Intelligent touch system and method based on user characteristics

Abstract

The invention discloses an intelligent touch system and method based on user characteristics, and belongs to the technical field of message pushing. According to the system, user portraits are generated through user behavior data, the user portraits comprise tags such as reach channels, reach time and reach content, and preference characteristics such as channel preference, time preference, algorithm preference and copywriting preference are generated based on click behaviors of users. And the system calculates the weight of each reaching channel, time and content packet through an element preferential module, and selects an optimal combination to generate a message sending scheme. And if the message is not successfully sent, the system automatically selects to generate a new sending scheme again and sends the new sending scheme again. According to the method, accurate and personalized message pushing can be realized according to historical behaviors and preferences of the user, the reaching rate of the message and the response rate of the user are remarkably improved, and the method is suitable for scenes needing large-scale personalized pushing such as e-commerce platforms and social media.

Description

Technical Field

[0001] The present invention belongs to the technical field of message push, and specifically, relates to an intelligent reaching system and method based on user characteristics. Background Art

[0002] With the rapid development of Internet technology, the way of information dissemination has undergone tremendous changes. The traditional message delivery model has gradually been replaced by intelligent and personalized message push systems. Especially with the popularization of mobile Internet and social media, the channels, time and content of users receiving information have become more diverse. However, although the intelligent message push system has made certain progress, there are still many problems in practical applications, especially in how to improve the message reach and user response rate.

[0003] Limitations of traditional message push methods:

[0004] In traditional message push systems, fixed channels, fixed times and unified content are usually used to send information to users. Although this method is simple and easy to implement, it has the following significant problems: Single channel limitation: Traditional push systems often rely on a single push channel, such as SMS, email or APP notifications. The use of this single channel can easily lead to information fatigue in users, and may even cause information to fail to reach effectively due to different user preferences. For example, some users may prefer to receive information through social media rather than email. Fixed time push: Traditional systems usually send messages to all users at fixed time points, ignoring the activity of different users in different time periods. This "one-size-fits-all" push strategy often fails to capture the best time for users to receive information, resulting in a large amount of information being ignored or missed. Content homogeneity: Traditional push systems lack in-depth mining of user interests and needs, and usually send the same or similar content to all users. This lack of personalized recommendations not only reduces the effectiveness of messages, but may also cause user disgust, thereby affecting brand image. The current development status of intelligent message push

[0005] In order to deal with the above problems, intelligent and personalized message push technology has been widely used in recent years. Through big data analysis, machine learning and artificial intelligence, intelligent message push systems can make accurate recommendations based on user behavior data, preference characteristics, etc., thereby improving information reach and user engagement. Modern intelligent push systems can support multiple channels to work in parallel, including SMS, email, APP notifications, social media, etc. By selecting the most suitable channel based on the user's historical behavior data, the probability of the message being opened and read can be significantly increased. Based on user portraits and behavioral data analysis, intelligent push systems can tailor content for each user. For example, by analyzing what types of information the user has clicked in the past, it is possible to predict topics that the user may be interested in in the future, so as to make accurate recommendations. Intelligent push systems can dynamically adjust push strategies based on real-time data. For example, if a message fails to successfully reach the target user within the scheduled time, the system can automatically select the suboptimal solution for resending to ensure that the information can eventually reach the user.

[0006] Although smart push technology has made significant progress, it still faces many challenges in practical applications. Most current smart push systems still have the following problems: Although many existing smart push systems can achieve multi-channel, multi-time point and personalized content recommendation, they often lack refined management of the entire process in actual operation. For example, when multiple messages need to be reached at the same time, how to choose the optimal combination (i.e., which channel and which time point to send which content) is still a difficult problem. Existing systems usually require manual configuration of multiple activities to decide when to send what content through which channel, which not only increases operating costs, but also is prone to configuration errors or omissions. When it is expected to send multiple messages to the target user group, existing technologies usually rely on manual recommendations instead of automatically selecting the best reach plan for each user. This approach is not only inefficient, but also easily leads to some important information failing to reach the target user in time. For example, in some cases, if a message fails to be successfully sent or is ignored, the existing system cannot automatically select the suboptimal plan for resending, resulting in information loss. Although some smart push systems can optimize to a certain extent based on historical data, most systems still lack real-time dynamic optimization mechanisms. That is to say, when user behavior changes (such as changes in receiving preferences or active time periods), existing systems are often unable to adjust their push strategies in a timely manner, thus affecting the effectiveness of information delivery.

[0007] As competition in the Internet market becomes increasingly fierce, major platforms and companies hope to increase user engagement and conversion rates by improving message reach. Therefore, an efficient and accurate intelligent reach system is crucial for companies. Future development trends will focus on the following aspects:

[0008] Comprehensive and multi-dimensional data collection and analysis: In order to achieve more accurate personalized recommendations, the future intelligent reach system needs to have more powerful data collection and analysis capabilities. It is necessary not only to collect basic behavioral data (such as click-through rate, open rate, etc.), but also to combine data from more dimensions, such as geographic location, device type, social relationships, etc., in order to generate a more comprehensive and accurate portrait for each user.

[0009] Automated decision-making and optimization: The future intelligent reach system will rely more on automated decision-making and optimization mechanisms. Through machine learning algorithms, the system can automatically select the best combination for each user (including the best channel, best time point and best content), and continuously adjust the strategy based on real-time feedback to ensure that each message can reach the target audience in the most efficient way.

[0010] Seamless connection across platforms and multiple terminals: With the development of the Internet of Things and 5G technology, future information dissemination will no longer be limited to a single terminal or platform. The intelligent reach system needs to have the ability to seamlessly connect across platforms and multiple terminals to ensure that users can receive relevant information in a timely manner regardless of where they are or what device they are using.

[0011] In summary, although the prior art has solved some problems in traditional message push to a certain extent, there are still many deficiencies. Therefore, there is a need for an intelligent reach method and system based on user characteristics, which can achieve more efficient and accurate information transmission by comprehensively optimizing multiple dimensions such as channels, time and content, thereby improving message reach and user participation. This is also the problem to be solved by the present invention. Summary of the invention

[0012] Problem to be solved

[0013] In view of the problems of low intelligent message reach rate and insufficient push accuracy in the prior art, the present invention provides an improved intelligent reach system and method based on user characteristics, aiming to improve the effectiveness and accuracy of message reach, and has the following advantages and innovations:

[0014] Multi-dimensional user portrait generation:

[0015] -The present invention generates user contact channel tags, contact time tags and contact content tags based on user behavior data (such as browsing, clicking, logging in, etc.) through the user portrait module. These tags can accurately reflect user preferences, making subsequent message push more personalized and accurate.

[0016] Dynamic preference generation and real-time updating:

[0017] -The preference generation module can dynamically generate and update the user's channel preference characteristics, time preference characteristics, algorithm preference characteristics, and text preference characteristics based on the user's click behavior on historical messages. Through iterative updates, the system can continuously optimize the push strategy to ensure that each push can best meet the user's current interests and needs.

[0018] Factor selection mechanism:

[0019] -The element selection module calculates the weights of each contact channel, contact time and content package, and selects the optimal combination to generate a message sending plan. This module includes:

[0020] -Channel selection unit: Calculate the weight of each channel based on user channel preference and platform normalized click-through rate.

[0021] -Content package selection unit: According to the user algorithm preference characteristics and copy preference characteristics, they are matched with the corresponding characteristics in the content package to calculate the content package weight.

[0022] -Timing optimization unit: Calculate the weight of each time period based on the user's time preference characteristics and the platform's normalized click-through rate.

[0023] -Through these weight calculations, the system can select the optimal combination in multiple dimensions, thereby greatly improving the accuracy of message push.

[0024] Automatic reissue mechanism:

[0025] -When a message fails to be sent or does not actually reach the user, the system can ignore the time factor and re-select a new delivery plan and resend it immediately. This mechanism ensures that the message can eventually be successfully delivered to the user, improving the overall reach rate.

[0026] Intelligent push of multiple messages:

[0027] -The present invention supports providing multiple channels, multiple content packages and multiple time points within one campaign, and can intelligently select the optimal combination for push based on preset conditions. This avoids the tedious operation of creating independent campaigns for each message in the traditional way, greatly improving operational efficiency.

[0028] Flexible scalability and security:

[0029] -The system interacts with the central server through the communication module, supporting flexible data transmission and processing. The central server is not only responsible for data storage and analysis, but also can issue instructions to optimize the processing logic of each module, while ensuring the security of data transmission.

[0030] The present invention achieves precise control over the intelligent message reach process by introducing multi-dimensional user portraits, dynamically updated preference generation mechanisms, and factor selection modules. Compared with the prior art, the present invention significantly improves message reach and user interaction rates, while reducing operating costs.

[0031] Technical Solution

[0032] To solve the above problems, the present invention adopts the following technical solutions.

[0033] An intelligent reach system based on user characteristics, a user portrait module, which is used to generate a user's reach channel label, reach time label, and reach content label based on the user's behavior data;

[0034] A preference generation module is used to generate the user's channel preference features, time preference features, algorithm preference features, and text preference features based on the user's click behavior on the message;

[0035] The element selection module is used to calculate the weight of each contact channel, the weight of each contact time and the weight of each content package according to the user preference characteristics, select the optimal combination of contact channel, contact time and content package, and generate a message sending plan, wherein:

[0036] The element selection module includes:

[0037] The channel optimization unit is used to calculate the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel;

[0038] A content package selection unit, used to match the algorithm features and text features of the content package with the algorithm features and text features of the user, and calculate the weight of each content package;

[0039] The timing optimization unit is used to calculate the weight of each touch time according to the user's time preference characteristics and the platform time normalized click rate;

[0040] A combining unit, configured to generate a message sending plan that satisfies a preset number of messages by combining the contact channel weight, the contact time weight, and the content package weight in descending order of weight;

[0041] An execution module is used to send a corresponding content package to a user through a corresponding contact channel at a corresponding contact time according to the message sending plan; if the message fails to be sent or the user is not actually reached, a message sending plan is re-generated based on the best performance, the contact time is ignored, and the message is resent;

[0042] The user portrait module, the preference generation module and the factor selection module are all connected to the central server through a communication module.

[0043] The communication module is used to realize data interaction between the user portrait module, the preference generation module and the factor selection module and the central server, and the communication module includes:

[0044] A data sending unit, used to send to the central server the user contact channel label, contact time label and contact content label generated by the user portrait module, the channel preference feature, time preference feature, algorithm preference feature and text preference feature generated by the preference generation module, and the contact channel weight, contact time weight and content package weight calculated by the factor selection module;

[0045] A data receiving unit, used to receive processing instructions and data update information issued by the central server, so that the user portrait module, preference generation module and factor selection module can process according to the latest instructions and data;

[0046] The communication control unit is used to manage and control the workflow of the data sending unit and the data receiving unit, ensuring the effective transmission and secure communication of data between each module and the central server.

[0047] The following reactive power regulation algorithm formula is used between the communication module and the central server: where r ij (i=1, 2, ..., k; j=1, 2, ..., 5) represents the degree of membership of data transmission between the i-th communication module and the j-th central server.

[0048] R (k×5) : The matrix represents the relationship between k communication modules and 5 central servers; r ij : Membership degree, which indicates the quality or priority of data transmission between the i-th communication module and the j-th central server. This formula is used to optimize the load distribution during data transmission, ensuring that key data can be transmitted first when the network load is high. By calculating the membership degree, resource allocation can be dynamically adjusted to improve the response speed and stability of the system.

[0049] The following adjustable capacity algorithm formula is used between the communication module and the factor selection module:

[0050]

[0051] where in The adjustable capacity provided for the transmitted data i in time period t, where V i t is the actual transmission speed of the transmitted data i in time period t, where is the upper limit of the regulation of the transmitted data in time period i, where is the lower limit of the data transmitted in time period i, where P it is the actual amount of data bytes transmitted in time period i, where E i is the capacity service provision for the data i transmitted, where β is the daily transmission rate of the data transmitted.

[0052] is used to calculate the adjustable data capacity in each time period, and E i The overall capacity service supply is calculated by accumulating the adjustable capacity of all time periods. It is used to define and calculate E i That is, first there is the adjustable capacity for each time period, and then the total service supply is obtained by accumulating these values.

[0053] Example parameters:

[0054] Time periods: 96

[0055] Actual conveying speed V i t : Random between 0.8 and 1.2

[0056] Adjustment upper limit

[0057] The lower limit ratio is 50% of the maximum capacity.

[0058] Actual data bytes P it : Randomly generated between minimum and maximum limits

[0059] Daily transmission rate β = 0.9

[0060] Calculation process

[0061] 1. Calculate the adjustable capacity for each time period:

[0062] For each time period, use the formula

[0063] 2. Calculate the total capacity service supply:

[0064] Use the formula to accumulate the adjustable capacity of all time periods to get the total supply E i =16.2.

[0065] Wherein, the user portrait module includes:

[0066] Behavior data collection unit, used to collect user behavior data, including information on receiving, opening, and clicking messages;

[0067] A label generation unit is used to analyze and generate a user's contact channel label, contact time label, and contact content label based on the behavior data, wherein:

[0068] The reach channel label indicates the user's preference on different channels;

[0069] The reach time label indicates the user's activity level in different time periods;

[0070] The access content tags indicate the user's interest preferences for different types of content.

[0071] Wherein, the preference generation module includes:

[0072] A channel preference generation unit is used to generate a user's channel preference feature based on the user's click behavior on messages of different reach channels, and iteratively update the feature based on the user's click rate;

[0073] A time preference generating unit, used to generate a user's time preference feature based on the user's click behavior on messages with different reach times, and iteratively update the feature based on the user's message click time;

[0074] An algorithm preference generating unit, used to generate the user's algorithm preference features according to the algorithm features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on the content recommended by different algorithms;

[0075] The copy preference generating unit is used to generate the user's copy preference features according to the copy features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on different copy styles.

[0076] Wherein, the central server includes:

[0077] A data processing unit, used to receive and store data from the communication module, including user portrait tags, user preference characteristics, weights of various contact channels, contact time weights, and content package weights;

[0078] An instruction control unit, used to generate and send processing instructions and data update information to the user portrait module, preference generation module and factor selection module;

[0079] Data analysis unit, used to analyze collected user data and optimize reach strategies and feature tag systems;

[0080] Security management unit, used to ensure the security of data during transmission and processing, and to prevent data leakage and illegal access;

[0081] The load balancing unit is used to distribute data transmission tasks among multiple communication modules to improve the stability and efficiency of the system.

[0082] An intelligent reaching method based on user characteristics comprises the following steps:

[0083] (1) User profile generation: obtaining user behavior data and generating user contact channel tags, contact time tags, and contact content tags based on the behavior data;

[0084] (2) Preference feature generation: Based on the user's click behavior on the message, the user's channel preference feature, time preference feature, algorithm preference feature, and copy preference feature are generated;

[0085] (3) Calculation of factor weights:

[0086] Calculate the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel;

[0087] Calculate the weight of each touch time based on the user's time preference characteristics and the platform time normalized click rate;

[0088] Match the user's algorithm preference features and copywriting preference features with the algorithm features and copywriting features of the content package, and calculate the weight of each content package;

[0089] (4) Optimal combination selection: Based on the calculated reach channel weight, reach time weight, and content package weight, the optimal combination is selected from high to low weights to generate a message sending plan that meets the preset number of messages;

[0090] (5) Message sending: according to the message sending scheme, at a corresponding reaching time and through a corresponding reaching channel, a corresponding content package is sent to the user;

[0091] (6) Resending mechanism: If a message fails to be sent or does not actually reach the user, a new message sending plan is generated based on the best method, ignoring the reach time and immediately resending the message;

[0092] (7) Data interaction: Data interaction with the central server is realized through the communication module, user portraits, preference characteristics and weight information are sent, and processing instructions and data update information from the central server are received to update and optimize the processing of each module.

[0093] In general: Figure 4 As shown in the figure, the intelligent reach system achieves accurate message push to users through four steps: layered profiling, user preference generation, factor selection and final execution. The following is a summary and optimization suggestions for each step:

[0094] Step 1: Layered portrait labels

[0095] Objective: Based on the user's historical behavior data (such as reception, opening rate, login habits, etc.), generate portrait labels for the user's access channel, access time, and access content.

[0096] Reach channel stratification: Select the optimal channel based on the user's reception and opening rates on different channels (such as Push, IM, and email).

[0097] Optimization suggestion: When selecting channels, not only should the user’s historical preferences be considered, but also the push strategy should be adjusted dynamically based on the current network status and channel congestion.

[0098] Reach time stratification: According to the user's historical login habits, the day is divided into multiple time slots (such as 8 o'clock, 9 o'clock, etc.), and the time period when the user is most active is selected for push.

[0099] Optimization suggestion: Further refine the time division, for example, into units of half an hour or 15 minutes, and make decisions based on the user's click behavior rather than login behavior to increase the message opening rate.

[0100] Content stratification: Based on the user opening rates of different types of content (such as job recommendations, browsing similar jobs, etc.), select the most suitable content package for push.

[0101] Optimization suggestion: Introduce the concept of "content package". A content package contains algorithm features and copywriting features. Make decisions based on the user's opening rate under different algorithms and copywriting to ensure that the pushed content is more accurate.

[0102] Step 2: Preference Generation

[0103] Objective: Generate and update user preference characteristics in terms of reach channels, reach time, and reach content based on the user’s click behavior on messages.

[0104] Channel preference generation: Generate and iteratively update the user's channel preference based on the user's click behavior on messages from different channels.

[0105] Optimization suggestion: Introduce a real-time feedback mechanism to dynamically adjust users’ channel preferences based on the click data after each push.

[0106] Time preference generation: Generate and update the user's time preference based on the specific time when the user clicks the message.

[0107] Optimization suggestion: Combine the best time to send different types of messages (such as promotional information and reminder information) and further refine the time preference.

[0108] Content package preference generation: Generate and update the user's algorithm and copy preferences based on the algorithm features and copy features corresponding to when the user clicks on the message.

[0109] Optimization suggestion: Analyze a large amount of historical data through machine learning models to automatically identify the algorithms and copywriting features that are most likely to attract user attention.

[0110] Step 3: Factor selection

[0111] Objective: To calculate the weight of each factor (channel, content package, timing), and select the best combination from high to low to generate a complete message sending plan.

[0112] Channel factor selection: Normalize the channel click-through rates stored by the data team and calculate the weight of each channel. If the data of a channel is missing, the overall data of the platform will be used to supplement it.

[0113] Optimization suggestion: Introduce a real-time monitoring mechanism to dynamically adjust the weight of each channel according to the current network conditions to avoid overloading a certain channel.

[0114] Content package elements selection:

[0115] Algorithm weight calculation: The weight is calculated based on the intersection of the user's algorithm features and the algorithm features in the content package. For example, if a user has a high preference for "high-paying jobs", the content package containing the "high-paying" feature will receive a higher weight.

[0116] Calculation of copy weight: The weight is calculated based on the intersection of the user copy features and the copy features in the content package. For example, if a user likes "urgent" style copy, the content package with this style of copy will get a higher weight.

[0117] Optimization suggestions: Continuously optimize the algorithm and copy matching strategy through A / B testing to improve the overall conversion rate.

[0118] Timing factor selection: The weight of each time period is calculated based on the data stored on the day. If the data for a time period is missing, the data from the previous day or the platform as a whole is used to supplement it.

[0119] Optimization suggestion: Further optimize the timing selection strategy by combining external factors (such as weather, holidays, etc.).

[0120] Combination unit selection: The system selects the best combination from each dimension to generate the first message, and then continues to select the second best combination from the remaining elements to generate the second message until the preset number of messages is met. The time and content package in each dimension cannot be selected repeatedly, but the channel can be selected repeatedly.

[0121] Optimization suggestion: Introduce a multi-round decision-making mechanism and dynamically adjust the subsequent message combination based on real-time feedback to improve the overall push effect.

[0122] Step 4: Implementation and execution

[0123] In the actual execution process, the system will derive the intelligent reach activity configured by the operation into multiple sub-activities, and find the target sub-activity to send the message after making a decision based on the best one. If a message fails to be sent successfully for some reason, the system will ignore the time factor, re-select the best one, and resend it immediately.

[0124] Optimization suggestions:

[0125] Introduce an intelligent reissue mechanism to select the most appropriate reissue method and time by analyzing the cause of failure (such as network problems, equipment problems, etc.).

[0126] Review and analyze the feedback data after each push, and continuously optimize the operation strategy and feature label system to improve the overall push effect.

[0127] Through the four steps of layered portrait labeling, preference generation, factor selection and intelligent execution, the intelligent reach system can achieve accurate and efficient information push to users. The system can not only automatically select the optimal reach channel, time and content, but also dynamically adjust the push strategy based on real-time feedback, thereby significantly improving the message reach rate and conversion rate.

[0128] Beneficial Effects

[0129] Compared with the prior art, the present invention has the following beneficial effects:

[0130] Improve message reach and accuracy:

[0131] In the existing technology, message push often adopts a simple fixed channel, fixed time, and fixed content package approach, which lacks in-depth understanding of user personalized needs, resulting in low message reach and user response rates. Traditional push methods usually require the configuration of multiple channels, times, and content packages for each activity, and each push requires human intervention, which cannot be automated or intelligent.

[0132] The present invention can generate user portraits based on user behavior data (such as browsing, clicking, logging in, etc.) by introducing a user portrait module, a preference generation module, and an element selection module, and dynamically generate channel preference features, time preference features, algorithm preference features, and text preference features based on the user's click behavior on the message. Subsequently, the channel weight, time weight, and content package weight are calculated through the element selection module to select the optimal combination for message sending. This intelligent push mechanism greatly improves the accuracy of message reach, ensuring that each message can reach the user at the most appropriate time through the most appropriate channel with the content that best suits the user's interests, thereby significantly improving the overall message reach rate.

[0133] Dynamic update and iterative optimization:

[0134] The present invention has a dynamic update mechanism, which can update the user's preference characteristics (such as channel preference, time preference, algorithm preference and text preference) in real time by continuously monitoring and analyzing the user's click behavior. This dynamic update mechanism enables the system to continuously optimize the push strategy based on the latest user behavior data, ensuring that each push can meet the user's current interests and needs to the greatest extent.

[0135] In addition, the present invention also has an iterative optimization function, which continuously optimizes the operation strategy and feature label system through the re-analysis of historical data (such as open rate, click rate, etc.). As time goes by, the system can more accurately predict user behavior, thereby further improving the push effect. This continuous optimization mechanism ensures the stability and efficiency of the system in long-term use.

[0136] Multi-dimensional factor selection mechanism:

[0137] Traditional push methods usually only consider factors of a single dimension (such as a fixed channel or fixed time), while ignoring other important dimensions (such as content packages or algorithms). This invention introduces a multi-dimensional factor selection mechanism, comprehensively considers the three key factors of channel, time and content package, and assigns weights to each dimension. Specifically:

[0138] The channel optimization unit calculates the weight of each reach channel based on the user's channel preference characteristics and the platform's normalized click-through rate.

[0139] The timing optimization unit calculates the weight of each reach time period based on the user's time preference characteristics and the platform's normalized click-through rate.

[0140] The content package selection unit matches the user's algorithm preference features and text preference features with the corresponding features in the content package and calculates the content package weight.

[0141] Through these weight calculations, the system can select the optimal combination in multiple dimensions, thereby greatly improving the accuracy of message push. Compared with traditional methods, the present invention not only considers the sending channel, but also comprehensively considers the sending timing and content matching, making the push process more personalized and intelligent.

[0142] Automatic reissue mechanism:

[0143] In the prior art, if a message fails to be sent successfully or does not actually reach the user, it usually needs to be resent manually, which not only increases operating costs, but may also cause important information to miss the best time to send. The present invention introduces an automatic resending mechanism. When a message fails to be sent or does not actually reach the user, the system will automatically re-select and generate a new sending plan, and immediately resend it regardless of time factors. This automatic resending mechanism ensures that every important message can eventually be successfully delivered to the user, thereby improving the reliability of overall information delivery.

[0144] Improve operational efficiency and flexibility:

[0145] Traditional push methods often require creating independent activities for each message and manually configuring multiple channels, content packages, and time points, which not only consumes a lot of human resources but is also prone to errors. The present invention realizes the automatic push of multiple messages through an intelligent system, and can provide multiple channels, multiple content packages, and multiple time points in one activity. The system will intelligently select the optimal combination for push based on preset conditions, greatly simplifying the operation process and improving efficiency.

[0146] In addition, the present invention also supports flexible expansion, and operators can adjust the push strategy at any time according to business needs, such as modifying channel weights, adjusting content packages, or adding new time periods, etc. The system can quickly respond to these adjustments and make corresponding optimizations to meet business needs in different scenarios.

[0147] Data security and stability guarantee:

[0148] The present invention exchanges data with the central server through the communication module to ensure the security of the data transmission process. The central server is responsible for receiving and storing data from each module, including user portrait labels, preference characteristics and weight information of each element, and issuing processing instructions to optimize the processing logic of each module. In addition, the present invention also introduces a load balancing unit to reasonably distribute data transmission tasks among multiple communication modules, thereby improving the data processing capability and stability of the system in high-concurrency scenarios.

[0149] Personalized experience improvement:

[0150] The push methods in the prior art are often too mechanical and cannot provide differentiated services for different types of users. The present invention uses a refined strategy to treat each user as an independent individual, generate a personalized portrait based on their historical behavior data, and dynamically adjust the push strategy based on their current interests. This highly personalized service not only improves the user experience, but also enhances the user's stickiness to the platform, helping to improve business conversion rates.

[0151] Real-time feedback and closed-loop management:

[0152] The present invention has a real-time feedback function. When the message is successfully reached and opened, the system will immediately collect relevant data and feed it back to the central server. These real-time feedback data will be used for subsequent analysis to further optimize the push strategy. At the same time, the present invention adopts a closed-loop management mode, from data collection to strategy execution to result feedback, the whole process is automated, effectively reducing human intervention and improving work efficiency.

[0153] In summary, compared with the prior art, the present invention not only significantly improves the reach and accuracy of intelligent messages, but also realizes comprehensive control of the intelligent message reach process through innovative mechanisms such as dynamic update, factor selection, and multi-dimensional comprehensive consideration. At the same time, the present invention also has an automatic reissue mechanism, flexible expansion capabilities, and efficient data security protection, providing operators with a set of efficient, reliable, and intelligent information push solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0154] Figure 1 This is an architecture diagram of the user feature-based intelligent reaching system of the present invention;

[0155] Figure 2 A flowchart of the behavior event triggering of the present invention;

[0156] Figure 3 It is the intelligent access entity relationship diagram of the present invention;

[0157] Figure 4 This is a flowchart of the actual application experience of the intelligent reaching system based on user characteristics of the present invention;

[0158] Figure 5 This is one of the user interface diagrams of the user feature-based intelligent reaching system of the present invention;

[0159] Figure 6 This is a second schematic diagram of the user interface of the intelligent reaching system based on user characteristics of the present invention;

[0160] Figure 7 This is the third user interface diagram of the intelligent reaching system based on user characteristics of the present invention;

[0161] Figure 8 This is a fourth schematic diagram of the user interface of the intelligent reaching system based on user characteristics of the present invention;

[0162] Fig. 9 The intelligent reaching method based on user characteristics of the present invention is as follows Figure 1 ;

[0163] Fig.10 The intelligent reaching method based on user characteristics of the present invention is as follows Figure 2 ;

[0164] Fig.11 The intelligent reaching method based on user characteristics of the present invention is as follows Figure 3 ;

[0165] Fig.12 A block diagram of the user feature-based intelligent reaching system of the present invention;

[0166] Fig.13A block diagram of a communication module of the user feature-based intelligent reaching system of the present invention;

[0167] Fig.14 A block diagram of a user portrait module of the user feature-based intelligent reach system of the present invention;

[0168] Fig.15 A block diagram of a preference generation module of the user feature-based intelligent reach system of the present invention;

[0169] Fig.16 A block diagram of a central server of the user feature-based intelligent reaching system of the present invention;

[0170] Fig.17 This is a flow chart of the user feature-based intelligent reaching method of the present invention. DETAILED DESCRIPTION

[0171] The present invention is further described below in conjunction with specific embodiments.

[0172] Implementation

[0173] Please refer to Figure 12-16 The present invention discloses an intelligent reaching system based on user characteristics, comprising:

[0174] The user portrait module is used to generate user contact channel labels, contact time labels and contact content labels based on user behavior data.

[0175] Among them, refer to Fig.14 , the user portrait module includes:

[0176] Behavior data collection unit, used to collect user behavior data, including information on receiving, opening, and clicking messages;

[0177] A label generation unit is used to analyze and generate a user's contact channel label, contact time label, and contact content label based on the behavior data, wherein:

[0178] The reach channel label indicates the user's preference on different channels;

[0179] The reach time label indicates the user's activity level in different time periods;

[0180] The access content tags indicate the user's interest preferences for different types of content.

[0181] In the intelligent reach system based on user characteristics, the user portrait module is used to generate the user's reach channel label, reach time label and reach content label based on the user's behavior data. Specifically, the user portrait module generates labels in the following ways:

[0182] Generation of reach channel labels:

[0183] Behavioral data collection: The user portrait module collects user behavior data on different contact channels, including message reception, message opening rate, click-through rate, etc. Contact channels include but are not limited to APP Push notifications, instant messaging (IM), emails, etc.

[0184] Channel preference analysis: Analyze the user's activity and response on each channel based on the collected behavioral data. Count the number of times the user opens messages, clicks, and response time on each channel.

[0185] Label generation: Generate user contact channel labels based on the results of channel preference analysis. The labels reflect the user's preference for each contact channel, identify the user's activity on different channels, and provide a basis for subsequent channel selection.

[0186] Generation of touch time label:

[0187] Behavioral data collection: collect user's historical login time, message opening time, active time period and other behavioral data.

[0188] Time segmentation: Divide the 24 hours of a day into multiple time periods, such as every hour or every half hour as a time period. For example, it can be divided into A period (00:00-00:59), B period (01:00-01:59) and so on until N period (23:00-23:59).

[0189] Time preference analysis: Analyze the user's activity in different time periods, and count the user's message opening rate and click rate in different time periods.

[0190] Label generation: Based on the results of time preference analysis, a user's reach time label is generated. This label reflects the user's activity at different time periods of the day, identifies the time period when the user is most likely to respond to messages, and assists the system in sending messages to users at the most appropriate time.

[0191] Generation of reach content tags:

[0192] Behavioral data collection: Collect user response behavior data to different types of content, including the opening rate and click-through rate of content recommended by different algorithms and messages with different copywriting styles.

[0193] Content type classification: Classify the content into different categories based on the nature of the content. For example:

[0194] Job content: regular recommended jobs, browse similar jobs, collect similar jobs, search for keyword matching jobs, quick pitch list, new jobs, etc.

[0195] Content news: industry information, career guidance, workplace topics, etc. (specific content types can be refined according to business needs).

[0196] Content preference analysis:

[0197] Algorithm feature analysis: Analyze user feedback on content recommended by different algorithms, count the user's message opening rate under different algorithm recommendations, and identify the user's preference for content recommended by different algorithms.

[0198] Copy feature analysis: Analyze users’ responses to messages with different copy styles and identify users’ preferred copy styles (e.g., urgent, lively, positive, etc.).

[0199] Tag generation: Based on the results of content preference analysis, the user's contact content tags are generated, including algorithm feature tags and copy feature tags. The tags reflect the user's preferences for different content types and styles, which helps the system provide users with more personalized content.

[0200] Through the above process, the user portrait module, based on the user's behavioral data, details the user's preferences in terms of contact channels, contact time, and contact content. These labels provide an important basis for the system's subsequent intelligent decision-making, enabling the system to select the most appropriate channel, time, and content for message contact based on each user's unique characteristics, thereby improving the effective contact rate of messages and user satisfaction.

[0201] The implementation of this module requires not only comprehensive collection and refined analysis of user historical behaviors, but also the establishment of a dynamically updated label system. As user behaviors change, the system can update user portrait labels in real time to ensure the accuracy and timeliness of labels, and achieve continuous optimization of intelligent reach strategies.

[0202] Continue to refer to Fig.12 The intelligent reach system based on user characteristics also includes a preference generation module, which is used to generate the user's channel preference characteristics, time preference characteristics, algorithm preference characteristics and copy preference characteristics based on the user's click behavior on the message.

[0203] Among them, refer to Fig.15 , the preference generation module includes:

[0204] A channel preference generation unit is used to generate a user's channel preference feature based on the user's click behavior on messages of different reach channels, and iteratively update the feature based on the user's click rate;

[0205] A time preference generating unit, used to generate a user's time preference feature based on the user's click behavior on messages with different reach times, and iteratively update the feature based on the user's message click time;

[0206] An algorithm preference generating unit, used to generate the user's algorithm preference features according to the algorithm features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on the content recommended by different algorithms;

[0207] The copy preference generating unit is used to generate the user's copy preference features according to the copy features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on different copy styles.

[0208] The specific operation examples are as follows:

[0209] The preference generation module is one of the key components of the present invention. It is responsible for generating the user's channel preference characteristics, time preference characteristics, algorithm preference characteristics and text preference characteristics based on the user's click behavior on the message. This module continuously iterates and optimizes the user's preference characteristics through continuous monitoring and analysis of user behavior data, thereby providing the system with a more accurate push strategy. The specific implementation plan is as follows:

[0210] Channel preference generation unit:

[0211] -Functional description: The channel preference generation unit is used to generate the user's channel preference characteristics based on the user's message click behavior on different contact channels. The unit can identify the user's response to different channels (such as APP Push, IM message, email, etc.) and iteratively update according to the user's click rate on each channel.

[0212] -Implementation steps:

[0213] 1. Data collection: The system monitors the user's behavior of receiving, opening, and clicking messages on different channels. For example, a user may be more inclined to receive and open messages through APP Push, and less likely to use email.

[0214] 2. Preference calculation: Based on the click rate of users on each channel, the system assigns a corresponding weight to each channel. The higher the weight, the stronger the user's preference for the channel.

[0215] 3. Iterative update: As time goes by, the system will continuously iteratively update the user's channel preference characteristics based on the latest click data. For example, if a user gradually reduces the response to APP Push messages and increases the clicks on IM messages, the weight of IM messages will gradually increase.

[0216] -Application scenario: When the system needs to push a message, the channel selection unit will select the optimal channel based on the user's current channel preference characteristics to ensure that the message can be delivered to the user through the most appropriate channel.

[0217] Time preference generation unit:

[0218] -Functional description: The time preference generation unit generates the user's time preference characteristics based on the user's click behavior on messages in different time periods. The unit can identify the time period in a day when the user is most active and most likely to open messages, and iterate based on these behavior data.

[0219] -Implementation steps:

[0220] 1. Data collection: The system records the specific time each time a user clicks on a message. For example, a user may be more likely to open messages between 9 and 11 a.m. and less likely to interact in the evening.

[0221] 2. Time period division: The system divides a day into multiple time periods (such as each hour or half hour as a time slot), and counts the user's click behavior in each time period.

[0222] 3. Weight calculation and update: The system assigns weights to each time period based on the click rate in each time period, and continuously updates iteratively based on the latest data. For example, if a user usually opens messages between 8 and 10 in the morning, this time period will be given a higher weight.

[0223] -Application scenario: When the system needs to push a message, the timing selection unit will select the optimal sending time based on the user's current time preference characteristics to ensure that the message can be delivered and opened at the most appropriate time.

[0224] Algorithm preference generation unit:

[0225] -Functional description: The algorithm preference generation unit is used to generate the user's algorithm preference characteristics based on the algorithm characteristics in the content package corresponding to the message clicked by the user. The unit can identify the user's response to different algorithm recommendation content (such as job recommendations, news recommendations, etc.) and iteratively update based on these behavioral data.

[0226] -Implementation steps:

[0227] 1. Algorithm feature extraction: Each push message corresponds to a content package, which contains algorithmic recommendation content (such as job matching recommendations, personalized news recommendations, etc.). When a user clicks on a message, the system will record the algorithm features used by the content package.

[0228] 2. Response analysis and weight calculation: The system will analyze the user's response to different algorithm recommendations (such as high-paying jobs, hot news, etc.) and assign corresponding weights to various algorithm features. For example, if a user is more inclined to click on the "high-paying job" recommendation, the "high salary" algorithm feature will be assigned a higher weight.

[0229] 3. Iterative update: As more click data is collected, the system will continuously update the user's algorithm preference features. For example, if the click rate of "hot positions" increases in a certain period, the weight of the algorithm feature will gradually increase.

[0230] -Application scenario: When the system needs to push information containing algorithm-recommended content, the content package selection unit will select the content package that best suits the user's interests based on the user's current algorithm preference characteristics to improve the push effect.

[0231] Copy preference generation unit:

[0232] -Function description: The copy preference generation unit is used to generate the user's copy preference characteristics based on the copy style in the content package corresponding to the message clicked by the user. By analyzing the impact of different copy styles (such as urgent, lively, formal, etc.) on user behavior, the unit can identify the copy style that can best attract the user's attention and perform iterative updates.

[0233] -Implementation steps:

[0234] 1. Copywriting style extraction: The copywriting in each push message has different styles (such as urgent, humorous, formal, etc.). When a user clicks on a message, the system will record the copywriting style used in the content package.

[0235] 2. Response analysis and weight calculation: The system will analyze the impact of different copywriting styles on the user and assign corresponding weights to each copywriting style. For example, if the "urgent" style copywriting has a higher click-through rate in a certain period, the "urgent" copywriting style will be assigned a higher weight.

[0236] 3. Iterative update: As more data is collected, the system will continuously update the user's response to different copywriting styles. For example, if "humorous" style copywriting receives more attention during a certain period, its corresponding weight will gradually increase.

[0237] -Application scenario: When the system needs to push information with copywriting information, the content package selection unit will select the copywriting style that best suits the user's preferences based on the user's current copywriting preference characteristics to improve the push effect.

[0238] Summarize

[0239] Through the above four sub-modules, combined Fig. 9 , Fig.10 and Fig.11, the present invention can dynamically generate and continuously optimize the personalized preferences of each target user based on multi-dimensional data analysis. This refined data processing method ensures that each push can meet the needs of the target group to the greatest extent, thereby significantly improving the intelligent reach effect. In actual applications, these modules work together to provide accurate data support for subsequent factor selection modules, making the entire intelligent reach process more efficient and accurate.

[0240] Continue to refer to Fig.12 The intelligent reaching system based on user characteristics also includes a factor selection module, which is used to calculate the weight of each reaching channel, the weight of each reaching time and the weight of each content package according to the user preference characteristics, select the optimal reaching channel, reaching time and content package combination, and generate a message sending plan, wherein:

[0241] The element selection module includes:

[0242] The channel optimization unit is used to calculate the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel;

[0243] A content package selection unit, used to match the algorithm features and text features of the content package with the algorithm features and text features of the user, and calculate the weight of each content package;

[0244] The timing optimization unit is used to calculate the weight of each touch time according to the user's time preference characteristics and the platform time normalized click rate;

[0245] A combining unit, configured to generate a message sending plan that satisfies a preset number of messages by combining the contact channel weight, the contact time weight, and the content package weight in descending order of weight;

[0246] An execution module is used to send a corresponding content package to a user through a corresponding contact channel at a corresponding contact time according to the message sending plan; if the message fails to be sent or the user is not actually reached, a message sending plan is re-generated based on the best performance, the contact time is ignored, and the message is resent;

[0247] The user portrait module, the preference generation module and the factor selection module are all connected to the central server through a communication module.

[0248] Among them, refer to Fig.15 , the preference generation module includes:

[0249] A channel preference generation unit is used to generate a user's channel preference feature based on the user's click behavior on messages of different reach channels, and iteratively update the feature based on the user's click rate;

[0250] A time preference generating unit, used to generate a user's time preference feature based on the user's click behavior on messages with different reach times, and iteratively update the feature based on the user's message click time;

[0251] An algorithm preference generating unit, used to generate the user's algorithm preference features according to the algorithm features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on the content recommended by different algorithms;

[0252] The copy preference generating unit is used to generate the user's copy preference features according to the copy features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on different copy styles.

[0253] The element selection module is one of the core modules of the present invention. It is responsible for calculating the weights of each contact channel, contact time and content package according to the user's preference characteristics, and selecting the optimal combination to generate a message sending plan. This module comprehensively considers the user's channel preference, time preference, algorithm preference and copywriting preference to ensure that each message can be pushed through the most appropriate channel at the most appropriate time with the content package that best suits the user's interests.

[0254] Channel selection unit:

[0255] Functional description: The channel selection unit calculates the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel. The system determines the user's preference for each channel by analyzing the user's historical click behavior on different channels (such as APP Push, IM messages, emails, etc.), and normalizes it based on the overall channel performance of the platform.

[0256] Implementation steps:

[0257] Data collection: The system monitors the user's clicking behavior on different channels and generates the user's channel preference characteristics based on historical click-through rates.

[0258] Weight calculation: Calculate the weight of each channel based on the user's channel preference characteristics and the overall click-through rate of the platform. For example, if a user prefers to receive messages through APP Push, and the click-through rate of IM messages is low, the weight of APP Push will be higher.

[0259] Iterative update: As more data is collected, the system will continuously update the user's channel preference characteristics and weights to ensure that the push strategy always meets the latest situation.

[0260] Content package selection unit:

[0261] Functional description: The content package selection unit matches the algorithmic features and copywriting features of the user with the algorithmic features and copywriting features in the content package, and calculates the weight of each content package. This unit ensures that the content package pushed to the user not only meets their interests, but also maximizes the click-through rate.

[0262] Implementation steps:

[0263] Algorithm matching: The system first extracts the algorithm features used in each content package and matches them with the user's algorithm preference features. For example, if a user has a strong interest in "high-paying jobs" recommendations, content packages containing such algorithm features will be given a higher weight.

[0264] Copy matching: The system also analyzes the copy style in each content package and matches it with the user's copy preference characteristics. For example, if a user is more likely to click on messages with "urgent" style copy, the content package with this copy style will be given a higher weight.

[0265] Weight calculation and update: The system assigns corresponding weights to each content package based on the matching results, and continuously updates these weights as more data is accumulated.

[0266] Timing Optimization Unit:

[0267] Function description: The timing optimization unit calculates the weight of each reach time period based on the user's time preference characteristics and the platform time normalized click rate. This unit ensures that push messages can be delivered in the most appropriate time period to increase the open rate.

[0268] Implementation steps:

[0269] Time period division: The system divides a day into multiple time periods (such as each hour or half hour as a time slot), and counts the user's click behavior in each time period.

[0270] Weight calculation: The system assigns corresponding weights to each time period based on the user's activity and the overall performance of the platform in each time period. For example, if a user usually opens messages between 9 am and 11 am, this time period will be given a higher weight.

[0271] Iterative update: As more data is collected, the system will continuously update the user's response to different time periods to ensure accurate push timing.

[0272] Combination Unit:

[0273] Functional description: The combination unit is responsible for generating a complete message sending plan based on the weights of the three dimensions of reach channel, reach time and content package, from high to low. This unit ensures that each push message can meet the needs of the target users to the greatest extent possible.

[0274] Implementation steps:

[0275] Optimal combination generation: The system first selects the highest-weighted combination of reach channel, reach time, and content package to generate the first message.

[0276] Subsequent combination generation: After generating the first message, the system will continue to select the suboptimal combination from the remaining elements to generate the second message, and so on, until the preset number of messages is met.

[0277] Resend mechanism: If a message fails to be sent successfully or does not actually reach the target user for some reason, the system will re-select and generate a new delivery plan and immediately resend it regardless of the time factor. This resend mechanism ensures that important information can be successfully delivered to the target object in the end.

[0278] Practical application cases

[0279] Take a recruitment platform as an example. The platform wants to push job recommendation information to its registered users. Assume that the platform operator configures a smart reach campaign, which needs to send two job recommendation messages to the target group. The operator provides the following elements:

[0280] Channel selection: APP Push, IM message

[0281] Content package selection: Four content packages with different job recommendation algorithms and copywriting styles (such as "High-paying job recommendation", "Urgent recruitment", etc.)

[0282] Time options: 8:00am, 9:00am, 10:00am, 11:00am

[0283] User A case:

[0284] User A has a strong preference for APP Push. He usually opens messages between 9:00 and 11:00 a.m., is particularly interested in "high-paying jobs" recommendations, and tends to click on messages with "urgent" style copy.

[0285] The system first found that APP Push was the most suitable push channel through the channel optimization unit, so it was given a higher weight. In the timing optimization unit, since user A is usually active between 9:00 and 11:00, the 9:00 time period was given the highest weight. In the content package optimization unit, since the "high-paying job" recommendation meets its algorithm preference and the "urgent" style copy meets its copy preference, the content package containing these elements receives the highest weight.

[0286] The system finally generates the first message: a message combining "high-paying positions" and "urgent recruitment" will be sent via APP Push at 9:00.

[0287] Next, the system continues to select the suboptimal combination from the remaining elements to generate the second message. Since the 9:00 time slot has been used, the system selects 10:00 as the suboptimal time, selects IM message as the suboptimal channel, and selects another content package that is relevant to his interests but slightly weaker.

[0288] The second message is: Send another set of job recommendation information via IM at 10:00.

[0289] If the second message fails to be sent successfully due to a network failure, the system will ignore the time factor and reselect the best combination for resending, such as immediately sending another set of job recommendation information via IM.

[0290] Effect: This intelligent, multi-dimensional factor selection mechanism ensures that each push can meet the needs of target users to the greatest extent, improving the open rate and click-through rate. At the same time, through the automatic resending mechanism, the success rate of information delivery is further improved, effectively avoiding the loss of important information due to technical problems.

[0291] Summary: The factor selection module in the present invention provides each target user with a personalized and accurate information push solution by comprehensively considering the preference characteristics in multiple dimensions (such as channels, timing, and content). This intelligent, multi-dimensional comprehensive evaluation mechanism significantly improves the intelligent reach effect, so that each push can meet the needs of the target group to the greatest extent, thereby improving the overall business conversion rate.

[0292] Among them, continue to refer to Fig.12 The intelligent reaching system based on user characteristics also includes a communication module for realizing data interaction between the user portrait module, preference generation module and factor selection module and the central server.

[0293] Reference Fig.13 , the communication module comprises:

[0294] A data sending unit, used to send to the central server the user contact channel label, contact time label and contact content label generated by the user portrait module, the channel preference feature, time preference feature, algorithm preference feature and text preference feature generated by the preference generation module, and the contact channel weight, contact time weight and content package weight calculated by the factor selection module;

[0295] A data receiving unit, used to receive processing instructions and data update information issued by the central server, so that the user portrait module, preference generation module and factor selection module can process according to the latest instructions and data;

[0296] The communication control unit is used to manage and control the workflow of the data sending unit and the data receiving unit, ensuring the effective transmission and secure communication of data between each module and the central server.

[0297] The communication module plays a key role as a bridge in the present invention, and is responsible for realizing data interaction between the user portrait module, preference generation module, and factor selection module and the central server. Through this module, the system can effectively transmit user-related data and receive processing instructions and data update information issued by the central server, thereby ensuring the smooth operation of the entire intelligent reach process. Specifically, the communication module includes the following three core units:

[0298] Data sending unit:

[0299] Function description: The data sending unit is responsible for sending the user contact channel labels, contact time labels and contact content labels generated by the user portrait module, the channel preference features, time preference features, algorithm preference features and copywriting preference features generated by the preference generation module, and the contact channel weights, contact time weights and content package weights calculated by the factor selection module to the central server.

[0300] Implementation steps:

[0301] Data packaging: The system first packages the data generated by the above modules, including user portrait information, preference characteristics and factor weights.

[0302] Data transmission: The packaged data is sent to a central server via a secure network protocol (such as HTTPS or WebSocket).

[0303] Transmission confirmation: The system monitors the data transmission process to ensure that the data successfully reaches the central server and retransmits it when necessary.

[0304] Data receiving unit:

[0305] Functional description: The data receiving unit is used to receive processing instructions and data update information issued by the central server. These instructions and update information will be used to guide the user portrait module, preference generation module and factor selection module to make processing adjustments.

[0306] Implementation steps:

[0307] Instruction reception: The system receives processing instructions and the latest data update information from the central server through communication protocols.

[0308] Instruction parsing: The received data will be parsed into specific operation instructions (such as updating user portraits, adjusting preference features, etc.).

[0309] Execution feedback: After executing these instructions, the system will feed back the execution results to the central server to ensure the success of the operation and perform subsequent optimization.

[0310] Communication control unit:

[0311] Function description: The communication control unit is responsible for managing and controlling the workflow of the data sending unit and the data receiving unit, ensuring the effective transmission and secure communication of data between each module and the central server.

[0312] Implementation steps:

[0313] Task scheduling: According to the task priority within the system, reasonably arrange data sending and receiving tasks to avoid conflicts.

[0314] Security assurance: Encryption technology (such as SSL / TLS) is used to ensure that data is not tampered with or leaked during transmission.

[0315] Error handling and retry mechanism: If a network failure or other error occurs during the communication process, the system will automatically handle the error and try to resend or receive data.

[0316] Practical application cases

[0317] Take an e-commerce platform as an example. The platform wants to push personalized promotional information to its user groups. Assume that the platform operator configures a smart reach campaign that needs to send three promotional messages to the target group. The operator provides the following elements:

[0318] Channel selection: APP Push, SMS, Email

[0319] Content package selection: Four content packages with different promotion algorithms and copywriting styles (such as "Limited Time Discount", "New Product Recommendation", etc.)

[0320] Time options: 8:00am, 10:00am, 2:00pm

[0321] User B case:

[0322] User B has a strong preference for APP Push. He is usually active between 10:00 a.m. and 2:00 p.m., is particularly interested in "limited time discount" promotions, and tends to click on information with "urgent" style copy.

[0323] Data sending process of communication module:

[0324] The system first generates the user's reach channel label (APP Push), reach time label (10:00 to 14:00) and reach content label (limited-time discount) through the user portrait module.

[0325] The preference generation module generates user B's channel preference characteristics (APP Push), time preference characteristics (10:00 to 14:00), algorithm preference characteristics (limited-time discounts), and copywriting preference characteristics (urgent style) based on historical click behavior.

[0326] The factor selection module calculates the highest weighted combination of the APP Push channel, the 10:00 time period, and the "limited time discount + emergency style" content package.

[0327] The system packages these calculation results and sends them to the central server through the data sending unit of the communication module.

[0328] Data receiving process of communication module:

[0329] The central server analyzes the received data and issues new processing instructions, such as adjusting the push strategy or updating certain content packages.

[0330] The system receives these instructions through the data receiving unit of the communication module and interprets them into specific operations, such as adjusting the push time of user B or changing the push content.

[0331] After the system completes these operations, it feeds back the execution results to the central server.

[0332] Application of reissue mechanism:

[0333] If a message fails to be successfully sent to user B via APP Push at 10:00 as planned due to a network failure, the system will ignore the time factor and immediately send a promotional message via the SMS channel to ensure that important information is not missed.

[0334] Effect: This intelligent, multi-dimensional factor selection mechanism combined with an efficient communication module enables each push to meet the needs of target users to the greatest extent, improving the message opening rate and click-through rate. At the same time, the automatic resending mechanism further improves the success rate of information delivery and effectively avoids the loss of important information due to technical problems. In addition, through real-time communication with the central server, the system can dynamically adjust the push strategy according to the latest business needs, thereby ensuring that the entire intelligent reach process is always in the best state.

[0335] Summary: The communication module in the present invention provides stable data transmission support for core functions such as user profiling, preference generation, and factor selection through an efficient and secure data interaction mechanism. Combined with actual application cases, it can be seen that the communication module can not only ensure smooth communication between various functional modules, but also respond to changes in business needs in a timely manner, and improve the overall message reach effect by adjusting the push strategy in real time.

[0336] The following reactive power regulation algorithm formula is used between the communication module and the central server: where r ij (i=1, 2, ..., k; j=1, 2, ..., 5) represents the degree of membership of data transmission between the i-th communication module and the j-th central server.

[0337] The following adjustable capacity algorithm formula is used between the communication module and the factor selection module:

[0338]

[0339] in in The adjustable capacity provided for the transmitted data i in time period t, where V i t is the actual transmission speed of the transmitted data i in time period t, where is the upper limit of the regulation of the transmitted data in time period i, where is the lower limit of the data transmitted in time period i, where P it is the actual amount of data bytes transmitted in time period i, where E i is the capacity service provision for the data i transmitted, where β is the daily transmission rate of the data transmitted.

[0340] In this invention, the communication and data transmission between the communication module and the central server and the factor selection module adopts reactive power regulation and adjustable capacity algorithm formulas. These algorithms are designed to optimize the efficiency and reliability of data transmission and ensure that the system can efficiently transmit information under different network conditions.

[0341] The algorithm formula of reactive power regulation:

[0342] The reactive power regulation algorithm is used to optimize the data transmission between the communication module and the central server. The "membership degree" mentioned here represents the degree of association between the data transmission between the i-th communication module and the j-th central server. Reactive power regulation is usually used to control reactive power in power systems to maintain voltage stability. In the present invention, this concept is borrowed to dynamically adjust the load distribution during data transmission.

[0343] - Membership explanation: Membership can be understood as the quality or priority of data transmission between the communication module and the central server. By calculating the membership, the system can decide how to allocate resources to ensure that critical data can be transmitted first when the network load is high.

[0344] -Practical application: For example, when multiple communication modules send data to the central server at the same time, the system will determine the priority based on the degree of affiliation between each module and the server. If the data transmission affiliation of a module is high (perhaps because the module is processing an important task), the system will give priority to the data request of the module, thereby improving the response speed of key tasks.

[0345] Algorithm formula for adjustable capacity:

[0346] The adjustable capacity algorithm is used to optimize the data transmission between the communication module and the factor selection module. The variables in the formula describe the data transmission in different time periods:

[0347] - Practical application: For example, during peak hours (such as 9am to 11am), the network bandwidth may be tight. At this time, the system can dynamically adjust the data capacity of each message according to the formula to ensure that important messages can be sent in time, while minor messages can be sent later or with reduced capacity. In this way, the system can balance the data load in different time periods and improve the overall communication efficiency.

[0348] Practical application cases:

[0349] A large e-commerce platform hopes to push personalized promotional information to users through an intelligent reach system. Due to the large number of users, the platform needs to ensure that important promotional information can be pushed to target users efficiently during peak hours (such as shopping festivals) while avoiding message delays or losses due to network congestion.

[0350] Reactive power regulation algorithm application:

[0351] Assume that the platform has multiple communication modules responsible for sending key information such as user portraits and preference characteristics to the central server. During the shopping festival, the number of user visits surges, and each communication module sends a large amount of data to the central server at the same time. At this time, the reactive power regulation algorithm begins to play a role:

[0352] -The system calculates the degree of data transmission between each communication module and the central server. If a communication module is processing information of VIP customers, its degree of membership will be set to a higher value.

[0353] -Based on the degree of affiliation, the system prioritizes data requests related to VIP customers to ensure that these customers can receive promotional information in a timely manner, while data requests from ordinary customers may be processed later.

[0354] Adjustable capacity algorithm application:

[0355] Assume that the platform needs to push different types of promotional information to different user groups, and each message contains a large amount of pictures and text content. During the shopping festival, due to bandwidth limitations, the platform needs to dynamically adjust the data capacity of each message to ensure that key promotional information can be delivered in time.

[0356] -The system calculates the bandwidth available for each message during different time periods based on the adjustable capacity formula. For example, during the peak hours of 9am to 11am, the system may reduce the data capacity of some minor promotional information (such as ordinary discounted items) and allocate more bandwidth to important promotional information (such as limited-time sales).

[0357] - If a message fails to be sent as planned (for example, due to network congestion), the system will readjust its capacity based on actual conditions and resend it through other channels (such as SMS).

[0358] Effect: Through reactive power regulation and adjustable capacity algorithms, the platform can dynamically adjust resource allocation when the network load is high to ensure that key tasks are given priority. At the same time, by dynamically adjusting the data capacity of each message, the platform can effectively utilize limited bandwidth resources and improve overall communication efficiency. This mechanism not only improves the message reach rate, but also avoids the loss or delay of important information due to network congestion, thereby improving user experience and business conversion rate.

[0359] The reactive power regulation and adjustable capacity algorithms in this invention provide a flexible and efficient data transmission optimization solution. Through these algorithms, the intelligent reach system can dynamically adjust resource allocation and data transmission strategies according to real-time network conditions, thereby ensuring that important information can be delivered to target users in a timely and reliable manner. This mechanism is particularly suitable for information push requirements in high-concurrency and large-scale user scenarios, and effectively improves the overall performance of the intelligent reach system.

[0360] Among them, continue to refer to Fig.12 ,The intelligent reaching system based on user characteristics also includes a central server.

[0361] Reference Fig.16 , the central server comprises:

[0362] A data processing unit, used to receive and store data from the communication module, including user portrait tags, user preference characteristics, weights of various contact channels, contact time weights, and content package weights;

[0363] An instruction control unit, used to generate and send processing instructions and data update information to the user portrait module, preference generation module and factor selection module;

[0364] Data analysis unit, used to analyze collected user data and optimize reach strategies and feature tag systems;

[0365] Security management unit, used to ensure the security of data during transmission and processing, and to prevent data leakage and illegal access;

[0366] The load balancing unit is used to distribute data transmission tasks among multiple communication modules to improve the stability and efficiency of the system.

[0367] In this intelligent reach system, the central server plays a core role, responsible for coordinating data transmission, processing, analysis and optimization between various modules. Through the management of the central server, the system can efficiently process information such as user portraits, preference characteristics and factor selection, and adjust the push strategy based on real-time data, thereby improving the accuracy and efficiency of message reach. The following are the functions of each unit of the central server and their actual application scenarios:

[0368] Data processing unit

[0369] Functional description: This unit is responsible for receiving and storing data from the communication module, including user portrait tags, user preference characteristics, weights of each contact channel, contact time weights, and content package weights. By storing and managing these data, the system can make more accurate decisions based on historical data in subsequent operations.

[0370] Practical Application:

[0371] Scenario: In the promotion activities of e-commerce platforms, the system needs to generate user portraits based on the user's historical behaviors (such as clicks, opens, etc.) and store the user's preferences for different channels, time and content packages. The data processing unit will store this data in the central server for reference during subsequent push.

[0372] Effect: By storing this data, the system can make more accurate decisions based on historical data in future push notifications, thereby improving message reach.

[0373] Command control unit

[0374] Functional description: The instruction control unit is used to generate and send processing instructions and data update information to the user profile module, preference generation module and factor selection module. It ensures that each module can be dynamically adjusted according to the latest data and strategies.

[0375] Practical Application:

[0376] Scenario: Assuming that the preference characteristics of a certain user group have changed (for example, they start to use AppPush more frequently instead of email), the instruction control unit will send an update instruction to the preference generation module, requiring it to adjust the channel preference of this user group.

[0377] Effect: This dynamic adjustment mechanism ensures that the system can quickly respond to changes based on real-time data, so that the push strategy always remains in the optimal state.

[0378] Data Analysis Unit

[0379] Functional description: This unit is responsible for analyzing the collected user data and optimizing the reach strategy and feature tag system. By analyzing a large amount of historical data, the system can discover potential trends and adjust the push strategy based on these trends.

[0380] Practical Application:

[0381] Scenario: After a large-scale promotion, the platform wants to know which push strategies are most effective. The data analysis unit will review and analyze all push behaviors (including open rate, click rate, etc.) to optimize the reach strategy in future activities.

[0382] Effect: Through continuous data analysis, the platform can continuously optimize its push strategy and improve message reach and user interaction rate.

[0383] Security Management Unit

[0384] Functional description: The security management unit is used to ensure the security of data during transmission and processing, and prevent data leakage and illegal access. It ensures that all data involving user privacy can be properly protected.

[0385] Practical Application:

[0386] Scenario: An e-commerce platform needs to push personalized promotional information to millions of users. This information contains sensitive user behavior data (such as purchase records, browsing records, etc.). The security management unit uses encryption technology to ensure that this data will not be illegally intercepted or tampered with during transmission.

[0387] Effect: Through strict data security management, the platform can effectively prevent data leakage, enhance user trust, and comply with relevant regulatory requirements (such as GDPR).

[0388] Load Balancing Unit

[0389] Functional description: The load balancing unit is used to distribute data transmission tasks among multiple communication modules to improve the stability and efficiency of the system. It ensures that the system can still run efficiently under high concurrency conditions and will not crash or delay due to excessive load.

[0390] Practical Application:

[0391] Scenario: During the Double 11 Shopping Festival, e-commerce platforms need to push promotional information to millions of users at the same time. The load balancing unit will reasonably distribute tasks to multiple communication modules based on real-time network conditions to avoid overloading any module.

[0392] Effect: Through the load balancing mechanism, the platform can ensure that personalized messages can be sent smoothly to each user even during peak hours, thereby improving the overall system performance.

[0393] Practical application cases

[0394] Scenario Overview:

[0395] A large e-commerce platform plans to push personalized promotional information to its millions of registered users during the "Double Eleven" period. In order to ensure that each user can receive promotional content that suits their interests in a timely manner, the platform uses the intelligent contact system in the present invention to coordinate data interaction and task execution between various modules through a central server.

[0396] Step 1: Data collection and processing

[0397] The user portrait module generates portrait tags for each user, including the channels they use most frequently (such as APP Push or SMS), the most active time periods (such as 9 am to 11 am), and their interests in different types of promotional content (such as "limited-time discounts" or "new product recommendations").

[0398] The data processing unit receives and stores these portrait tags and preference features to provide a basis for subsequent push notifications.

[0399] Step 2: Send instructions and adjust strategies

[0400] Before the event starts, the command control unit issues commands based on real-time data changes. For example, if the APP Push channel is congested during a certain period of time, the unit will issue a command to the factor selection module to give priority to the SMS channel for push.

[0401] Step 3: Real-time analysis and optimization

[0402] The data analysis unit continuously monitors the opening rate and click-through rate of each message. If it is found that a certain type of promotional content performs poorly (such as a low opening rate for "new product recommendations"), the system will automatically adjust the subsequent push strategy and allocate more resources to the better performing "limited-time discount" promotional content.

[0403] Step 4: Security

[0404] During the entire process, the security management unit uses encryption technology to ensure that all data related to user privacy (such as purchase records) will not be illegally intercepted or tampered with, thereby protecting user privacy.

[0405] Step 5: Load balancing and efficient transmission

[0406] During the peak period of "Double Eleven", with millions of messages to be sent at the same time, the load balancing unit distributes the tasks to multiple communication modules to avoid network congestion or delays. In this way, the platform ensures that every user can receive personalized promotional information in a timely manner.

[0407] Through the various functional units in the central server, the present invention realizes efficient coordination and optimization in the process of intelligent message reaching. Whether it is data collection and storage, real-time instruction issuance, or security and load balancing, the various functional units cooperate with each other to jointly improve the performance and stability of the entire intelligent reaching system. In practical applications, this architecture not only improves the message reach rate, but also significantly optimizes the user experience, bringing higher business conversion rates to enterprises.

[0408] The present invention also relates to an intelligent reaching method based on user characteristics, comprising the following steps:

[0409] S1, user portrait generation: obtaining user behavior data, and generating the user's contact channel label, contact time label, and contact content label based on the behavior data;

[0410] S2, preference feature generation: based on the user's click behavior on the message, generate the user's channel preference feature, time preference feature, algorithm preference feature and copy preference feature;

[0411] S3, factor weight calculation:

[0412] Calculate the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel;

[0413] Calculate the weight of each touch time based on the user's time preference characteristics and the platform time normalized click rate;

[0414] Match the user's algorithm preference features and copywriting preference features with the algorithm features and copywriting features of the content package, and calculate the weight of each content package;

[0415] S4, optimal combination selection: Based on the calculated reach channel weight, reach time weight, and content package weight, the optimal combination is selected from high to low weights to generate a message sending plan that meets the preset number of messages;

[0416] S5, message sending: according to the message sending scheme, at a corresponding reaching time and through a corresponding reaching channel, a corresponding content package is sent to the user;

[0417] S6, resending mechanism: If the message fails to be sent or does not actually reach the user, a new message sending plan is generated based on the best method, the reaching time is ignored, and the message is resent immediately;

[0418] S7, data interaction: Realize data interaction with the central server through the communication module, send user portraits, preference characteristics and weight information, receive processing instructions and data update information from the central server to update and optimize the processing of each module.

[0419] The method for improving the effective reach rate of smart messages aims to solve the problems of low reach rate and insufficient push accuracy of smart messages in the existing technology. With the rapid development of the Internet and smart message reach technology, the form of message push has undergone tremendous changes. The traditional push method of fixed channels and time can no longer meet the personalized needs of users, resulting in a decrease in message reach rate. The existing technology lacks refined control over the smart message reach process, especially in the scenario of multiple message pushes, and cannot automatically select the optimal reach channel, time and content combination for users.

[0420] Traditional smart message push methods usually require configuring multiple independent tasks for each campaign. Each task identifies target users based on tags and sends messages to users through fixed channels, content packages, and time points. When multiple messages need to be sent to the same target user group, operators need to manually create multiple campaigns to decide to send different content through different channels at different times. This method has the following problems:

[0421] - Lack of automation: Each push requires human intervention, and the push strategy cannot be dynamically adjusted based on the user's real-time behavior.

[0422] - Low push efficiency: Multiple messages require the creation of multiple campaigns, increasing operational costs and complexity.

[0423] - Lack of precision: Unable to automatically select the optimal contact channel, time and content combination based on user preferences, resulting in poor push results.

[0424] The present invention introduces intelligent and automated message reach strategies to achieve dynamic management of user portraits, preference characteristics, and factor selection, thereby significantly improving the effective reach rate of intelligent messages. The specific innovations are as follows:

[0425] Automatic multi-dimensional optimization mechanism:

[0426] The system can provide multiple channels, multiple content packages and multiple time points in one campaign, and automatically calculate the weight of each factor based on the user's historical behavior data (such as click-through rate, open rate, etc.), and select the optimal, suboptimal and suboptimal combinations in turn to generate a complete message sending plan. Compared with the traditional method, the present invention does not require manual creation of multiple campaigns, but automatically selects the most appropriate contact plan for each user through the system.

[0427] Real-time feedback and dynamic adjustment:

[0428] The system will dynamically update the user profile and preference characteristics based on the feedback data after each push (such as open rate, click rate, etc.). Through this iterative optimization mechanism, the system can continuously improve the accuracy of the push strategy, so that each push can meet the user's current needs to the greatest extent.

[0429] Reissue mechanism:

[0430] If a message fails to be sent or does not reach the user due to network failure or other reasons, the system will re-select and generate a new delivery plan regardless of time factors and resend it immediately. This resending mechanism ensures that important information will not miss the best delivery opportunity due to unexpected factors.

[0431] By reviewing and summarizing large-scale marketing recommendations in ToC scenarios, the present invention can achieve the following technical effects:

[0432] - Accurate crowd prediction: Based on data analysis of all registered users, the system can automatically predict the optimal crowd and its preferences, thereby achieving more accurate target crowd selection.

[0433] -Intelligent channel and content recommendation: The system can recommend the most suitable contact channel and content package combination for each target user based on business requirements, thereby improving conversion rate.

[0434] -Efficient operation support: Through the automation and multi-dimensional optimization mechanism, the present invention greatly reduces the workload of operators to manually configure tasks, while improving the push efficiency and effect.

[0435] The present invention is implemented in three steps through a refined strategy:

[0436] Layered portrait label generation:

[0437] The system first generates hierarchical portrait tags based on user behavior data, including contact channel tags (such as APP Push, IM message, email), contact time tags (based on historical login habits), and contact content tags (based on historical open rates). These tags provide data support for subsequent preference generation and factor selection.

[0438] Preference generation and strategy optimization:

[0439] Based on the user's click behavior on historical messages, the system will dynamically generate and update the user's channel preference characteristics, time preference characteristics, algorithm preference characteristics, and copy preference characteristics. Through iterative optimization of these preferences, the system can continuously adjust the push strategy so that each push can meet the user's current needs to the greatest extent.

[0440] Factor selection and combination generation:

[0441] The system calculates the weights of the multiple channels, content packages, and time points provided based on user portraits and preference characteristics, and selects the optimal combination in turn to generate a complete message sending plan. For example, when sending three promotional messages to a target user group, the system will select the "limited time discount" copy sent via APP Push at 9:00 as the first message based on the weights, and then select the second best combination from the remaining elements to generate the second message, and so on until the preset number of messages is met.

[0442] Reissue mechanism and dynamic adjustment:

[0443] If a message fails to be sent successfully or does not actually reach the target user, the system will ignore the time factor and re-select a new delivery plan and resend it immediately. For example, if the delivery via APP Push at 9:00 fails, the system may choose to send another promotional message via IM at 10:00 as a resend.

[0444] The present invention achieves precise control over the intelligent message reach process by introducing a multi-dimensional intelligent optimization mechanism. Compared with the traditional method, the present invention not only improves the message reach rate, but also reduces the workload of operators to manually configure tasks, greatly improving operational efficiency. At the same time, through real-time feedback and dynamic adjustment, the present invention can continuously optimize the push strategy and bring more efficient and accurate information delivery solutions to enterprises.

[0445] For the above method, apply the following:

[0446] like Figure 1 As shown in the figure, the intelligent reach system architecture diagram shows a decision service system based on the rule engine (Drools), which aims to realize automated decision execution and rule management. The system realizes the automated processing of business logic through interaction with intelligent services and combined with the operation of the rule engine. The following is a detailed analysis of each module in the architecture diagram:

[0447] Smart Services

[0448] Function: The intelligent service module is responsible for receiving external requests or triggering events and passing these requests to the decision service for processing. It can be an event triggered by user behavior or a system scheduled task.

[0449] Function: As the entrance to the entire system, the intelligent service passes the data to be processed to the decision execution module, which processes it according to preset rules.

[0450] Decision Service (Drools)

[0451] Function: The decision service is the core part of the entire system, which makes automated decisions based on the rule engine (Drools) for business logic. It includes multiple sub-modules:

[0452] Decision execution module: receives data from intelligent services and performs corresponding operations based on the output of the rule engine. It is responsible for calling the rule engine to process business logic and returning the results to the intelligent service.

[0453] Rule engine module: This is the core module implemented based on Drools, which is responsible for reasoning and calculating the input data according to predefined business rules to obtain decision results.

[0454] Rule management module: used to manage and maintain business rules, supporting dynamic loading, updating and deletion of rules. The rule management module interacts with the database (PgSQL) and cache (Redis) to ensure that the latest business rules can take effect in a timely manner.

[0455] Decision service management: provides a management interface for operators or administrators to view, modify and configure business rules.

[0456] Data storage

[0457] Redis: used to cache business rules and improve query efficiency. In practical applications, Redis can store commonly used or frequently accessed rules to reduce the direct query pressure on the database.

[0458] PgSQL: used to persist all business rules and related data. All business rules are stored in PgSQL after creation or update, so that they can be loaded into the rule engine for execution when needed.

[0459] Rule Management

[0460] Function: This module is responsible for managing all business rules, including creation, modification, and deletion. Operation personnel can maintain these rules through the decision service management terminal. All updated rules will be synchronized to the Redis cache to increase system access speed.

[0461] Function: Through this module, the system can flexibly adjust the business logic without modifying the code. This design greatly improves the flexibility and maintainability of the system.

[0462] Decision execution

[0463] Function: When the intelligent service sends a request, the decision execution module calls the Drools engine to make judgments and calculations based on the currently effective business rules and returns the results to the intelligent service. The decision execution module also interacts with Redis and PgSQL to ensure that the latest business rules are used.

[0464] Function: This module is the part of the entire system that actually executes the business logic. It combines input data with existing rules to obtain corresponding decision results.

[0465] like Figure 2As shown in the figure, the flowchart shows the message push process triggered by behavioral events in the smart reach system. Through the cooperation of decision-making services and rule engines, the system can automatically select the optimal activity, time and content for message push based on user behavior. The following is a detailed analysis of each part of the flowchart: Behavior event service: Function: The behavior event service module is responsible for receiving the user's behavior data in the system and determining whether these behaviors meet the triggering conditions of smart reach activities. Process description: Receive message: When users perform certain operations on the platform (such as browsing, clicking, purchasing, etc.), the system will capture these behavior data and pass them to the behavior event service as a message. Get activity ID and details: The behavior event service will obtain the corresponding activity ID based on user behavior and query the detailed information of the activity. Determine whether it is a smart reach activity: The system will determine whether the behavior belongs to a smart reach activity. If yes, continue to execute the subsequent steps; if not, the process ends. Smart reach service: Function: The smart reach service module is responsible for processing qualified behavior events and selecting the optimal push strategy (such as the best activity, time point, etc.) through the rule engine. Process description: Consume Topic messages: When a behavior event is identified as a smart reach activity, the smart reach service consumes the corresponding Topic message and enters the rule factory class for processing. Rule factory class processing: The rule factory class contains multiple rule types (such as custom rules, market rules, user preference rules, polling rules, etc.). The system analyzes user behavior based on these rules and determines the optimal push strategy. Call decision service: According to the analysis results of the rule factory class, the smart reach service calls the decision service to further optimize the push strategy. Decision service (rule engine): Function: The decision service module runs predefined business logic based on the rule engine (such as Drools) to ultimately determine the optimal push solution. Process description: Return the optimal activity ID and time point: The decision service calculates the activity ID and the optimal push time point that are most suitable for the current user through the rule engine, and returns the results to the smart reach service. Push execution: According to the decision result, the system will send the corresponding message to the user to complete the entire push process. Push execution: Function: After the decision is completed, the system will generate a specific message and actually push it through the traffic framework or marketing engine. Process description: The system will generate messages containing the best activities, time points and content combinations based on the decision results, and send them to target users through preset channels (such as APP Push, SMS, etc.). This architecture achieves efficient and accurate information reach by capturing user behavior and dynamically adjusting the push strategy based on preset rules. It can not only respond quickly based on real-time user behavior, but also support flexible adjustment of business logic, making the entire system highly automated and flexible. In actual applications, this architecture is suitable for scenarios that require real-time response to user behavior and personalized recommendations or marketing push, such as personalized recommendations in e-commerce platforms and risk warnings in financial platforms.

[0466] like Figure 3 As shown in the figure, the entity relationship diagram of the smart reach system shows the relationship between multiple core entities, mainly including content packages, smart reach activities, reach content, and reach configuration. The following is a detailed analysis of each entity and its relationship:

[0467] Content Package

[0468] property:

[0469] id: A unique identifier used to distinguish different content packages.

[0470] Name: The name of the content pack, describing the purpose or type of the content pack.

[0471] Activity ID: The activity ID associated with the content package, indicating which smart reach activity the content package belongs to.

[0472] Data source type: describes the data source type of the content package, such as whether it is obtained from an internal database or an external interface.

[0473] Status: The status of the current content pack (such as enabled, disabled, etc.).

[0474] Version number: used to identify the version information of the content package for easy management and updating.

[0475] Creation Time / Update Time: Records the time when the content pack was created and last updated.

[0476] relation:

[0477] There is a one-to-many relationship between a content package and **Content Package Details**. Each content package can have multiple details, such as the specific content to be sent, the type of delivery channel, etc.

[0478] Content Package Details

[0479] property:

[0480] id: A unique identifier used to distinguish different content package details.

[0481] Content pack id: The content pack ID associated with it, indicating which content pack this detail belongs to.

[0482] Send content: The specific message or information to be sent to the user.

[0483] Sending channel type: describes the channel through which the message is sent (such as App Push, SMS, email, etc.).

[0484] Status: The status of the current detail item (such as enabled or disabled).

[0485] Version number / creation time / update time: used to manage the version and time information of the detail item.

[0486] relation:

[0487] There is a one-to-many relationship between a Content Pack Detail and the Content Packs to which it belongs.

[0488] Smart Reach Activity

[0489] property:

[0490] Activity ID: A unique identifier used to distinguish different smart reach activities.

[0491] Event Name: A name that describes the event, usually reflecting the purpose or theme of the event.

[0492] Basic information of the activity: Contains some basic descriptions of the activity, such as the target user group, push strategy, etc.

[0493] relation:

[0494] Smart reach activities are associated with multiple entities under them, including reach content and reach configuration.

[0495] Reach Content

[0496] property:

[0497] Activity ID: Associated with the smart reach activity, indicating which activity the reach content belongs to.

[0498] Activity content: specific information or messages to be pushed to users.

[0499] Content tags / algorithm tags: used to mark the specific tags or algorithm recommendation features contained in the message so that the system can match it according to user preferences.

[0500] relation:

[0501] Each smart reach activity can contain multiple reach contents, and each reach content is associated with the corresponding smart reach activity.

[0502] Reach Configuration

[0503] property:

[0504] Activity ID: Associated with the Smart Reach Activity, indicating which activity the configuration belongs to.

[0505] Audience ID: indicates the target user group for this configuration.

[0506] Data source information / algorithm data source information / frequency configuration information: used to describe the data source, algorithm recommendation rules, message push frequency and other information involved in the configuration.

[0507] relation:

[0508] Each smart reach campaign can have multiple reach configurations to define how to execute push strategies for different audience groups.

[0509] The process of using this method:

[0510] 1) If Figure 5 As shown, it schematically represents the user group that you want to reach based on user characteristics.

[0511] 2) If Figure 6 As shown, it schematically shows the configuration of the reach strategy, the selection of pre-fetched channel elements, the total number of items, and whether each element is optimized based on user preferences.

[0512] 3) If Figure 7 As shown in the figure, it schematically shows that multiple content packages are provided. The content packages consist of algorithms and texts. Each content package should provide a set of texts for each expected channel. After the creation is completed, the text features of this content package need to be defined.

[0513] 4) If Figure 8 It schematically shows the setting of the reach date and provides multiple reach times for selection.

[0514] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the field of message push technology can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An intelligent reaching system based on user characteristics, characterized in that: User portrait module, used to generate user contact channel labels, contact time labels and contact content labels based on user behavior data; A preference generation module is used to generate the user's channel preference features, time preference features, algorithm preference features, and text preference features based on the user's click behavior on the message; The element selection module is used to calculate the weight of each contact channel, the weight of each contact time and the weight of each content package according to the user preference characteristics, select the optimal combination of contact channel, contact time and content package, and generate a message sending plan, wherein: The element selection module includes: The channel optimization unit is used to calculate the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel; A content package selection unit, used to match the algorithm features and text features of the content package with the algorithm features and text features of the user, and calculate the weight of each content package; The timing optimization unit is used to calculate the weight of each touch time according to the user's time preference characteristics and the platform time normalized click rate; A combining unit, configured to generate a message sending plan that satisfies a preset number of messages by combining the contact channel weight, the contact time weight, and the content package weight in descending order of weight; An execution module is used to send a corresponding content package to a user through a corresponding contact channel at a corresponding contact time according to the message sending plan; if the message fails to be sent or the user is not actually reached, a message sending plan is re-generated based on the best performance, the contact time is ignored, and the message is resent; Among them, the user portrait module, the preference generation module and the factor selection module are all connected to the central server through a communication module.

2. The intelligent reaching system based on user characteristics according to claim 1 is characterized in that: The communication module is used to realize data interaction between the user portrait module, the preference generation module and the factor selection module and the central server, and the communication module includes: A data sending unit, used to send to the central server the user contact channel label, contact time label and contact content label generated by the user portrait module, the channel preference feature, time preference feature, algorithm preference feature and text preference feature generated by the preference generation module, and the contact channel weight, contact time weight and content package weight calculated by the factor selection module; A data receiving unit, used to receive processing instructions and data update information issued by the central server, so that the user portrait module, preference generation module and factor selection module can process according to the latest instructions and data; The communication control unit is used to manage and control the workflow of the data sending unit and the data receiving unit, ensuring the effective transmission and secure communication of data between each module and the central server.

3. The user-feature-based intelligent reaching system according to claim 2 is characterized in that: The following reactive power regulation algorithm formula is used between the communication module and the central server: where r ij =(i=1, 2, ..., k; j=1, 2, ..., 5) represents the degree of membership of data transmission between the i-th communication module and the j-th central server.

4. The user-feature-based intelligent reaching system according to claim 3 is characterized in that: The following algorithm formula for the adjustable capacity is used between the communication module and the factor selection module: in, The adjustable capacity provided for the transmitted data i in time period t, V i t is the actual transmission speed of the transmitted data i in time period t, is the upper limit of the regulation of the transmitted data in time period i, is the lower limit of the data transmitted in time period i, P it is the actual amount of data bytes transmitted in time period i, E i is the capacity service supply of the transmitted data i, and β is the daily transmission rate of the transmitted data.

5. The user-feature-based intelligent reaching system according to claim 4, characterized in that: The user portrait module includes: Behavior data collection unit, used to collect user behavior data, including information on receiving, opening, and clicking messages; A label generation unit is used to analyze and generate a user's contact channel label, contact time label, and contact content label based on the behavior data, wherein: The reach channel label indicates the user's preference on different channels; The reach time label indicates the user's activity level in different time periods; The access content tags indicate the user's interest preferences for different types of content.

6. The user feature-based intelligent reaching system according to claim 5, characterized in that: The preference generation module comprises: A channel preference generation unit is used to generate a user's channel preference feature based on the user's click behavior on messages of different reach channels, and iteratively update the feature based on the user's click rate; A time preference generating unit, used to generate a user's time preference feature based on the user's click behavior on messages with different reach times, and iteratively update the feature based on the user's message click time; An algorithm preference generating unit, used to generate the user's algorithm preference features according to the algorithm features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on the content recommended by different algorithms; The copy preference generating unit is used to generate the user's copy preference features according to the copy features in the content package corresponding to the message clicked by the user, and iteratively update according to the user's click rate on different copy styles.

7. The user feature-based intelligent reaching system according to claim 6, characterized in that: The central server comprises: A data processing unit, used to receive and store data from the communication module, including user portrait tags, user preference characteristics, weights of various contact channels, contact time weights, and content package weights; An instruction control unit, used to generate and send processing instructions and data update information to the user portrait module, preference generation module and factor selection module; Data analysis unit, used to analyze collected user data and optimize reach strategies and feature tag systems; Security management unit, used to ensure the security of data during transmission and processing, and to prevent data leakage and illegal access; The load balancing unit is used to distribute data transmission tasks among multiple communication modules to improve the stability and efficiency of the system.

8. An intelligent reaching method based on user characteristics, characterized in that: The following steps are involved: S1: User portrait generation: Obtain user behavior data, and generate user contact channel tags, contact time tags, and contact content tags based on the behavior data; S2: Preference feature generation: Based on the user's click behavior on the message, the user's channel preference feature, time preference feature, algorithm preference feature and copy preference feature are generated; S3: Element weight calculation: Calculate the weight of each reach channel based on the user's channel preference characteristics and the normalized click-through rate of the platform channel; Calculate the weight of each touch time based on the user's time preference characteristics and the platform time normalized click rate; Match the user's algorithm preference features and copywriting preference features with the algorithm features and copywriting features of the content package, and calculate the weight of each content package; S4: Optimal combination selection: Based on the calculated reach channel weight, reach time weight, and content package weight, the optimal combination is selected from high to low weights to generate a message sending plan that meets the preset number of messages; S5: message sending: according to the message sending scheme, at a corresponding reaching time and through a corresponding reaching channel, a corresponding content package is sent to the user; S6: Resend mechanism: If the message fails to be sent or does not actually reach the user, a new message sending plan is generated based on the best method, ignoring the reach time, and the message is resent immediately; S7: Data interaction: Realize data interaction with the central server through the communication module, send user portraits, preference characteristics and weight information, receive processing instructions and data update information from the central server to update and optimize the processing of each module.

Images