CN117591320B

CN117591320B - Optimized pushing method and system based on multi-channel message

Info

Publication number: CN117591320B
Application number: CN202410064177.1A
Authority: CN
Inventors: 王心安
Original assignee: Zhilin Technology Co ltd; Beijing Zhilin Technology Co ltd
Current assignee: Zhilin Technology Co ltd; Beijing Zhilin Technology Co ltd
Priority date: 2024-01-17
Filing date: 2024-01-17
Publication date: 2024-04-26
Anticipated expiration: 2044-01-17
Also published as: CN117591320A

Abstract

The invention relates to the technical field of data processing, and discloses an optimized pushing method and system based on multi-channel information, which are used for improving the accuracy of optimized pushing based on the multi-channel information. Comprising the following steps: obtaining a message to be pushed, and carrying out pushing channel analysis on the message to be pushed to obtain an initial pushing channel; collecting current network state data of a target user in real time, and analyzing push time of an initial push channel to obtain initial push time; the method comprises the steps of pushing a message to a target user, and collecting feedback data of the target user in real time to obtain real-time feedback data; and carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, carrying out pushing index correction based on the pushing fitness score to obtain a corrected pushing index, and carrying out strategy optimization on an initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy.

Description

Optimized pushing method and system based on multi-channel message

Technical Field

The invention relates to the technical field of data processing, in particular to an optimized pushing method and system based on multi-channel information.

Background

With the continuous development of the information age, multi-channel message pushing becomes one of the main modes of communication and information transfer between enterprises and users. However, conventional message pushing methods often lack personalization and intelligence, resulting in difficulty for users to screen and obtain truly interesting content in an information overload environment. In order to solve the problem, a message pushing method based on multiple channels has been developed, and the purpose is to construct an intelligent pushing strategy by deeply mining user behavior and preference data, so as to improve the effect of message pushing and user satisfaction.

Although there are many technical solutions for multi-channel message pushing, there are some disadvantages. First, the integration capability of part of the system to the data source is limited in the user portrait construction stage, and it is difficult to comprehensively and accurately grasp the multidimensional information of the user. Second, existing push strategies often lack sufficient personalization to accommodate the user's varying behaviors and preferences. In addition, the processing mode of the real-time feedback data needs to be further optimized to adjust the pushing strategy more intelligently. Therefore, how to integrate data better and improve individuation and intelligence of pushing is still a problem to be solved in the current multi-channel message pushing technology.

Disclosure of Invention

In view of the above, the embodiment of the invention provides an optimized pushing method and system based on multi-channel messages, which are used for improving the accuracy of optimized pushing based on the multi-channel messages.

The invention provides an optimized pushing method based on multi-channel information, which comprises the following steps: collecting historical message data of a target user from a plurality of preset data sources to obtain target message data of each data source, and constructing a user portrait of the target user according to the target message data of each data source to obtain a target user portrait; performing priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source; carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set; obtaining a message to be pushed, and carrying out push channel analysis on the message to be pushed through the initial message push strategy to obtain an initial push channel; collecting current network state data of the target user in real time, and analyzing the pushing time of the initial pushing channel based on the current network state data to obtain initial pushing time; message pushing is carried out on the target user based on the initial pushing channel and the initial pushing time, and feedback data of the target user are collected in real time to obtain real-time feedback data; and carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, carrying out pushing index correction based on the pushing fitness score to obtain a corrected pushing index, and carrying out strategy optimization on the initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy.

In the invention, the step of collecting the historical message data of the target user from a plurality of preset data sources to obtain the target message data of each data source, and constructing the user portrait of the target user according to the target message data of each data source to obtain the target user portrait comprises the following steps: analyzing the data acquisition interfaces of the data sources to obtain the data acquisition interface of each data source; carrying out identity information verification on the target user based on the data acquisition interface of each data source to obtain a verification result; when the verification result is that verification is passed, acquiring historical message data of the target user from a plurality of data sources to obtain initial message data of each data source; performing stem extraction on the initial message data of each data source respectively to obtain a stem data set of the initial message data of each data source; generating target message data for each of the data sources based on a stem data set of the initial message data for each of the data sources; and constructing the user portrait of the target user according to the target message data of each data source to obtain the target user portrait.

In the present invention, the step of constructing the user portrait for the target user according to the target message data of each data source to obtain the target user portrait includes: analyzing the content of interest of the user to the target user according to the target message data of each data source to obtain the content of interest of the user; analyzing the active time of the target user according to the target message data of each data source to obtain the user active time of each data source; based on the user active time of each data source, carrying out active time sequencing on a plurality of data sources to obtain a time sequencing result; based on the time sequencing result, respectively carrying out weight data matching on each data source to obtain weight data of each data source; and constructing a user portrait of the target user through the user interested content and the user active time of each data source based on the weight data of each data source, so as to obtain the target user portrait.

In the present invention, the step of analyzing the message type of the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set includes: respectively carrying out message structure analysis on the target message data of each data source to obtain a plurality of message structures; based on a plurality of message structures, respectively splitting message fields of target message data of each data source to obtain a message field set corresponding to each data source; message type analysis is carried out on the target message data of each data source to obtain a message type set; analyzing the type pushing priority of the message type set to obtain type pushing priority data; carrying out push frequency analysis on each message type in the message type set to obtain push frequency data of each message type; carrying out pushing period analysis on each message type in the message type set to obtain pushing period data of each message type; based on the type push priority data, push frequency data of each message type and push period data of each message type, and constructing an initial message push strategy according to the priority sequence of each data source.

In the invention, the step of obtaining the message to be pushed, and analyzing the pushing channel of the message to be pushed through the initial message pushing strategy to obtain the initial pushing channel comprises the following steps: analyzing the message type of the message to be pushed to obtain a target message type; analyzing the message importance of the message to be pushed based on the target message type to obtain the importance of the target message; carrying out matching index analysis on the target message type through the target user image to obtain an image matching index; extracting user associated information from the importance of the target message based on the portrait matching index to obtain target user associated information; and carrying out push channel analysis on the target user associated information through the initial message push strategy to obtain the initial push channel.

In the invention, the step of pushing the message to the target user based on the initial pushing channel and the initial pushing time and collecting the feedback data of the target user in real time to obtain the real-time feedback data comprises the following steps: carrying out identifier analysis on the target user to obtain unique identifier data of the target user; analyzing the pushing trigger point of the initial pushing channel and the initial pushing time through the unique identifier data to obtain a target pushing trigger point; based on the target pushing trigger point, pushing the message to the target user; collecting click events of the target user in real time to obtain user click events; collecting reading time data of the target user to obtain the reading time data; collecting feedback content of the target user to obtain target feedback content; based on a preset time interval, carrying out click rate analysis on the user click event to obtain a target click rate; performing user behavior report analysis on the target user based on the reading time data and the target feedback content to obtain a target behavior report; and merging the target click rate and the target behavior report into the real-time feedback data.

In the invention, the step of calculating the push fitness score of the real-time feedback data to obtain the push fitness score, correcting the push index based on the push fitness score to obtain a corrected push index, and performing policy optimization on the initial message push policy by the corrected push index to obtain a target message push policy comprises the following steps: performing weight matching on the target click rate to obtain first weight data; performing user preference analysis on the user behavior report to obtain target user preference data; user response rate analysis is carried out on the target user preference data to obtain target user response rate; performing weight matching on the target user response rate to obtain second weight data; based on the first weight data and the second weight data, carrying out weighted summation on the target click rate and the target user response rate to obtain the push fitness score; analyzing the score range of the push fitness score to obtain a target score range; performing index correction parameter analysis based on the target score range to obtain the correction pushing index; and carrying out strategy optimization on the initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy.

The invention also provides an optimized pushing system based on the multi-channel information, which comprises the following steps:

The acquisition module is used for acquiring historical message data of a target user from a plurality of preset data sources to obtain target message data of each data source, and constructing a user portrait of the target user according to the target message data of each data source to obtain a target user portrait;

The first analysis module is used for carrying out priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source;

The construction module is used for carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set;

The second analysis module is used for acquiring the message to be pushed, and carrying out push channel analysis on the message to be pushed through the initial message push strategy to obtain an initial push channel;

The third analysis module is used for collecting current network state data of the target user in real time, and analyzing the pushing time of the initial pushing channel based on the current network state data to obtain initial pushing time;

The pushing module is used for pushing the message to the target user based on the initial pushing channel and the initial pushing time, and collecting feedback data of the target user in real time to obtain real-time feedback data;

The correction module is used for carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, correcting pushing indexes based on the pushing fitness score to obtain corrected pushing indexes, and carrying out strategy optimization on the initial message pushing strategy through the corrected pushing indexes to obtain a target message pushing strategy.

According to the technical scheme provided by the invention, historical message data of a target user are acquired from a plurality of preset data sources to obtain target message data of each data source, and user portrait construction is carried out on the target user according to the target message data of each data source to obtain target user portrait; carrying out priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source; carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set; the method comprises the steps of obtaining a message to be pushed, and carrying out pushing channel analysis on the message to be pushed through an initial message pushing strategy to obtain an initial pushing channel; collecting current network state data of a target user in real time, and analyzing pushing time of an initial pushing channel based on the current network state data to obtain initial pushing time; message pushing is carried out on the target user based on the initial pushing channel and the initial pushing time, and feedback data of the target user are collected in real time, so that real-time feedback data are obtained; and carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, carrying out pushing index correction based on the pushing fitness score to obtain a corrected pushing index, and carrying out strategy optimization on an initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy. The historical message data of the target user is acquired from a plurality of preset data sources, so that the target message data of each data source is obtained, and the activity track of the user on different platforms can be comprehensively known. Based on the data, user portrayal construction is performed on the target users according to the target message data of each data source, so that more accurate and comprehensive target user portrayal is formed. By performing priority analysis on multiple data sources, the priority order of each data source can be determined, and more flexibility and efficiency in multi-channel pushing are ensured. And obtaining a message type set by analyzing the message type of the target message data of each data source, and providing a finer information basis for the subsequent push strategy. When the initial message pushing strategy is constructed, the priority order of the message type set and the data source is comprehensively considered, the initial pushing strategy is formulated in a more intelligent and personalized mode, and the accuracy of message pushing and the satisfaction of users are improved. After the message to be pushed is obtained, pushing channel analysis is carried out through an initial message pushing strategy, so that the channel in which each message is pushed can be accurately determined, and excessive disturbance and redundant pushing of the message are effectively avoided. The current network state data of the target user is collected in real time, and pushing time analysis is carried out based on the data, so that message pushing can be carried out in a period when the user is more active, and the instantaneity of the message and the user response rate are improved. By collecting feedback data of the target user in real time, real-time user interaction information is obtained, and important data support is provided for pushing fitness score calculation. By calculating the push fitness score, the push index can be adjusted more intelligently, so that the message push meets the personalized requirements of the user. By correcting the pushing index, the initial message pushing strategy is subjected to strategy optimization, and the message pushing effect and user experience are improved. Overall, the beneficial effects brought by the process are that individuation, accuracy and user satisfaction of message pushing are improved, and an efficient and intelligent optimization method is provided for multi-channel message pushing.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of an optimized push method based on a multi-channel message in an embodiment of the present invention.

FIG. 2 is a flow chart of user portrayal construction for a target user based on target message data for each data source in an embodiment of the invention.

Fig. 3 is a schematic diagram of an optimized push system based on multi-channel messages in an embodiment of the present invention.

Reference numerals:

301. an acquisition module; 302. a first analysis module; 303. constructing a module; 304. a second analysis module; 305. a third analysis module; 306. a pushing module; 307. and a correction module.

Detailed Description

The following description of the embodiments of the present invention will be made apparent and fully in view of the accompanying drawings, in which some, but not all embodiments of the invention are shown. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. are based on the directions or positional relationships shown in the drawings, are merely for convenience of describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and thus should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.

In addition, the technical features of the different embodiments of the present invention described below may be combined with each other as long as they do not collide with each other.

For easy understanding, the following describes a specific flow of an embodiment of the present invention, referring to fig. 1, fig. 1 is a flowchart of an optimized push method based on multi-channel messages according to an embodiment of the present invention, and as shown in fig. 1, the method includes the following steps:

S101, collecting historical message data of a target user from a plurality of preset data sources to obtain target message data of each data source, and constructing a user portrait of the target user according to the target message data of each data source to obtain a target user portrait;

Specifically, firstly, historical message data of a user in different scenes is acquired through a plurality of preset data sources. Data sources include social media, email, mobile applications, etc., each of which records user communication, interaction, and consumption behavior in a particular environment. For example, user praise, comments on social media, subscription records in email, and browsing and purchase history in shopping applications are important data that helps build user portraits. By extracting and sorting the target message data of these data sources, important information about the user contained by each data source is obtained. Taking social media as an example, text, pictures, videos, etc. that are posted by users on a platform form a series of target message data. These data not only reflect the personal interests of the user, but also reveal the social relationship of the user with others. Next, a target user representation is constructed based on the target message data for each data source. By analyzing the behavior patterns and the relevance of the user on each data source, the user can know the characteristics of multiple aspects, such as hobbies, social circles, shopping habits and the like of the user. For example, users frequently read articles about technical news in emails, and mainly focus on fashion and travel topics on social media, which reflects the diverse areas of interest of users. By comprehensively analyzing the target message data of each data source, a more comprehensive target user portrait is gradually obtained. The target user image is then associated with the weight of each data source. This involves evaluating the importance and impact of different data sources to determine priority in push services. For example, if the user's liveness on social media is higher, the social media data sources may be given higher weight because the information on this platform more reflects the user's real-time interests and interactions.

S102, carrying out priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source;

Specifically, through previous data analysis, behavior and preference information of the target user in different data sources are obtained, which forms the basis of the target user portrait. For example, while the target user is interested in fashion and travel on social media, technological news is also more of a concern in email. Such a target user representation enables more accurate insight into the field of interest and the active time period of the target user.

Further, priority analysis is performed on the plurality of data sources based on the target user image, and the priority order of each data source is obtained by this analysis. The prioritization is typically based on a weighted evaluation of the various information in the target user representation. Taking the example of the target user's liveness in social media, if the target user's interaction frequency on the platform is high, this data source may take up a relatively high weight in the target user representation. This means that the social media data source may be forward in prioritization because it more accurately reflects the real-time interests and social relationships of the target user.

Further, considering the timeliness and information update frequency of different data sources, the importance of the data sources needs to be evaluated. Taking news applications and social media as examples, news applications may focus more on pushing up-to-date information, while social media focuses more on real-time social interactions of target users. This requires a comprehensive consideration of the timeliness of the data sources in the priority analysis, so that the push service is closer to the current demands of the target users.

An important goal of priority analysis is to find the most critical data sources in the multi-source information for more targeted personalized pushing. By analyzing the representation of the target user and the various data sources, more decisive data sources for the target user can be identified. For example, if a target user's subscription history in an email indicates that his attention to a particular topic is high, the email data source may occupy a more important location in priority.

This multiple data source priority analysis based on the target user profile will make the push service more intelligent and personalized. By comprehensively considering multiple dimensions such as interest, liveness, timeliness and the like of the target user, the priority order of each data source is formed, and the push service can better meet the expectations of the target user and improve the push hit rate. For example, the target user usually uses more social media on weekends, but pays more attention to the e-mail on weekdays, and the priority of each data source can be adjusted according to the rule, so that the pushing effect in the time period with high attention of the target user is improved.

S103, carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set;

In particular, each data source can provide a unique perspective with respect to user preferences, behavior patterns, and interaction history. In the process of analyzing the message type, the message is classified according to the content, format, purpose and other factors, such as information updating, interactive request, feedback collection and the like.

After the message type sets are obtained, an initial message pushing strategy is constructed according to the priority sequence of each data source based on the sets. The term "priority order" as used herein refers to ordering according to the importance of the data sources and the size of the impact on the target user. For example, if one data source is a user's interaction record on social media and the other is a browsing history of a shopping website, then the social media interaction record may be given a higher priority based on the characteristics of the target user population, as it may be more likely to reveal the user's personal preferences and behavior patterns.

The message type and priority of the data source will be considered when constructing the initial message push policy. This process requires a comprehensive analysis of the importance of the message, the expected user response and the best timing for pushing. For example, if the message type is an urgent update (such as a security alert), then the message type may be assigned a higher push priority even if the priority of the data source is not high. In contrast, for some conventional updates or promotional information, even from high priority data sources, it may be scheduled to push during periods of higher user activity to improve user engagement and responsiveness.

For example, consider an e-commerce platform whose targeted message data may come from a user's purchase history, product browsing records, and user interactions on social media. The message type analysis may reveal the user's preferences for certain types of products, the period of time the user is active, and the user's response to different marketing campaigns. When the initial message pushing policy is constructed, the platform may decide to push the product updates related to the past purchase history of the user preferentially, and then push the content generated by social media interaction (such as user comments or shared products) as the supplemental content according to the product recommended by the browsing record.

S104, obtaining a message to be pushed, and analyzing a pushing channel of the message to be pushed through an initial message pushing strategy to obtain an initial pushing channel;

In particular, the message to be pushed refers to information that needs to be delivered to the user, and may include various formats and contents, such as product updates, service notifications, marketing information, and the like. The sources of these messages may be multiple, including internal content management systems, user behavior analysis systems, and the like. For example, an e-commerce website may have a message regarding a special offer promotion to push to the user, or an online service platform may need to send a system upgrade notification to the user. Further, the process of pushing channel analysis of the message to be pushed through the initial message pushing strategy is to determine the most suitable delivery path based on deep understanding of the user population and analysis of the message content. The initial message pushing strategy is a set of rules and parameters preset based on various factors such as past user interaction data, message types, user preferences and the like. For example, if the historical data shows that young users are more inclined to receive promotional information via social media, and older users are more inclined to email, then the push policy will be set accordingly. Push channel analysis refers to analyzing and determining the optimal message delivery channel according to the initial policy and the characteristics of specific messages in a specific implementation process. Specifically, the method is based on factors such as urgency of the message, receiving habit of the user, coverage range and efficiency of the channel and the like. For example, for urgent secure update notifications, push channel analysis may tend to select a direct and immediate communication mode, such as a cell phone short message or APP push notification; whereas e-mail may be a more appropriate choice for general content updates or periodic news feeds. In the push channel analysis process, the characteristics and applicable scenes of each channel need to be carefully considered. Social media platforms are generally adapted to disseminate more attractive and interactive content, while emails are adapted to send more formal or detailed information. At the same time, feedback and preferences of the user for different channels are also considered, and this information can typically be analyzed by historical user interaction data. For example, if a user is very reactive to a certain type of message on social media, a similar message may be more suitable for pushing through this channel.

For example, an online educational platform requires pushing information about new courses to users. The platform, by analyzing the historical interaction data of the users, finds that most users prefer to use the platform at night and access through the mobile device more frequently. Thus, in push channel analysis, the platform may choose to send this message by way of APP push notifications at night, which takes into account both the user's active time and their device usage habits. In this way, the platform can more effectively attract the attention of the user, and the interest and participation of the user in the new course are improved.

S105, collecting current network state data of a target user in real time, and analyzing push time of an initial push channel based on the current network state data to obtain initial push time;

Specifically, the current network state data of the target user is collected in real time, and the method is mainly used for monitoring and recording the information such as the online state, the network connection quality, the equipment type and the like of the user. This data collection is similar to network traffic monitoring, which can help identify whether a user is online, how fast the user's internet is connected, and the type of device the user is using (e.g., a cell phone or computer). For example, if a user is using a mobile device and has access to only slower mobile data connections, this information is critical to deciding when and how to send messages to the user.

The step of push time analysis based on current network state data aims at determining an optimal message sending time. This process involves not only considering the presence of users, but also analyzing their network connection quality. This is because even if the user is online, a poor network connection may result in a delay or failure of the push notification. In addition, the type of device currently being used by the user is also an important factor. For example, if the user is using a mobile device, it may be more appropriate to send an immediate, short push notification; and may be more suitable for sending longer, detailed emails if the user is using a desktop device.

By this analysis, the initial push time, i.e. the point in time at which the sent message is most likely seen by the user and interacted with in an optimal way, can be derived. This initial push time is determined based on a complex series of algorithms and assumptions that take into account a number of factors such as the user's behavior pattern, device usage habits, and network status. For example, by analyzing past data, it may be found that the user has a higher probability of using a smartphone at night, and a higher tendency to use a desktop computer during work hours. Thus, for important messages requiring timely response by the user, it may be selected to be pushed through the mobile device at night, while for content requiring deep reading and interaction by the user, it is selected to be sent through email at work.

In practice, for example, an online shopping platform has a special value promotion about to begin, and the platform wishes to notify as many potential customers as possible. To this end, the platform monitors in real time the network status of users, such as which users are currently online, how fast the users are connected to the network, and the devices they are using. From these data, an analysis then determines the optimal push time. If the data shows that most target users use a high speed Wi-Fi connection at night, the platform may decide to inform the user of the upcoming promotional event via APP push notification during this period.

S106, pushing the message to the target user based on the initial pushing channel and the initial pushing time, and collecting feedback data of the target user in real time to obtain real-time feedback data;

Specifically, the initial push channel may include various platforms and media, such as email, social media, short messages or application notifications, etc., and the initial push time is the best point in time predicted according to the network status, activity time and behavior pattern of the user. This process is similar to finding the most direct path in a vast network space to deliver information to a user while ensuring that this path is utilized at the point in time when the user is most likely to receive and respond to the information.

Next, once the message is pushed out, the next task is to collect feedback data of the user in real time. These data are key to evaluating the effectiveness of the push strategy, including whether the user opened the message, read how long, clicked on which links, and even including direct replies or comments from the user. The collection of this information is similar to the way network analysis tools work, they can track and record each interaction of the user with the push message, providing rich data for further analysis.

In the process of obtaining real-time feedback data, it is important to be able to process these data quickly and accurately in order to adjust and optimize the push strategy in time. The processing of real-time feedback data involves converting the raw data collected into meaningful insight and metrics. For example, if most users close immediately after receiving a message, this may mean that the content is not attractive enough or the sending time is poor. Conversely, if the user spends time reading the message and explores further by clicking on the link, this may indicate that both the content and the push time are very suitable.

For example, assume an online retailer plans to push a notification to a user regarding an upcoming special offer. Merchants choose to send notifications in the afternoon hours via mobile applications based on factors such as the user's shopping history, past response data, and network activity. Once the notification is pushed, the merchant immediately begins tracking the user's reaction to this message: whether they have opened a notification, how long they have remained in the application, whether they have viewed the details of the special merchandise, or even whether they have completed the purchase.

Through analysis of these real-time feedback data, merchants can gain valuable insight, such as which user groups are most interested in special sales, which time periods of push best, and even which types of goods are most popular. Such information is critical to adjusting future push strategies, which can help merchants locate target users more accurately, optimize push content and time, and thereby improve user engagement and sales conversion.

S107, pushing fitness score calculation is conducted on the real-time feedback data to obtain pushing fitness score, pushing index correction is conducted on the basis of the pushing fitness score to obtain corrected pushing index, and strategy optimization is conducted on the initial message pushing strategy through the corrected pushing index to obtain the target message pushing strategy.

Specifically, performing weight matching on the target click rate to obtain first weight data; user preference analysis is carried out on the user behavior report to obtain target user preference data; user response rate analysis is carried out on the target user preference data to obtain target user response rate; performing weight matching on the target user response rate to obtain second weight data; based on the first weight data and the second weight data, carrying out weighted summation on the target click rate and the target user response rate to obtain a push fitness score; carrying out score range analysis on the push fitness score to obtain a target score range; performing index correction parameter analysis based on the target score range to obtain correction pushing indexes; and carrying out strategy optimization on the initial message pushing strategy by correcting the pushing index to obtain the target message pushing strategy.

By executing the steps, the historical message data of the target user are collected from a plurality of preset data sources to obtain the target message data of each data source, and the user portrait construction is carried out on the target user according to the target message data of each data source to obtain the target user portrait; carrying out priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source; carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set; the method comprises the steps of obtaining a message to be pushed, and carrying out pushing channel analysis on the message to be pushed through an initial message pushing strategy to obtain an initial pushing channel; collecting current network state data of a target user in real time, and analyzing pushing time of an initial pushing channel based on the current network state data to obtain initial pushing time; message pushing is carried out on the target user based on the initial pushing channel and the initial pushing time, and feedback data of the target user are collected in real time, so that real-time feedback data are obtained; and carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, carrying out pushing index correction based on the pushing fitness score to obtain a corrected pushing index, and carrying out strategy optimization on an initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy. The historical message data of the target user is acquired from a plurality of preset data sources, so that the target message data of each data source is obtained, and the activity track of the user on different platforms can be comprehensively known. Based on the data, user portrayal construction is performed on the target users according to the target message data of each data source, so that more accurate and comprehensive target user portrayal is formed. By performing priority analysis on multiple data sources, the priority order of each data source can be determined, and more flexibility and efficiency in multi-channel pushing are ensured. And obtaining a message type set by analyzing the message type of the target message data of each data source, and providing a finer information basis for the subsequent push strategy. When the initial message pushing strategy is constructed, the priority order of the message type set and the data source is comprehensively considered, the initial pushing strategy is formulated in a more intelligent and personalized mode, and the accuracy of message pushing and the satisfaction of users are improved. After the message to be pushed is obtained, pushing channel analysis is carried out through an initial message pushing strategy, so that the channel in which each message is pushed can be accurately determined, and excessive disturbance and redundant pushing of the message are effectively avoided. The current network state data of the target user is collected in real time, and pushing time analysis is carried out based on the data, so that message pushing can be carried out in a period when the user is more active, and the instantaneity of the message and the user response rate are improved. By collecting feedback data of the target user in real time, real-time user interaction information is obtained, and important data support is provided for pushing fitness score calculation. By calculating the push fitness score, the push index can be adjusted more intelligently, so that the message push meets the personalized requirements of the user. By correcting the pushing index, the initial message pushing strategy is subjected to strategy optimization, and the message pushing effect and user experience are improved. Overall, the beneficial effects brought by the process are that individuation, accuracy and user satisfaction of message pushing are improved, and an efficient and intelligent optimization method is provided for multi-channel message pushing.

In a specific embodiment, the process of executing step S101 may specifically include the following steps:

(1) Analyzing the data acquisition interfaces of the plurality of data sources to obtain the data acquisition interface of each data source;

(2) Verifying the identity information of the target user based on the data acquisition interface of each data source to obtain a verification result;

(3) When the verification result is that the verification is passed, acquiring historical message data of a target user from a plurality of data sources to obtain initial message data of each data source;

(4) Performing stem extraction on the initial message data of each data source respectively to obtain a stem data set of the initial message data of each data source;

(5) Generating target message data for each data source based on the stem data set of the initial message data for each data source;

(6) And constructing the user portrait of the target user according to the target message data of each data source to obtain the target user portrait.

Specifically, first, data acquisition interface analysis is started for a plurality of data sources. There is a particular need to identify and understand the data interfaces provided by various data sources (e.g., social media platforms, online shopping websites, search engines, etc.), which are channels for obtaining data. The purpose of the data interface analysis is to determine what types of data each data source can provide, such as browsing history, shopping habits, social interactions, etc. of the user, and how such data is accessed and downloaded. Next, identity information verification is performed on the target user based on the data acquisition interface of each data source. Authentication typically involves the validation of user login credentials or the assurance of legitimate acquisition of data by technical means such as API keys. Once the user identity is verified, the user's historical message data can be securely collected from the various data sources. After the initial message data for each data source is collected, the next step performed is stem extraction. Stem extraction is a text analysis technique that simplifies complex text data into basic word elements or root forms, which helps identify and analyze key information in text. For example, keywords related to a particular topic or interest are extracted from the user's social media dynamics. Based on these stem data sets, target message data for each data source is then generated. This process involves integrating information in the stem dataset, translating it into meaningful insight, such as points of interest, activity patterns, or preference trends of the user. In this way, the raw message data provided by each data source is converted into more structured and operational information. Finally, the user portraits of the target users are constructed using the target message data. The user profile is a comprehensive user profile that includes information in multiple dimensions of user interests, preferences, behavior patterns, and the like. The representation is constructed by comprehensively analyzing the target message data provided by each data source, and can reflect the overall characteristics and habits of the user. For example, if a user frequently browses and purchases athletic equipment on multiple shopping sites, their user portraits may emphasize their interest in sports and fitness.

In one embodiment, as shown in fig. 2, the process of performing the user profile construction step for the target user based on the target message data for each data source may specifically include the steps of:

S201, analyzing the content of interest of the user to the target user according to the target message data of each data source to obtain the content of interest of the user;

S202, analyzing the active time of a target user according to the target message data of each data source to obtain the user active time of each data source;

s203, based on the user active time of each data source, sorting the active time of the plurality of data sources to obtain a time sorting result;

S204, respectively carrying out weight data matching on each data source based on a time sequencing result to obtain weight data of each data source;

S205, constructing a user portrait of the target user through the user interested content and the user active time of each data source based on the weight data of each data source, and obtaining the target user portrait.

It should be noted that, first, the target message data of each data source needs to be analyzed to identify the content of interest to the target user. This step involves mining the user's behavioral data, such as the content of the user's endorsements, comments on social media, or the types of merchandise that is viewed and purchased on the e-commerce platform. In this way, points of interest of the user, such as concerns about a particular brand, entertainment activity, or news event, may be revealed. For example, if a user interacts frequently on car-related social media posts, the user may be identified as having a high interest in the car. Next, an activity time analysis will be performed on the targeted message data for each data source to determine the user's activity period on each platform. This includes analyzing the user's online behavior over a particular time, such as early morning, hours of operation, evening, or weekend. The user's liveness may vary over different time periods, and this information is critical to understanding the user's daily habits and optimal interaction time.

The plurality of data sources will then be ranked in order of active time based on the user's active time for each data source. This means that which platforms are most important to the user will be determined based on the user's liveness and time distribution in the various platforms. By this ordering, priority may be given to the platform where the user is most active, thereby collecting and analyzing data more efficiently. After the time ordering result is obtained, weight data matching is carried out on each data source. This step is to assign a weight to each data source that reflects the relative importance of the data source in building the user representation. For example, if a user spends a significant amount of time on a particular social media platform and exhibits significant interest preferences on that platform, the platform will be weighted higher in user portrayal construction. And finally, carrying out user portrait construction on the target user by combining the content of interest of the user and the user active time of each data source based on the weight data of each data source. The user representation will comprehensively reflect the behavior patterns and interest preferences of the user on different platforms, thereby providing a multi-dimensional, comprehensive user profile. For example, for a user who is shopping often in the evening through a cell phone, his portrayal may emphasize his interest in the electronic product and high activity during the evening.

Through the process, the target user can be comprehensively analyzed from multiple dimensions, and an accurate and comprehensive user portrait is finally constructed. This image may be used for a variety of applications such as personalized marketing, user experience optimization, or behavioral prediction. For example, a home electronics sub-business company may utilize a user representation to push product advertisements that may be of interest to a user, or to send promotional information at times when the user is most active, thereby enhancing the effectiveness of a marketing campaign.

In a specific embodiment, the process of executing step S103 may specifically include the following steps:

(1) Respectively carrying out message structure analysis on the target message data of each data source to obtain a plurality of message structures;

(2) Based on a plurality of message structures, respectively splitting message fields of target message data of each data source to obtain a message field set corresponding to each data source;

(3) Message type analysis is carried out on the target message data of each data source to obtain a message type set;

(4) Performing type pushing priority analysis on the message type set to obtain type pushing priority data;

(5) Carrying out push frequency analysis on each message type in the message type set to obtain push frequency data of each message type;

(6) Carrying out pushing period analysis on each message type in the message type set to obtain pushing period data of each message type;

(7) Based on the type push priority data, push frequency data of each message type and push period data of each message type, an initial message push strategy is built according to the priority sequence of each data source.

Specifically, first, message structure analysis is performed on the target message data of each data source, and the purpose of the message structure analysis is to reveal constituent elements of the message, such as a header, a body, a link, or an attachment, etc. For example, for a social media post, the message structure may include a user name, post content, picture or video link, or the like. Next, based on these message structures, field splitting will be performed on the target message data for each data source. This process involves breaking down each message into smaller parts in order to better understand and process the information for each part. Message field splitting helps identify key information in a message, such as keywords, topic labels, or link addresses. In this way, meaningful information elements can be extracted from the original message data, providing a basis for subsequent analysis. Then, message type analysis is performed for the set of message fields corresponding to each data source. Message type analysis is the process of identifying the category to which different messages belong, such as news updates, promotional advertisements, user feedback, and the like. This step is critical to understanding the nature and purpose of the message content. For example, the message types in an email service may include promotional mail, subscription notifications, or service updates. Next, a type push priority analysis will be performed on the set of message types to determine the relative importance and push order of the different types of messages. This involves evaluating which types of messages are more important or urgent for the user and should be pushed with priority. For example, urgent service notifications may be given higher push priority, while conventional marketing mail may have lower priority. In addition, a push frequency analysis is performed for each message type in the set of message types to determine how often each type of message should be pushed. This step ensures that the frequency of message pushing can meet the information requirements of the user and is not too frequent to cause the user to feel trouble. For example, daily news updates may be pushed once a day, while promotional event notifications may be pushed only once a week. At the same time, a push period analysis is also performed for each message type in the set of message types to determine an optimal message transmission time. This analysis helps determine which time of day a particular type of message is most appropriate to push to the user. For example, entertainment related content may be more effective to push in the evening or weekend. Finally, after comprehensively considering the type pushing priority data, the pushing frequency data of each message type and the pushing period data of each message type, an initial message pushing strategy is constructed according to the priority sequence of each data source. This strategy is a plan that integrates all of the foregoing analysis results, with the aim of maximizing the effectiveness of message pushing. For example, for an online retail platform, emergency notifications (high priority, high frequency, specific periods) may be pushed first about upcoming ending promotions, followed by regular product updates (medium priority, lower frequency, flexible periods).

In a specific embodiment, the process of executing the step S104 may specifically include the following steps:

(1) Carrying out message type analysis on the message to be pushed to obtain a target message type;

(2) Analyzing the importance degree of the message to be pushed based on the type of the target message to obtain the importance degree of the target message;

(3) Carrying out matching index analysis on the type of the target message through the target user image to obtain an image matching index;

(4) Extracting user associated information from the importance of the target message based on the portrait matching index to obtain target user associated information;

(5) And carrying out push channel analysis on the target user associated information through an initial message push strategy to obtain an initial push channel.

In particular, the message types may include, for example, news updates, promotional information, user notifications, and the like. By identifying the type of message, a basic understanding of the content and purpose of the message can be obtained. For example, if the message to be pushed is about a product promotion, then the type of such message may be classified as "promotion". After the target message type is determined, a message importance analysis is then performed. This process involves evaluating the potential value and urgency of the message to the user. The message importance analysis helps determine the priority of the message and the degree of interest that the user may have in it. For example, warning messages regarding account security may be given high importance, whereas conventional marketing updates may be less important. Next, a matching metric analysis is performed on the target message type through the target user image. In this step, the matching of the message to the user interests and preferences is evaluated using the user images that have been created. The user portrayal may contain information about the user's interests, shopping history, social media activities, etc. The purpose of the match index analysis is to determine the relevance of the message to be pushed to the user specific feature. For example, if a user portrait shows that the user is highly interested in a scientific product, then a message about the new scientific product will have a high match index. And further extracting user associated information from the importance of the target message based on the portrait matching index. This step aims to gain insight into the specific relevance of a message to a particular user. By analyzing the relationship of the user representation to the importance of the message, the information that is likely to be most valuable to the user can be more accurately identified. For example, if a message is closely related to a user's historical purchasing behavior, the message may have a higher relevance to the user. And finally, carrying out push channel analysis on the target user associated information through an initial message push strategy. At this stage, the characteristics of the different push channels (e.g., email, in-application notifications, social media, etc.), as well as the user's liveness and response patterns in these channels, are considered. The goal of push channel analysis is to determine which channel is best suited to push a particular message to improve message visibility and user engagement. For example, if a user is often active on social media, social media may be the best channel for pushing messages. By combining these steps, the most suitable push channel can be selected for each message to be pushed while ensuring that the message content is highly correlated with the interests and behavior patterns of the user. The method not only improves the effect of message pushing, but also improves the user experience, because the message received by the user is more fit with the interests and demands of the user. For example, for a user who is often shopping online, shopping preference information related to pushing via email may be selected because the analysis indicates that this type of message and pushing channel best matches the user's behavior pattern and preferences.

In a specific embodiment, the process of executing the step S106 may specifically include the following steps:

(1) Carrying out identifier analysis on the target user to obtain unique identifier data of the target user;

(2) Analyzing the pushing trigger point of the initial pushing channel and the initial pushing time through the unique identifier data to obtain a target pushing trigger point;

(3) Based on the target pushing trigger point, carrying out message pushing on the target user;

(4) Collecting click events of a target user in real time to obtain user click events;

(5) Acquiring reading time data of a target user to obtain the reading time data;

(6) The feedback content acquisition is carried out on the target user, so that target feedback content is obtained;

(7) Based on a preset time interval, carrying out click rate analysis on a user click event to obtain a target click rate;

(8) Performing user behavior report analysis on the target user based on the reading time data and the target feedback content to obtain a target behavior report;

(9) And merging the target click rate and the target behavior report into real-time feedback data.

Specifically, first, the step starts with performing identifier analysis on the target user to acquire unique identifier data of the target user. At this stage, the identity of the user is determined by analyzing the user's account information, device ID, or other unique identification code. These unique identifiers are key to identifying and tracking user activity, and they ensure the accuracy and personalization of message pushing. For example, when a user logs in on to an application, it can be identified as a particular user by account information. The initial push channel and the initial push time are then analyzed for push trigger points by the unique identifier data. This process is to determine when and through which channel the message is most appropriate to push. The push trigger point analysis considers the online behavior pattern of the user, the preference of the time interval and the response condition of the user to different channels in the past. For example, if a user frequently views information through a cell phone application during the evening hours, it may choose to push messages through the application at that time. And then, based on the determined target pushing trigger point, pushing the message to the target user. This step ensures that messages are delivered to the user through the most appropriate channel at the best opportunity. For example, for a promotional event notification, if the analysis shows that the user is more active during the weekday noon, the push will be made during this period. Then, the click event of the target user is collected in real time. These click events include user response actions to push messages, such as opening a message, clicking on a link, etc. Such data is critical to assessing the user's interest in push content. For example, if a user frequently clicks on a promotional message of some type, this indicates that they have a high interest in such content. Meanwhile, reading time data of the target user can be acquired. Reading time data reflects the length of time a user spends on a message, which is another key indicator of the user's engagement. For example, the act of reading a news update for a long period of time indicates that the user is very interested in the message content. In addition, feedback content collection is performed on the target user, including possible direct replies, comments or feedback of the user. These feedback content provide a deeper understanding of user satisfaction and preferences. For example, a positive comment by a user under a promotional message may indicate that they are highly satisfied with such activity. And then, carrying out click rate analysis on the user click event based on a preset time interval to obtain a target click rate. Click rate analysis involves evaluating the frequency of response of a user to push messages over a period of time and is an important indicator for measuring the attractiveness of the message. For example, a high click-through rate indicates that a certain type of message is very attractive to the user. In addition, based on the reading time data and the target feedback content, user behavior report analysis is performed on the target user, so that a target behavior report is obtained. This report provides an integrated view of how the user interacted with the message, including the user's behavioral patterns, preferences, and feedback trends. And finally, merging the target click rate and the target behavior report into real-time feedback data. These real-time feedback data are critical to optimizing future message pushing strategies because they provide direct evidence as to which types of messages are most popular, which times and channels are most efficient. For example, by analyzing real-time feedback data, it may be found that users react very aggressively to messages for certain types of product updates, increasing the push frequency and priority of such messages in the future.

In a specific embodiment, the process of executing step S107 may specifically include the following steps:

(1) Performing weight matching on the target click rate to obtain first weight data;

(2) User preference analysis is carried out on the user behavior report to obtain target user preference data;

(3) User response rate analysis is carried out on the target user preference data to obtain target user response rate;

(4) Performing weight matching on the target user response rate to obtain second weight data;

(5) Based on the first weight data and the second weight data, carrying out weighted summation on the target click rate and the target user response rate to obtain a push fitness score;

(6) Carrying out score range analysis on the push fitness score to obtain a target score range;

(7) Performing index correction parameter analysis based on the target score range to obtain correction pushing indexes;

(8) And carrying out strategy optimization on the initial message pushing strategy by correcting the pushing index to obtain the target message pushing strategy.

It should be noted that, first, the target click rate is subjected to weight matching to obtain first weight data. In this step, the click rate, i.e. the frequency of clicking on a particular message by the user, is given a particular weight value to reflect its importance in evaluating the push strategy. For example, click-through rates may be given higher weights for message types that are highly attractive to users, such as urgent updates or special offers, because these message types generally indicate stronger user engagement and interest. Then, user preference analysis is performed on the user behavior report, so that target user preference data is obtained. The user behavior report contains the user's response and interaction patterns to various messages, and by analyzing this data, the user's preference trends, such as preferences for certain types of products or content, can be revealed. For example, if users often read and interact with technical news or product update class messages, this indicates that they are interested in such topics. And then, carrying out user response rate analysis on the target user preference data to obtain target user response rate. User response rate refers to the frequency of user responses or feedback to messages, and is an important indicator for measuring user engagement and feedback activity. For example, a user who often gives feedback to promotional messages may indicate that they are of particular interest to such messages. And then, carrying out weight matching on the target user reply rate to obtain second weight data. In this step, the weight of the user reply reflects the importance of the user feedback in adjusting the push strategy. For example, if a user often gives an aggressive reply to a certain type of push content, this behavior pattern will be given a higher weight in the adjustment of the push policy. And then, based on the first weight data and the second weight data, carrying out weighted summation on the target click rate and the target user response rate, thereby obtaining the push fitness score. This score is calculated by considering the clicking and replying actions of the user in combination, which provides a quantified indicator for assessing the overall response of the user to the current push content. For example, if a user has a high click rate and a high response rate for a certain type of message, the push fitness score for such a message will be high. And then, analyzing the score range of the push fitness score, thereby obtaining a target score range. This analysis step aims at determining the location of the push fitness score within the effective range to decide whether adjustments to the push strategy are needed. For example, an fitness score above a certain threshold may indicate that the current push strategy is very effective, while a score below the threshold may indicate that improvement is needed. Then, index correction parameter analysis is performed based on the target score range, thereby obtaining a correction push index. This step involves evaluating which push parameters need to be adjusted to promote the effect of the push strategy. For example, if the analysis finds that the fitness score of a certain type of message is low, it may be necessary to adjust the push frequency or push period of such messages. And finally, carrying out strategy optimization on the initial message pushing strategy by correcting the pushing index to obtain the target message pushing strategy. This means that based on the collected user data and analysis results, its pushing method may be adjusted, such as changing the pushing frequency or priority of certain messages, to ensure that future message pushing more conforms to the user's preferences and behavior patterns. For example, if the analysis shows that the user is more aggressive with content pushed at night, the policy may be optimized, scheduling more important content to be pushed at night.

The embodiment of the invention also provides an optimized pushing system based on the multi-channel message, as shown in fig. 3, which specifically comprises the following steps:

The acquisition module 301 is configured to acquire historical message data of a target user from a plurality of preset data sources, obtain target message data of each data source, and perform user portrait construction on the target user according to the target message data of each data source, so as to obtain a target user portrait;

A first analysis module 302, configured to perform priority analysis on a plurality of data sources according to the target user image, so as to obtain a priority order of each data source;

A construction module 303, configured to perform message type analysis on the target message data of each data source to obtain a message type set, and construct an initial message pushing policy according to the priority order of each data source based on the message type set;

The second analysis module 304 is configured to obtain a message to be pushed, and perform push channel analysis on the message to be pushed through the initial message push policy, so as to obtain an initial push channel;

The third analysis module 305 is configured to collect current network state data of the target user in real time, and perform push time analysis on the initial push channel based on the current network state data, so as to obtain an initial push time;

The pushing module 306 is configured to push a message to the target user based on the initial pushing channel and the initial pushing time, and collect feedback data of the target user in real time, so as to obtain real-time feedback data;

The correction module 307 is configured to perform push fitness score calculation on the real-time feedback data to obtain a push fitness score, correct a push index based on the push fitness score to obtain a corrected push index, and perform policy optimization on the initial message push policy by using the corrected push index to obtain a target message push policy.

Through the cooperative work of the modules, historical message data of a target user are collected from a plurality of preset data sources to obtain target message data of each data source, and user portrait construction is carried out on the target user according to the target message data of each data source to obtain target user portrait; carrying out priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source; carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set; the method comprises the steps of obtaining a message to be pushed, and carrying out pushing channel analysis on the message to be pushed through an initial message pushing strategy to obtain an initial pushing channel; collecting current network state data of a target user in real time, and analyzing pushing time of an initial pushing channel based on the current network state data to obtain initial pushing time; message pushing is carried out on the target user based on the initial pushing channel and the initial pushing time, and feedback data of the target user are collected in real time, so that real-time feedback data are obtained; and carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, carrying out pushing index correction based on the pushing fitness score to obtain a corrected pushing index, and carrying out strategy optimization on an initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy. The historical message data of the target user is acquired from a plurality of preset data sources, so that the target message data of each data source is obtained, and the activity track of the user on different platforms can be comprehensively known. Based on the data, user portrayal construction is performed on the target users according to the target message data of each data source, so that more accurate and comprehensive target user portrayal is formed. By performing priority analysis on multiple data sources, the priority order of each data source can be determined, and more flexibility and efficiency in multi-channel pushing are ensured. And obtaining a message type set by analyzing the message type of the target message data of each data source, and providing a finer information basis for the subsequent push strategy. When the initial message pushing strategy is constructed, the priority order of the message type set and the data source is comprehensively considered, the initial pushing strategy is formulated in a more intelligent and personalized mode, and the accuracy of message pushing and the satisfaction of users are improved. After the message to be pushed is obtained, pushing channel analysis is carried out through an initial message pushing strategy, so that the channel in which each message is pushed can be accurately determined, and excessive disturbance and redundant pushing of the message are effectively avoided. The current network state data of the target user is collected in real time, and pushing time analysis is carried out based on the data, so that message pushing can be carried out in a period when the user is more active, and the instantaneity of the message and the user response rate are improved. By collecting feedback data of the target user in real time, real-time user interaction information is obtained, and important data support is provided for pushing fitness score calculation. By calculating the push fitness score, the push index can be adjusted more intelligently, so that the message push meets the personalized requirements of the user. By correcting the pushing index, the initial message pushing strategy is subjected to strategy optimization, and the message pushing effect and user experience are improved. Overall, the beneficial effects brought by the process are that individuation, accuracy and user satisfaction of message pushing are improved, and an efficient and intelligent optimization method is provided for multi-channel message pushing.

The above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the scope of the claims.

Claims

1. An optimized pushing method based on multi-channel information is characterized by comprising the following steps:

collecting historical message data of a target user from a plurality of preset data sources to obtain target message data of each data source, and constructing a user portrait of the target user according to the target message data of each data source to obtain a target user portrait;

Performing priority analysis on a plurality of data sources according to the target user image to obtain the priority sequence of each data source;

Carrying out message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message pushing strategy according to the priority order of each data source based on the message type set;

obtaining a message to be pushed, and carrying out push channel analysis on the message to be pushed through the initial message push strategy to obtain an initial push channel;

Collecting current network state data of the target user in real time, and analyzing the pushing time of the initial pushing channel based on the current network state data to obtain initial pushing time;

Message pushing is carried out on the target user based on the initial pushing channel and the initial pushing time, and feedback data of the target user are collected in real time to obtain real-time feedback data, and the method specifically comprises the following steps: carrying out identifier analysis on the target user to obtain unique identifier data of the target user; analyzing the pushing trigger point of the initial pushing channel and the initial pushing time through the unique identifier data to obtain a target pushing trigger point; based on the target pushing trigger point, pushing the message to the target user; collecting click events of the target user in real time to obtain user click events; collecting reading time data of the target user to obtain the reading time data; collecting feedback content of the target user to obtain target feedback content; based on a preset time interval, carrying out click rate analysis on the user click event to obtain a target click rate; performing user behavior report analysis on the target user based on the reading time data and the target feedback content to obtain a target behavior report; merging the target click rate and the target behavior report into the real-time feedback data;

Performing push fitness score calculation on the real-time feedback data to obtain a push fitness score, performing push index correction based on the push fitness score to obtain a corrected push index, and performing policy optimization on the initial message push policy through the corrected push index to obtain a target message push policy, wherein the method specifically comprises the following steps: performing weight matching on the target click rate to obtain first weight data; performing user preference analysis on the user behavior report to obtain target user preference data; user response rate analysis is carried out on the target user preference data to obtain target user response rate; performing weight matching on the target user response rate to obtain second weight data; based on the first weight data and the second weight data, carrying out weighted summation on the target click rate and the target user response rate to obtain the push fitness score; analyzing the score range of the push fitness score to obtain a target score range; performing index correction parameter analysis based on the target score range to obtain the correction pushing index; and carrying out strategy optimization on the initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy.

2. The optimized pushing method based on multi-channel message as claimed in claim 1, wherein the step of collecting historical message data of a target user from a plurality of preset data sources to obtain target message data of each data source, and constructing a user portrait for the target user according to the target message data of each data source to obtain a target user portrait comprises the steps of:

Analyzing the data acquisition interfaces of the data sources to obtain the data acquisition interface of each data source;

carrying out identity information verification on the target user based on the data acquisition interface of each data source to obtain a verification result;

When the verification result is that verification is passed, acquiring historical message data of the target user from a plurality of data sources to obtain initial message data of each data source;

performing stem extraction on the initial message data of each data source respectively to obtain a stem data set of the initial message data of each data source;

generating target message data for each of the data sources based on a stem data set of the initial message data for each of the data sources;

and constructing the user portrait of the target user according to the target message data of each data source to obtain the target user portrait.

3. The optimized push method based on multi-channel messages according to claim 2, wherein said step of constructing a user representation of said target user based on target message data of each of said data sources, comprises the steps of:

Analyzing the content of interest of the user to the target user according to the target message data of each data source to obtain the content of interest of the user;

Analyzing the active time of the target user according to the target message data of each data source to obtain the user active time of each data source;

Based on the user active time of each data source, carrying out active time sequencing on a plurality of data sources to obtain a time sequencing result;

based on the time sequencing result, respectively carrying out weight data matching on each data source to obtain weight data of each data source;

And constructing a user portrait of the target user through the user interested content and the user active time of each data source based on the weight data of each data source, so as to obtain the target user portrait.

4. The optimized push method based on multi-channel messages according to claim 1, wherein said step of performing message type analysis on the target message data of each data source to obtain a message type set, and constructing an initial message push policy by a priority order of each data source based on the message type set, comprises:

Respectively carrying out message structure analysis on the target message data of each data source to obtain a plurality of message structures;

based on a plurality of message structures, respectively splitting message fields of target message data of each data source to obtain a message field set corresponding to each data source;

Message type analysis is carried out on the target message data of each data source to obtain a message type set;

Analyzing the type pushing priority of the message type set to obtain type pushing priority data;

carrying out push frequency analysis on each message type in the message type set to obtain push frequency data of each message type;

carrying out pushing period analysis on each message type in the message type set to obtain pushing period data of each message type;

Based on the type push priority data, push frequency data of each message type and push period data of each message type, and constructing an initial message push strategy according to the priority sequence of each data source.

5. The optimized pushing method based on multi-channel messages according to claim 1, wherein the step of obtaining the message to be pushed, performing pushing channel analysis on the message to be pushed through the initial message pushing policy, and obtaining an initial pushing channel comprises the following steps:

analyzing the message type of the message to be pushed to obtain a target message type;

analyzing the message importance of the message to be pushed based on the target message type to obtain the importance of the target message;

Carrying out matching index analysis on the target message type through the target user image to obtain an image matching index;

extracting user associated information from the importance of the target message based on the portrait matching index to obtain target user associated information;

and carrying out push channel analysis on the target user associated information through the initial message push strategy to obtain the initial push channel.

6. A multi-channel message based optimization push system for performing the multi-channel message based optimization push method as claimed in any one of claims 1 to 5, comprising:

The pushing module is used for pushing the message to the target user based on the initial pushing channel and the initial pushing time, collecting feedback data of the target user in real time, and obtaining real-time feedback data, and specifically comprises the following steps: carrying out identifier analysis on the target user to obtain unique identifier data of the target user; analyzing the pushing trigger point of the initial pushing channel and the initial pushing time through the unique identifier data to obtain a target pushing trigger point; based on the target pushing trigger point, pushing the message to the target user; collecting click events of the target user in real time to obtain user click events; collecting reading time data of the target user to obtain the reading time data; collecting feedback content of the target user to obtain target feedback content; based on a preset time interval, carrying out click rate analysis on the user click event to obtain a target click rate; performing user behavior report analysis on the target user based on the reading time data and the target feedback content to obtain a target behavior report; merging the target click rate and the target behavior report into the real-time feedback data;

The correction module is used for carrying out pushing fitness score calculation on the real-time feedback data to obtain a pushing fitness score, correcting a pushing index based on the pushing fitness score to obtain a corrected pushing index, and carrying out strategy optimization on the initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy, and specifically comprises the following steps: performing weight matching on the target click rate to obtain first weight data; performing user preference analysis on the user behavior report to obtain target user preference data; user response rate analysis is carried out on the target user preference data to obtain target user response rate; performing weight matching on the target user response rate to obtain second weight data; based on the first weight data and the second weight data, carrying out weighted summation on the target click rate and the target user response rate to obtain the push fitness score; analyzing the score range of the push fitness score to obtain a target score range; performing index correction parameter analysis based on the target score range to obtain the correction pushing index; and carrying out strategy optimization on the initial message pushing strategy through the corrected pushing index to obtain a target message pushing strategy.