the target object is not in the idle time. The historical click probability is the click probability obtained by predicting a plurality of times before the current prediction by the user.

In the embodiment of the application, by recording the historical click probability of the user, if the position of the click probability predicted by the user in the current scene on the statistical distribution of the historical click probability is earlier, the user can be determined to be currently in the idle time, and a message can be pushed to the user. For example, the messages are sorted in a descending order (from large to small), the click probability of the current time of the user in the historical click probability of the user is sorted second, and if the preset order range is the first three of the sorting, the user can be determined to be in the space time at the moment, and the messages can be pushed.

In the embodiment, the personalized threshold values for different users are realized, and the user coverage rate of message pushing can be effectively improved.

In an optional implementation manner, if the ranking of the click probabilities in the historical click probabilities corresponding to the target object is within the preset sequence range, before determining that the current time of the target object is in the idle time, it is further necessary to determine whether the priority of the scene corresponding to the current time is the highest in the preset time duration according to the expected times corresponding to each scene, that is, no scene with a higher priority than the current scene will occur at the future time in the preset time duration, and if the priority of the scene is also higher than the current scene, it is determined that the user is not in the idle time, no message is pushed, and the triggering of a better scene is waited; otherwise, determining that the user is not in idle time, and pushing the message.

The priority sequence is determined according to the click probability of the target object in each scene, and the higher the click probability is, the higher the priority is. That is, the user's priority under each scenario is determined according to the prediction model f (x) ═ y. Assume prioritization is scene 1> scene 2> scene 3> scene 4.

In the embodiment of the present application, the expected times are determined according to the historical operation distribution of the target object, and are used to indicate the expected occurrence times, that is, the possible occurrence times of each scene in the preset time duration, and may be used to determine whether the scene will occur at a future time within the preset time duration.

Wherein the expected times of each scene may be averaged according to the number of times of occurrence of each scene in a period of time (e.g., the last 5 days) before the current time.

Taking a scene when the current time is predicted as a scene 2 as an example, assuming that the preset time duration is one day, 3 times of the scene 1 have occurred before the current time of today, and 4 times of the expected times of the scene 1, it indicates that the scene 1 with higher priority than the scene 2 may occur again at the future time of today, and thus it may be determined that the user is not in the idle time at this time, but the low-frequency application is triggered to push a message to the user when the next scene 1 occurs.

Taking a scene when the current time is predicted as a scene 1 as an example, the priority of the scene 1 is the highest at the time, and the click probability of the prediction is within the preset sequence range, so that the user can be determined to be in idle time at the time, and the low-frequency application is triggered to push a message to the user.

It should be noted that, in the embodiment of the present application, each scenario may be defined according to a structured data event, and other manners may also be adopted, for example, when a user opens a hot news module in the chat application a, the scenario may be defined as one scenario, so that when the user opens the hot news module, whether the user is idle or not may be predicted through the prediction model, and if the user is idle, the low frequency application may be triggered to send a push message to the user, for example, the low frequency application of some news categories is triggered to send a push message to the user, as shown in fig. 3, S30 in the figure is the push message sent by the news application B.

In the embodiment, the scene pushing is effectively considered, and each user can be ensured to push the message.

S24: and triggering the low-frequency application to send the push message in idle time.

In this embodiment of the present application, the low frequency application may be triggered to send the push message in the following ways:

and in the first triggering mode, the idle time is sent to the low-frequency application, and the low-frequency application displays the push message to the target object according to the idle time.

And in the triggering mode II, a triggering signal is sent to the low-frequency application in idle time, and the low-frequency application is triggered to display the push message to the target object.

And a triggering mode III is used for sending the push message to be sent to the low-frequency application in idle time, and the low-frequency application displays the push message to the target object.

It should be noted that the above triggering manners are only examples, and any manner of triggering the low-frequency application to send the push message in the idle time is applicable to the embodiment of the present application.

Referring to fig. 4A, a schematic diagram of a method for pushing a message when a prediction model provided in the embodiment of the present application is deployed on a server side is shown, where a client collects operation data generated by a user in a high-level application and a low-frequency application, and sends the operation data to a back-end server, the back-end server determines idle time of the user for the low-frequency application, and pushes a message to the low-frequency application in the idle time, where the pushed message may be a screen locking message or hot news, and the like.

In addition, in this embodiment of the present application, the prediction model may also be deployed on the client side, where the client determines the idle time of the user and triggers the low frequency application to send a push message to the user, as shown in fig. 4B. Specifically, the client sends the idle time to the low frequency application server, and the low frequency application server triggers the low frequency application to send the push message, where the manner when the low frequency application server triggers the low frequency application to send the push message may be any one of the above triggering manners, and details are not repeated here.

In the embodiment, the client executes some calculation logics of prediction model judgment, so that background pressure can be relieved, and high performance of distributed calculation is fully utilized.

Fig. 5 is an interaction timing chart of a message pushing method according to an embodiment of the present application.

The specific implementation flow of the method is as follows:

s51: the method comprises the steps that a high-frequency application client side obtains a large amount of behavior data generated when a user uses a high-frequency application in a preset time period;

s52: the high-frequency application client side constructs structured data in the behavior data into structured data events;

s53: the high-frequency application client sends the structured data event to a server;

s54: the server extracts features according to the structured data events to obtain behavior habit features of the target object;

s55: the server determines the idle time of the target object capable of using the low-frequency application according to the behavior habit characteristics;

s56: the server sends the push message to the low-frequency application client in idle time;

s57: and the low-frequency application client displays the push message to the user.

In the embodiment of the application, the user behavior portrait is established through the behavior data of the user on the high-frequency application, the idle time of the user is explored, and the low-frequency application is triggered to push messages to the user in the idle time, so that the click rate of the user on the low-frequency application is improved, and the use duration and the activity of the user on the low-frequency application are improved.

As shown in fig. 6, which is a schematic structural diagram of a message pushing apparatus 600 provided in an embodiment of the present application, the message pushing apparatus may include:

a data obtaining unit 601, configured to obtain behavior data of a target object generated for a high-frequency application in a preset time period, where the high-frequency application is an application whose usage frequency is greater than a preset frequency threshold in the preset time period;

a characteristic obtaining unit 602, configured to obtain behavior habit characteristics of a target object based on behavior data;

an idle time determining unit 603, configured to determine, according to the behavior habit characteristics, an idle time during which the target object can use the low-frequency application;

the triggering unit 604 is configured to trigger, in an idle time, a low-frequency application to send a push message, where the low-frequency application is an application whose usage frequency is not greater than a preset frequency threshold within a preset time period.

In an optional implementation manner, the idle time determining unit 603 is specifically configured to:

inputting the behavior habit characteristics into a trained prediction model, and acquiring the click probability output by the prediction model, wherein the trained prediction model is obtained by training a training sample labeled with the click probability, the training sample contains the behavior habit characteristics, and the click probability is used for expressing the probability that a target object clicks the push message after the push message is displayed to the target object;

and determining the idle time according to the click probability.

if the click probability is larger than a preset probability threshold, determining that the target object is in idle time; or

And if the sequence of the click probability in the historical click probability corresponding to the target object is within the preset sequence range, determining that the target object is in idle time.

In an optional implementation manner, if the ranking of the click probability in the historical click probability corresponding to the target object is within a preset order range, the feature obtaining unit 602 is further configured to, before determining that the target object is in the idle time:

determining the highest priority of the scene corresponding to the current time in the preset time length according to the expected times corresponding to each scene, wherein the expected times are determined according to the historical operation distribution of the target object and are used for representing the expected times of each scene in the preset time length, the priority sequence is determined according to the click probability of the target object in each scene, and the higher the click probability is, the higher the priority is.

In an alternative embodiment, the feature obtaining unit 602 is specifically configured to determine the behavior habit feature of the target object through any one or more of the following manners:

acquiring the activity of each operation carried out by the target object in each unit time of a first statistical time interval according to at least one structured data event, taking the average value of the activity as the long-term activity characteristic of the target object, and determining the behavior habit characteristic of the target object according to the long-term activity characteristic, wherein the first statistical time interval is within a preset time interval range;

acquiring the times of carrying out various operations of the target object in a second statistical time period according to at least one structured data event, taking the acquired times as short-term characteristics of the target object, and determining behavior habit characteristics of the target object according to the short-term characteristics, wherein the second statistical time period is within a preset time period range, and the duration of the second statistical time period is less than that of the first statistical time period;

and acquiring instant state information of the high-frequency application at the current moment according to at least one structured data event, taking the instant state information as the instant characteristic of the target object, and determining the behavior habit characteristic of the target object according to the instant characteristic.

In an alternative embodiment, the behavioral habit feature further comprises a feedback feature; the feature obtaining unit 602 is further configured to:

after the push message is displayed on the target object, feedback information used for indicating whether the target object clicks the push message or not is obtained according to at least one structured data event, and the feedback information is used as the feedback characteristic of the target object; and/or

The behavioral habit characteristics also include static characteristics; the feature obtaining unit 602 is further configured to:

In an alternative embodiment, the apparatus further comprises a model training unit 605;

the model training unit 605 is used for training a trained prediction model by:

acquiring behavior habit characteristics corresponding to the training samples;

inputting the behavior habit characteristics of the training samples into the untrained prediction model to obtain the click probability of the training samples output by the untrained prediction model;

and continuously optimizing the parameters of the untrained prediction model until the click probability output by the untrained prediction model and the click probability marked on the training sample are within an allowable difference range, thereby obtaining the trained prediction model.

For convenience of description, the above parts are separately described as modules (or units) according to functional division. Of course, the functionality of the various modules (or units) may be implemented in the same one or more pieces of software or hardware when implementing the present application.

After introducing the message pushing method and apparatus according to the exemplary embodiment of the present application, an electronic device according to an exemplary embodiment of the present application is introduced next.

As will be appreciated by one skilled in the art, aspects of the present application may be embodied as a system, method or program product. Accordingly, various aspects of the present application may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.) or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" system.

Fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. As shown in fig. 7, an electronic device 700 may include at least one processor 701, and at least one memory 702. The memory 702 stores therein program codes, which, when executed by the processor 701, cause the processor 701 to execute the steps in the message pushing method according to various exemplary embodiments of the present application described above in the present specification. For example, the processor 701 may perform the steps as shown in fig. 2.

In some possible embodiments, a computing device according to the present application may include at least one processing unit, and at least one memory unit. Wherein the storage unit stores program code which, when executed by the processing unit, causes the processing unit to perform the steps of the service invocation method according to various exemplary embodiments of the present application described above in the present specification. For example, the processing unit may perform the steps as shown in fig. 2.

The computing device 80 according to this embodiment of the present application is described below with reference to fig. 8. The computing device 80 of fig. 8 is only one example and should not impose any limitations on the functionality or scope of use of embodiments of the application.

As shown in fig. 8, computing device 80 is embodied in the form of a general purpose computing device. Components of computing device 80 may include, but are not limited to: the at least one processing unit 81, the at least one memory unit 82, and a bus 83 connecting the various system components (including the memory unit 82 and the processing unit 81).

Bus 83 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a processor, or a local bus using any of a variety of bus architectures.

The storage unit 82 may include readable media in the form of volatile memory, such as a Random Access Memory (RAM)821 and/or a cache storage unit 822, and may further include a Read Only Memory (ROM) 823.

The storage unit 82 may also include a program/utility 825 having a set (at least one) of program modules 824, such program modules 824 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

The computing device 80 may also communicate with one or more external devices 84 (e.g., keyboard, pointing device, etc.), may also communicate with one or more devices that enable a user to interact with the computing device 80, and/or may communicate with any devices (e.g., router, modem, etc.) that enable the computing device 80 to communicate with one or more other computing devices. Such communication may be through input/output (I/O) interfaces 85. Also, computing device 80 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet) through network adapter 86. As shown, network adapter 86 communicates with other modules for computing device 80 over bus 83. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with computing device 80, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

In some possible embodiments, the aspects of the message pushing method provided in the present application may also be implemented in the form of a program product, which includes program code for causing a computer device to perform the steps in the message pushing method according to various exemplary embodiments of the present application described above in this specification when the program product is run on the computer device, for example, the computer device may perform the steps as shown in fig. 2.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The program product for message pushing of the embodiments of the present application may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a computing device. However, the program product of the present application is not limited thereto, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with a command execution system, apparatus, or device.

A readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with a command execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user computing device, partly on the user equipment, as a stand-alone software package, partly on the user computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).

While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.

Claims

1. A message pushing method, the method comprising:

acquiring historical behavior data of a target object generated for high-frequency application in a preset historical time period, wherein the high-frequency application is an application of which the use frequency is greater than a preset frequency threshold in the preset historical time period;

acquiring behavior habit characteristics of the target object based on the historical behavior data, wherein the behavior habit characteristics at least comprise part or all of the following: long-term liveness characteristics, short-term characteristics, instantaneous characteristics;

determining the idle time of the target object not using the high-frequency application according to the behavior habit characteristics;

and in the idle time, triggering a low-frequency application to send a push message, wherein the low-frequency application is an application of which the use frequency is not greater than the preset frequency threshold value in the preset historical time period, and the push message comprises a screen locking message or hot news.

2. The method of claim 1, wherein the determining an idle time during which the target object does not use the high frequency application based on the behavioral habit feature comprises:

inputting the behavior habit characteristics into a trained prediction model, and acquiring a click probability output by the prediction model, wherein the trained prediction model is obtained by training a training sample labeled with the click probability, the training sample contains the behavior habit characteristics, and the click probability is used for expressing the probability that a target object clicks a push message after the target object is displayed with the push message;

and determining the idle time according to the click probability.

3. The method of claim 2, wherein said determining the idle time based on the click probability comprises:

And if the sequence of the click probability in the historical click probability corresponding to the target object is within a preset sequence range, determining that the target object is in idle time.

4. The method of claim 3, wherein if the click probability is ranked within a preset order range in the historical click probability corresponding to the target object, before determining that the target object is in idle time, further comprising:

determining the highest priority of the scenes corresponding to the current time within a preset time length according to the expected times corresponding to the scenes, wherein the expected times are determined according to the historical operation distribution of the target object and are used for representing the expected times of the scenes within the preset time length, the priority sequence is determined according to the click probability of the target object under the scenes, and the higher the click probability is, the higher the priority is.

5. The method according to any one of claims 1 to 4, wherein the obtaining of the behavior habit characteristics of the target object based on the historical behavior data comprises:

constructing the structured data in the historical behavior data into at least one structured data event, wherein the structured data event is used for representing an event which occurs when the target object carries out various operations;

and performing feature extraction according to the at least one structured data event to obtain the behavior habit features of the target object.

6. The method of claim 5, wherein the behavioral habit features of the target object are determined by any one or more of the following means:

acquiring the activity of the target object in each unit time of a first statistical time period according to the at least one structured data event, taking the average value of the activity as the long-term activity characteristic of the target object, and determining the behavior habit characteristic of the target object according to the long-term activity characteristic, wherein the first statistical time period is within the range of the preset historical time period;

acquiring the times of the target object for implementing each operation in a second statistical time period according to the at least one structured data event, taking the acquired times as short-term characteristics of the target object, and determining behavior habit characteristics of the target object according to the short-term characteristics, wherein the second statistical time period is within the preset historical time period range, and the duration of the second statistical time period is less than that of the first statistical time period;

7. The method of claim 6, wherein the behavioral habit features further include a feedback feature;

performing feature extraction on the at least one structured data event to generate behavior habit features of the target object, including:

The behavioral habit features also include static features;

the performing feature extraction on the at least one structured data event to generate behavior habit features of the target object includes:

8. The method of any of claims 2 to 4, wherein the trained predictive model is trained by:

acquiring behavior habit characteristics corresponding to the training samples;

inputting the behavior habit characteristics of the training sample into an untrained prediction model to obtain the click probability of the training sample output by the untrained prediction model;

and continuously optimizing the parameters of the untrained prediction model until the click probability output by the untrained prediction model and the click probability marked on the training sample are within an allowed difference range, so as to obtain the trained prediction model.

9. A message push apparatus, comprising:

the data acquisition unit is used for acquiring historical behavior data of a target object generated for high-frequency application in a preset historical time period, wherein the high-frequency application is an application of which the use frequency is greater than a preset frequency threshold in the preset historical time period;

a characteristic obtaining unit, configured to obtain behavior habit characteristics of the target object based on the historical behavior data, where the behavior habit characteristics at least include part or all of the following: long-term liveness characteristics, short-term characteristics, instantaneous characteristics;

the idle time determining unit is used for determining the idle time of the target object not using the high-frequency application according to the behavior habit characteristics;

and the triggering unit is used for triggering a low-frequency application to send a push message in the idle time, wherein the low-frequency application is an application of which the use frequency is not more than the preset frequency threshold value in the preset historical time period, and the push message comprises a screen locking message or hot news.

10. The apparatus of claim 9, wherein the idle time determination unit is specifically configured to:

and determining the idle time according to the click probability.

11. The apparatus of claim 10, wherein the idle time determination unit is specifically configured to:

12. The apparatus of claim 11, wherein if the click probability is ranked within a preset order range in the historical click probability corresponding to the target object, the feature obtaining unit is further configured to, before determining that the target object is in idle time:

13. The apparatus according to any one of claims 10 to 12, further comprising a model training unit;

the model training unit is used for obtaining the trained prediction model through the following training:

acquiring behavior habit characteristics corresponding to the training samples;

14. An electronic device, comprising a processor and a memory, wherein the memory stores program code which, when executed by the processor, causes the processor to perform the steps of the method of any of claims 1 to 8.

15. Computer-readable storage medium, characterized in that it comprises program code for causing an electronic device to carry out the steps of the method according to any one of claims 1 to 8, when said program product is run on said electronic device.