Method and device for intelligently and strictly selecting short message channel

By receiving user needs and collecting candidate channel attributes, dynamically adjusting the SMS channel selection model, the problem of low accuracy in SMS channel selection in the existing technology is solved, and more accurate and flexible channel selection is achieved.

CN118972791AInactive Publication Date: 2024-11-15SHENZHEN SHANGZHI TECHNOLOGY CO LTD
View PDF 0 Cites 9 Cited by

Patent Information

Application Number
CN202410960424.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing technology cannot dynamically adapt to user needs, resulting in low accuracy in selecting SMS channels.

Method used

By receiving the channel demand information of the target user and collecting multi-dimensional attribute parameters of the candidate SMS channel, the standard SMS channel selection model is retrieved, the weight is adjusted dynamically, the target SMS channel selection model is generated for the target user, the candidate channels are scored, and the comprehensive performance is selected The optimal channel.

Benefits of technology

It improves the accuracy and adaptability of SMS channel selection, and can dynamically configure the selection model according to user needs and select the optimal channel from multiple candidate channels.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118972791A_ABST
    Figure CN118972791A_ABST
Patent Text Reader

Abstract

The invention discloses a short message channel intelligent strict selection method and device, and belongs to the field of information communication, and the method comprises the steps: receiving channel demand information of a target user, and collecting channel attribute parameters respectively corresponding to N candidate short message channels; calling a standard short message channel selection model, configuring the standard short message channel selection model according to the channel demand information, and generating a target short message channel selection model; scoring the N candidate short message channels, and determining the candidate short message channel with the highest score as a target short message channel; and when the target user sends the short message, sending the short message sent by the target user through the target short message channel. The technical problem that the short message channel selection accuracy is low due to the fact that user requirements cannot be dynamically met in the prior art is solved, and the technical effects that the short message channel selection model is dynamically configured according to the user requirements, the optimal channel is selected from the multiple candidate channels, and the short message channel selection accuracy is improved are achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of information communication, and in particular to a method and device for intelligent and strict selection of short message channels. Background Art

[0002] SMS service is an important information transmission method and is widely used in many scenarios such as personal communication and corporate marketing. In order to improve the delivery rate and timeliness of SMS, SMS service providers will provide multiple SMS channels for users to choose from. However, traditional SMS channel selection usually adopts static rules or strategies, such as selection based on the inherent attributes of the channel, without fully considering the actual needs of users. It is difficult to adapt to the differentiated requirements of different users for SMS services, resulting in low accuracy in SMS channel selection. Summary of the invention

[0003] The present application provides a method and device for intelligent and strict selection of SMS channels, aiming to solve the technical problem that the prior art cannot dynamically adapt to user needs, resulting in low accuracy in SMS channel selection.

[0004] In view of the above problems, the present application provides a method and device for intelligent and strict selection of SMS channels.

[0005] The first aspect disclosed in the present application provides a method for intelligent and strict selection of SMS channels, the method comprising: receiving channel demand information of a target user, and simultaneously collecting channel attribute parameters corresponding to N candidate SMS channels, the channel attribute parameters comprising channel price, delay, arrival rate and channel status parameters; calling a standard SMS channel selection model, configuring the weights of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generating a target SMS channel selection model for the target user; using the target SMS channel selection model to score the N candidate SMS channels, and determining the candidate SMS channel with the highest score as the target SMS channel; and when the target user sends a SMS, sending the SMS sent by the target user through the target SMS channel.

[0006] Another aspect disclosed in the present application provides a device for intelligently and strictly selecting SMS channels, the device comprising: an information acquisition module, for receiving channel demand information of a target user, and simultaneously collecting channel attribute parameters corresponding to N candidate SMS channels, respectively, wherein the channel attribute parameters include channel price, delay, arrival rate and channel status parameters; a model configuration module, for calling a standard SMS channel selection model, configuring the weights of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generating a target SMS channel selection model for the target user; a channel scoring module, for scoring the N candidate SMS channels using the target SMS channel selection model, and determining the candidate SMS channel with the highest score as the target SMS channel; and a SMS sending module, for sending the SMS sent by the target user through the target SMS channel when the target user sends the SMS.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By receiving the channel demand information of the target user and collecting the channel attribute parameters corresponding to N candidate SMS channels at the same time, the channel attribute parameters include channel price, delay, arrival rate and channel status parameters, the personalized needs of the user are obtained, and the multi-dimensional attribute information of the candidate channels is comprehensively collected, which provides a data basis for the subsequent channel selection; the standard SMS channel selection model is called, and the weights of each channel attribute parameter in the standard SMS channel selection model are configured according to the channel demand information to generate a target SMS channel selection model for the target user, so as to achieve the matching of the selection model with the user needs. The standard model provides a general selection framework, and the dynamic adjustment of the weight enables the model to flexibly adapt to the differentiated needs of different users; using the target SMS channel selection model N candidate SMS channels are scored, and the candidate SMS channel with the highest score is determined as the target SMS channel. A customized selection model is used to conduct a multi-dimensional comprehensive evaluation of the candidate channels, fully considering the influence of various attribute parameters, and screening out the channel with the best comprehensive performance. Compared with simple rule screening, it can comprehensively weigh various indicators and make more intelligent and accurate selection decisions; when the target user sends a text message, the technical solution of sending the text message sent by the target user through the target SMS channel solves the technical problem that the existing technology cannot dynamically adapt to user needs, resulting in low accuracy of SMS channel selection, and achieves the technical effect of dynamically configuring the SMS channel selection model according to user needs, selecting the optimal channel from multiple candidate channels, and improving the accuracy of SMS channel selection.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flowchart of a method for intelligent and strict selection of SMS channels is provided for an embodiment of the present application; Figure 2 A structural schematic diagram of a device for intelligent and strict selection of SMS channels is provided for an embodiment of the present application.

[0010] Explanation of reference numerals: information acquisition module 11, model configuration module 12, channel scoring module 13, SMS sending module 14. DETAILED DESCRIPTION

[0011] The overall idea of ​​the technical solution provided by this application is as follows: The embodiments of the present application provide a method and device for intelligent and strict selection of SMS channels. By introducing a standard selection model and dynamically adjusting its attribute weights, a target selection model that is highly matched with user needs is generated, and the model is then used to comprehensively evaluate and optimize candidate channels, thereby significantly improving the accuracy and adaptability of channel selection while meeting the differentiated needs of users.

[0012] First, the channel demand information of the target user is received, and the price, delay, arrival rate, status and other multi-dimensional attribute parameters of the candidate channels are comprehensively collected to obtain the quantitative demand characteristics and channel characteristics. On this basis, the weights of each attribute in the standard SMS channel selection model are dynamically configured to form a highly targeted and distinctive target SMS channel selection model. Then, the SMS channel selection model is used to score the candidate SMS channels one by one, comprehensively weigh the impact of each parameter, accurately select the target SMS channel with the best comprehensive performance, and use it for actual SMS sending.

[0013] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification. Example

[0014] like Figure 1 As shown, the embodiment of the present application provides a method for intelligent and strict selection of SMS channels, the method comprising: S1: receiving channel demand information of a target user, and collecting channel attribute parameters corresponding to N candidate SMS channels respectively, wherein the channel attribute parameters include channel price, delay, arrival rate and channel status parameters.

[0015] Specifically, first, the channel demand information provided by the target user is received. The channel demand information includes but is not limited to the target user's requirements for the performance indicators such as the price, delay, and arrival rate of the SMS channel. Through the channel demand information of the target user, the target user's preferences and expectations for the SMS channel can be understood, laying the foundation for the subsequent generation of a personalized SMS channel selection model. While obtaining the channel demand information of the target user, the channel attribute parameters of N candidate SMS channels are collected. Among them, N is a positive integer, representing the number of candidate SMS channels. The N candidate SMS channels come from different SMS service providers or different channel resources of the same provider. The channel attribute parameters include but are not limited to channel price, delay, arrival rate, and channel status parameters. Among them, the channel price represents the tariff cost of sending SMS; the delay represents the time required for SMS to be sent and received; the arrival rate represents the proportion of successfully delivered SMS; and the channel status parameter represents the current availability and health of the channel. By collecting the channel attribute parameters, the performance and quality of the candidate SMS channels can be comprehensively evaluated.

[0016] By receiving the channel demand information of the target user and collecting the channel attribute parameters of the candidate SMS channels, a data basis is provided for selecting the target SMS channel that meets the user's needs.

[0017] S2: Retrieve a standard SMS channel selection model, configure the weight of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generate a target SMS channel selection model for the target user.

[0018] Specifically, first, a pre-built standard SMS channel selection model is retrieved. The standard model can be trained by a machine learning algorithm based on a large amount of historical SMS sending data and expert experience. The standard SMS channel selection model contains multiple channel attribute parameters, such as channel price, delay, arrival rate and channel status parameters, etc. Each attribute parameter corresponds to a weight coefficient, which reflects the importance of different attribute parameters to the SMS channel selection decision. Then, according to the channel demand information of the target user obtained, the channel attribute parameter weights in the standard SMS channel selection model are dynamically adjusted to generate a target SMS channel selection model. Among them, the specific weight adjustment depends on the actual application scenario and user needs. For example, according to the degree of attention paid by the target user to each channel indicator, the weight of the corresponding attribute parameter is increased, while the weight of other attribute parameters is reduced. If the target user pays attention to SMS delay, the weight coefficient of the delay attribute is increased. If the target user is not sensitive to the channel price, the weight coefficient of the price attribute is reduced. Through weight adjustment, the standard SMS channel selection model is converted into a target SMS channel selection model, which can evaluate the comprehensive performance of the candidate SMS channel according to the actual needs of the target user.

[0019] By calling the standard SMS channel selection model and dynamically adjusting the attribute parameter weights according to the channel demand information of the target user, a target SMS channel selection model personalized for the target user is generated to intelligently select the target SMS channel from the candidate SMS channels and improve the accuracy of SMS channel selection.

[0020] S3: Use the target SMS channel selection model to score the N candidate SMS channels, and determine the candidate SMS channel with the highest score as the target SMS channel.

[0021] Specifically, the channel attribute parameters of the N candidate SMS channels collected are input into the target SMS channel selection model. The target SMS channel selection model comprehensively evaluates the comprehensive score of each candidate SMS channel according to the weight of each attribute parameter. The higher the score, the more the channel meets the channel requirements of the target user. After calculating the comprehensive scores of all candidate SMS channels, the candidate SMS channel with the highest score is determined as the target SMS channel that best meets the requirements of the target user.

[0022] Through the target SMS channel selection model, multiple attribute parameters of candidate SMS channels are comprehensively evaluated, and the target SMS channel is selected to match the most appropriate channel resources for the target users, thereby improving SMS sending effects and user satisfaction.

[0023] S4: When the target user sends a text message, the text message sent by the target user is sent through the target text message channel.

[0024] Specifically, when the target user needs to send a text message, the sending request is associated with the determined target SMS channel. After the association is completed, the text message sent by the target user is imported into the target SMS channel and sent through the channel. The text message sending process can be real-time or delayed. In the real-time sending mode, the text message is sent through the target channel immediately after the target user submits the sending request. In the delayed sending mode, the text message is sent through the target channel at a specific time or after meeting specific conditions, so as to optimize the channel resource utilization and sending effect.

[0025] In the process of SMS sending, a monitoring and feedback mechanism is introduced. For example, the real-time performance indicators of the target SMS channel, such as success rate and latency, are monitored. If the indicators are obviously abnormal, the predefined abnormality handling process is initiated, such as switching to a backup channel, reselecting the channel, etc. Another example is to collect feedback from target users on the effect of SMS sending, which is used to evaluate and optimize the accuracy and effectiveness of the channel selection model.

[0026] By applying the selected target SMS channel to the actual SMS sending, the SMS of the target user can be delivered on demand, thus improving the user experience of the SMS service.

[0027] Furthermore, the embodiment of the present application also includes: Obtaining a historical SMS sending record, wherein the historical SMS sending record includes a plurality of historical SMS sending samples corresponding to N candidate SMS channels, wherein each historical SMS sending sample corresponds to a channel attribute parameter when a candidate SMS channel performs multiple SMS sending within a preset time window; Based on the channel attribute parameters of each historical SMS sending sample, the channel performance score corresponding to the historical SMS sending sample is marked; Summarize the channel attribute parameters and corresponding channel performance scores of all historical SMS sending samples to obtain a model training sample set, wherein the model training sample set includes multiple model training samples corresponding to multiple historical SMS sending samples one by one, and each model training sample includes a channel attribute parameter sample and a channel performance score sample; Based on the model training sample set, with the channel attribute parameter samples as input and the channel performance score samples as supervision targets, a standard SMS channel selection model is constructed.

[0028] In a feasible implementation, first, obtain the historical SMS sending record, which contains the actual sending data of N candidate SMS channels in the historical time, and these data exist in the form of multiple historical SMS sending samples. Each historical SMS sending sample corresponds to the statistical information of multiple SMS sending by a candidate SMS channel within a preset time window (such as 10 minutes, half an hour, etc.), including the average channel attribute parameters within the time window, such as average delay, average arrival rate, etc. By collecting and arranging historical SMS sending samples, the real performance of the candidate SMS channels can be fully reflected. Then, according to the channel attribute parameters contained in each historical SMS sending sample, such as delay, arrival rate, price, etc., according to the user's feedback data, a comprehensive channel performance score is marked to quantify the overall service quality of the channel corresponding to the historical SMS sending sample within the preset time window. The higher the score, the better the channel performance.

[0029] Subsequently, the channel attribute parameters and corresponding channel performance scores of all historical SMS sending samples are summarized to obtain a complete model training sample set. This sample set contains multiple model training samples, each of which corresponds to a historical SMS sending sample, and is composed of the channel attribute parameters (as features) and channel performance scores (as labels) of the historical sample. Next, the obtained model training sample set is used to construct a standard SMS channel selection model. The model uses channel attribute parameters as input features and channel performance scores as supervision targets, and is trained through machine learning algorithms such as decision trees, support vector machines, neural networks, etc. Through training, the model can learn the association between channel attribute parameters and performance scores, so that it has the ability to predict its performance score based on given channel attribute parameters.

[0030] By building a standard SMS channel selection model, we can make full use of the historical performance information of candidate channels, objectively and accurately evaluate channel quality, and provide important decision support for channel selection. At the same time, since the standard SMS channel selection model is trained based on a large amount of historical data, it has good generalization and robustness, and can adapt to different scenarios and user needs.

[0031] Furthermore, the embodiment of the present application also includes: Acquire channel demand information of historical users, and classify the historical users into multiple user categories based on the similarity of preferences of the historical users for channel attributes; For each user category, a corresponding adaptive weight generation model is trained respectively, wherein the adaptive weight generation model takes channel demand information as input and outputs the weight of each channel attribute parameter; Determine the user category to which the target user belongs; Inputting the channel demand information of the target user into the corresponding adaptive weight generation model to generate a target weight coefficient combination, including the weight of each channel attribute parameter; The standard SMS channel selection model is configured according to the target weight coefficient combination to obtain a target SMS channel selection model.

[0032] In a feasible implementation, first, the channel demand information of historical users is obtained, which reflects the preferences and importance of different users for attributes such as SMS channel price, delay, and arrival rate. Then, historical users are divided into multiple user categories based on the similarity of user preferences for channel attributes through clustering algorithms, similarity metrics, and other methods. The purpose of user classification is to group users with similar preferences together to provide a basis for personalized model customization. Then, for each user category, an adaptive weight generation model is trained separately. The model takes channel demand information as input and outputs weight coefficients of each channel attribute parameter. Among them, the adaptive weight generation model is constructed using a neural network. By learning the preference characteristics and rules within the user category, the attribute weights are automatically adjusted to generate a weight distribution scheme that meets the needs of users in this category.

[0033] When a new target user appears, the user category to which he belongs is first determined. Based on the historical behavior data, personal attributes and other characteristics of the target user, the category of the target user is determined through classification algorithms or similarity matching methods, which is the premise for selecting a suitable adaptive weight generation model for him. Then, the channel demand information of the target user is input into the adaptive weight generation model of the corresponding category. After model calculation and reasoning, a set of target weight coefficients is generated to represent the weight distribution of each channel attribute parameter. This set of weight coefficients is dynamically generated based on the personalized needs of the target user and the common characteristics of the category to which it belongs, and can reflect the user's emphasis and preference on different channel attributes. After that, the standard SMS channel selection model is configured and adjusted using the target weight coefficient combination to obtain a target SMS channel selection model customized for the target user. The configuration process includes replacing the weights of each attribute parameter in the standard model with the target weight coefficient, and fine-tuning the model structure and hyperparameters as needed. The adjusted model can evaluate the comprehensive performance of candidate channels and select the target SMS channel according to the specific needs of the target user.

[0034] By introducing user classification and adaptive weight generation mechanism, the SMS channel selection model is personalized and dynamically optimized. The weight generation model is trained for different categories of users, which can automatically learn and adapt to the preferences of user groups. At the same time, according to the actual needs of the target users, the weight coefficient is dynamically generated, and the standard model is configured and adjusted to obtain the target SMS channel selection model. This personalized channel selection method improves the flexibility and adaptability of the model, can more accurately match user needs, select the optimal channel, and thus improve user experience and satisfaction.

[0035] Furthermore, the embodiment of the present application also includes: The adaptive weight generation model is a generative adversarial network model; A generator network encodes the channel demand information of the target user into a latent vector and outputs an initial weight coefficient combination based on the latent vector; A discriminator network outputs a authenticity score based on the initial weight coefficient combination output by the generator network; According to the authenticity score output by the discriminator network, the parameters of the generator network are updated, and the operation is iteratively performed to generate a target weight coefficient combination.

[0036] In a preferred embodiment, the adaptive weight generation model is a generative adversarial network model, which uses the two sub-networks of the generator and the discriminator of the generative adversarial network to automatically optimize and generate a target weight coefficient combination that meets the needs of the target user through adversarial learning.

[0037] First, a generator network is constructed to convert the channel demand information of the target user into a combination of weight coefficients. The generator network encodes the channel demand information of the target user into a low-dimensional latent vector. The encoding process is implemented through embedding layers, encoders, etc., with the aim of extracting the key features and semantic representation of the demand information; then the generator network takes the latent vector as input, and after nonlinear transformation and generation layers, outputs the initial weight coefficient combination, which represents the weight distribution of each channel attribute parameter. At the same time, a discriminator network is constructed to evaluate the authenticity and rationality of the weight coefficient combination output by the generator network. The discriminator network receives the initial weight coefficient combination generated by the generator, extracts and analyzes its features through a multi-layer neural network, and outputs an authenticity score. The higher the score value, the closer the weight coefficient combination is to the real user demand distribution and the more it conforms to the target user's preference characteristics.

[0038] Afterwards, the parameters of the generator network are updated and optimized according to the authenticity score output by the discriminator network. Specifically, by maximizing the objective function of the discriminator score, the gradient of the generator network parameters is calculated using the back propagation algorithm, and the parameters are iteratively updated using optimizers (such as Adam, SGD, etc.). Through continuous iterative training, the generator network can gradually generate more realistic and reasonable weight coefficient combinations, approaching the actual demand distribution of target users; at the same time, the discriminator network is also constantly improving its discriminative ability, prompting the optimization direction of the generator network to develop towards generating higher-quality weight coefficient combinations. After multiple rounds of iterations and adversarial learning, the generator network and the discriminator network finally reach a dynamic balance, and the generator network can generate a weight coefficient combination that meets the characteristics of the target user's needs, that is, the target weight coefficient combination, which can accurately reflect the target user's emphasis and preference on the attribute parameters of each channel, and is used for the subsequent construction of a personalized SMS channel selection model.

[0039] Through the architecture and learning mechanism of the generative adversarial network, the construction of an adaptive weight generation model is realized. Through the mutual game and optimization of the generator network and the discriminator network, a weight coefficient combination that meets the personalized needs of the target users is generated.

[0040] Furthermore, the embodiment of the present application also includes: Obtaining the scheduled time for the target user to send a text message; For the scheduled time, predict the scheduled time channel attribute parameters corresponding to the N candidate SMS channels respectively; Using the target SMS channel selection model to score the N candidate SMS channels based on the predetermined time channel attribute parameters, and determining the candidate SMS channel with the highest score as the predetermined target SMS channel; At the predetermined time, the short message sent by the target user is sent through a predetermined target short message channel.

[0041] In a feasible implementation, a method for preselecting SMS channels for future time points is provided. By predicting channel attribute parameters at future times, the optimal channel is evaluated and selected in advance, and the sending task is automatically executed at the scheduled time, thereby achieving the foresight and adaptability of channel selection.

[0042] First, obtain the scheduled time for the target user to send SMS. The scheduled time is set manually by the user or automatically inferred based on the user's historical sending rules and behavior patterns. The purpose of obtaining the scheduled time is to lock the future sending time point in advance and provide a time reference for advance planning and optimization of channel selection. Then, for the scheduled time, predict the channel attribute parameters of N candidate SMS channels at that time point. Among them, the prediction process uses time series prediction models, such as ARIMA, LSTM, etc., to learn the time trend and change rules of historical channel attribute data, and extrapolate and estimate the attribute values ​​at future time points. The prediction result is a set of channel attribute parameters for each candidate channel at the scheduled time, reflecting the performance of the channel at the future moment.

[0043] Subsequently, the predicted scheduled time channel attribute parameters are input into the target SMS channel selection model to score and sort the N candidate SMS channels. The target SMS channel selection model comprehensively evaluates the expected performance of each channel at the scheduled time based on the weight of each attribute parameter and obtains a sorting result. The candidate SMS channel with the highest score is determined as the scheduled target SMS channel, that is, the channel that is most suitable for sending SMS to the target user at the scheduled time. When the scheduled time arrives, the SMS sent by the target user is sent through the scheduled target SMS channel. Since the channel selection is completed in advance, there is no need to evaluate and switch channels in real time during the sending process, which improves the sending efficiency and user experience.

[0044] By introducing the scheduled time and channel attribute prediction mechanism, advance planning and automatic switching of SMS channels are realized. According to the user's sending time, the future channel attribute parameters are predicted in advance, and the trained channel selection model is used for evaluation and selection to determine the optimal scheduled target channel. When the scheduled time arrives, the sending task is automatically executed without manual intervention, which can make full use of the dynamic change rules of the channel, plan and respond in advance, and improve channel utilization and sending success rate.

[0045] Furthermore, the embodiment of the present application also includes: Acquire a channel attribute parameter record for a historical time period, wherein the channel attribute parameter record for a historical time period includes channel attribute parameters of the N candidate SMS channels in the historical time period; Based on the channel attribute parameter records of the historical time period, a time series prediction model is constructed, wherein the time series prediction model takes any given time point as input and outputs the predicted channel attribute parameters of the N candidate SMS channels at that time point; The scheduled time is input into the time series prediction model, the predicted channel attribute parameters of the N candidate short message channels at the scheduled time are obtained, and the predicted channel attribute parameters are determined as the scheduled time channel attribute parameters.

[0046] In a feasible implementation, first, the channel attribute parameter records of N candidate SMS channels in the historical time period are obtained. The historical time period can be several days, weeks or months, which are set according to the granularity of the data and the time scale of the prediction. The channel attribute parameter records of the historical time period include the price, delay, arrival rate and other indicator data of each candidate SMS channel at different historical time points, reflecting the historical changes in channel performance. Then, based on the obtained channel attribute parameter records of the historical time period, a time series prediction model is constructed. The time series prediction model can characterize the law and trend of the change of channel attribute parameters over time, and predict the attribute values ​​at future time points by learning historical data. The time series prediction model can be a statistical model such as ARIMA, SARIMA, Prophet and a deep learning model such as LSTM, GRU. The input of the model is any given time point, and the output is the predicted channel attribute parameters of the N candidate SMS channels at that time point. The training process of the model continuously optimizes the model parameters by minimizing the error between the predicted value and the true value, thereby improving the accuracy and reliability of the prediction.

[0047] Subsequently, the scheduled time is used as input into the trained time series prediction model to obtain the predicted channel attribute parameters of N candidate SMS channels at the scheduled time. The predicted channel attribute parameters are future attribute values ​​obtained by the model based on historical data and the time characteristics of the scheduled time. These predicted values ​​reflect the expected performance of the candidate channels at the scheduled time, providing an important reference for subsequent channel selection and optimization. The predicted channel attribute parameters are determined as the scheduled time channel attribute parameters, which are used to replace the real-time collected attribute values ​​as the input for channel evaluation selection.

[0048] By introducing the time series prediction model, the prediction and estimation of SMS channel attribute parameters at future time points are realized. By learning the time change rules of historical channel attribute data, a prediction model is constructed, and the channel attributes at a predetermined time are predicted using the model to obtain the channel attribute parameters at the predetermined time. The prediction mechanism can grasp the changing trend of channel performance in advance, providing support for advance planning and optimization of channel selection.

[0049] Furthermore, the embodiment of the present application also includes: Continuously monitoring the real-time channel attribute parameters of the N candidate SMS channels within a preset time period before the scheduled time; Comparing the real-time channel attribute parameters with the predicted predetermined time channel attribute parameters to obtain a channel attribute deviation; If the channel attribute deviation of a candidate SMS channel exceeds a preset threshold, the scheduled channel switching process is triggered, including: Inputting the real-time channel attribute parameters into the target SMS channel selection model, and re-scoring the N candidate SMS channels; If the candidate SMS channel with the highest score is different from the predetermined target SMS channel, the candidate SMS channel with the highest score is determined as a new predetermined target SMS channel; At the predetermined time, the short message sent by the target user is sent through the new predetermined target short message channel.

[0050] In a preferred embodiment, a predetermined channel switching mechanism is provided to deal with the situation where the real-time channel attributes deviate greatly from the predicted values. By continuously monitoring the real-time attribute parameters of the candidate SMS channels and comparing them with the predicted values, when the deviation exceeds the preset threshold, the predetermined channel is re-evaluated and switched to ensure that the optimal channel is used for SMS sending at the scheduled time.

[0051] First, the real-time channel attribute parameters of N candidate SMS channels are continuously monitored for a period of time before the scheduled time (i.e., the preset time period), which reflects the actual performance of the candidate SMS channels at the current moment. Through real-time monitoring, the latest changes in the attributes of the candidate channels are mastered to provide data support for subsequent deviation judgment. Then, the channel attribute parameters collected in real time are compared with the previously predicted scheduled time channel attribute parameters, and the difference between them is calculated to obtain the channel attribute deviation. The channel attribute deviation indicates the degree of deviation between the predicted value and the actual value, reflecting the accuracy of the prediction model and the volatility of the candidate channel performance. Among them, the larger the deviation, the lower the match between the predicted value and the actual situation, which may affect the result of the pre-channel selection.

[0052] Determine whether there is a case where the channel attribute deviation of the candidate SMS channel exceeds the preset threshold. The preset threshold is a critical value set according to historical experience and business tolerance, indicating an acceptable deviation range. If there is a real-time channel attribute parameter of a candidate SMS channel that exceeds the threshold, it is considered that the difference between the predicted value and the actual value is too large, and the scheduled channel switching process is triggered at this time to re-evaluate and adjust the scheduled target SMS channel. First, the real-time channel attribute parameters of the N candidate SMS channels are input into the target SMS channel selection model, the N candidate SMS channels are re-scored, and the comprehensive performance score of each channel is calculated; then, the highest-scoring candidate channel obtained by the re-scoring is compared with the original scheduled target SMS channel. If the two are inconsistent, the candidate channel with the highest score is determined as the new scheduled target SMS channel. When the scheduled time arrives, the SMS sent by the target user is sent through the new scheduled target SMS channel. Since the new scheduled target SMS channel is selected based on the re-evaluation of real-time data, it can meet the user's sending needs with higher actual performance.

[0053] By monitoring the attribute parameters of candidate channels in real time and comparing them with the predicted values, it is determined whether there is a large deviation. When the deviation exceeds the preset threshold, the channel re-evaluation and switching is triggered, and the predetermined target channel is dynamically adjusted to ensure that the real-time optimal channel is used for sending at the scheduled time. It can effectively respond to sudden fluctuations in channel performance, improve the accuracy and reliability of channel selection, and maximize the guarantee that users' SMS sending meets their needs. At the same time, the setting of the preset time period also reserves buffer time for channel switching, avoiding the impact of temporary switching.

[0054] In summary, the method for intelligent and strict selection of SMS channels provided in the embodiments of the present application has the following technical effects: Receive the channel demand information of the target user, and collect the channel attribute parameters corresponding to N candidate SMS channels. The channel attribute parameters include channel price, delay, arrival rate and channel status parameters. By receiving user needs, clarify the user's specific requirements and preferences for SMS channels. By collecting channel attribute parameters, comprehensively grasp the various performance indicators of candidate channels, and provide a data basis for subsequent channel selection. Retrieve the standard SMS channel selection model, configure the weights of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generate a target SMS channel selection model for the target user. By dynamically configuring the weights of the attribute parameters in the standard model, it matches the user's channel needs and generates a highly targeted target SMS channel selection model. Use the target SMS channel selection model to score the N candidate SMS channels, determine the candidate SMS channel with the highest score as the target SMS channel, and score each candidate channel using the target SMS channel selection model to quantitatively measure the degree of matching with user needs. When the target user sends a text message, send the text message sent by the target user through the target SMS channel. By importing the user's text messages into the selected target SMS channel, ensure the on-demand transmission of the user's text messages. Compared with static selection, dynamically matched channels can better adapt to changes in user needs and channel status, and continuously optimize SMS sending effects. Example

[0055] Based on the same inventive concept as the method for intelligent and strict selection of SMS channels in the aforementioned embodiment, Figure 2 As shown, an embodiment of the present application provides a device for intelligent and strict selection of SMS channels, the device comprising: The information acquisition module 11 is used to receive the channel demand information of the target user and collect the channel attribute parameters corresponding to the N candidate SMS channels respectively, wherein the channel attribute parameters include channel price, delay, arrival rate and channel status parameters; The model configuration module 12 is used to call the standard SMS channel selection model, configure the weight of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generate a target SMS channel selection model for the target user; A channel scoring module 13 is used to score the N candidate SMS channels using the target SMS channel selection model, and determine the candidate SMS channel with the highest score as the target SMS channel; The short message sending module 14 is used for sending the short message sent by the target user through the target short message channel when the target user sends a short message.

[0056] Furthermore, the embodiment of the present application also includes a model building module, which includes the following execution steps: Obtaining a historical SMS sending record, wherein the historical SMS sending record includes a plurality of historical SMS sending samples corresponding to N candidate SMS channels, wherein each historical SMS sending sample corresponds to a channel attribute parameter when a candidate SMS channel performs multiple SMS sending within a preset time window; Based on the channel attribute parameters of each historical SMS sending sample, the channel performance score corresponding to the historical SMS sending sample is marked; Summarize the channel attribute parameters and corresponding channel performance scores of all historical SMS sending samples to obtain a model training sample set, wherein the model training sample set includes multiple model training samples corresponding to multiple historical SMS sending samples one by one, and each model training sample includes a channel attribute parameter sample and a channel performance score sample; Based on the model training sample set, with the channel attribute parameter samples as input and the channel performance score samples as supervision targets, a standard SMS channel selection model is constructed.

[0057] Furthermore, the model configuration module 12 includes the following execution steps: Acquire channel demand information of historical users, and classify the historical users into multiple user categories based on the similarity of preferences of the historical users for channel attributes; For each user category, a corresponding adaptive weight generation model is trained respectively, wherein the adaptive weight generation model takes channel demand information as input and outputs the weight of each channel attribute parameter; Determine the user category to which the target user belongs; Inputting the channel demand information of the target user into the corresponding adaptive weight generation model to generate a target weight coefficient combination, including the weight of each channel attribute parameter; The standard SMS channel selection model is configured according to the target weight coefficient combination to obtain a target SMS channel selection model.

[0058] Furthermore, the model configuration module 12 also includes the following execution steps: A generator network encodes the channel demand information of the target user into a latent vector and outputs an initial weight coefficient combination based on the latent vector; A discriminator network outputs a authenticity score based on the initial weight coefficient combination output by the generator network; According to the authenticity score output by the discriminator network, the parameters of the generator network are updated, and the operation is iteratively performed to generate a target weight coefficient combination.

[0059] Furthermore, the embodiment of the present application also includes a channel reservation module, which includes the following execution steps: Obtaining the scheduled time for the target user to send a text message; For the scheduled time, predict the scheduled time channel attribute parameters corresponding to the N candidate SMS channels respectively; Using the target SMS channel selection model to score the N candidate SMS channels based on the predetermined time channel attribute parameters, and determining the candidate SMS channel with the highest score as the predetermined target SMS channel; At the predetermined time, the short message sent by the target user is sent through a predetermined target short message channel.

[0060] Furthermore, the channel reservation module further includes the following execution steps: Acquire a channel attribute parameter record for a historical time period, wherein the channel attribute parameter record for a historical time period includes channel attribute parameters of the N candidate SMS channels in the historical time period; Based on the channel attribute parameter records of the historical time period, a time series prediction model is constructed, wherein the time series prediction model takes any given time point as input and outputs the predicted channel attribute parameters of the N candidate SMS channels at that time point; The scheduled time is input into the time series prediction model, the predicted channel attribute parameters of the N candidate short message channels at the scheduled time are obtained, and the predicted channel attribute parameters are determined as the scheduled time channel attribute parameters.

[0061] Furthermore, the embodiment of the present application also includes a predetermined channel switching module, which includes the following execution steps: Continuously monitoring the real-time channel attribute parameters of the N candidate SMS channels within a preset time period before the scheduled time; Comparing the real-time channel attribute parameters with the predicted predetermined time channel attribute parameters to obtain a channel attribute deviation; If the channel attribute deviation of a candidate SMS channel exceeds a preset threshold, the scheduled channel switching process is triggered, including: Inputting the real-time channel attribute parameters into the target SMS channel selection model, and re-scoring the N candidate SMS channels; If the candidate SMS channel with the highest score is different from the predetermined target SMS channel, the candidate SMS channel with the highest score is determined as a new predetermined target SMS channel; At the predetermined time, the short message sent by the target user is sent through the new predetermined target short message channel.

[0062] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0063] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these modifications and variations.

Claims

1. A method for intelligent and strict selection of SMS channels, characterized in that: The method comprises: Receive channel demand information of target users, and simultaneously collect channel attribute parameters corresponding to N candidate SMS channels, wherein the channel attribute parameters include channel price, delay, arrival rate and channel status parameters; Retrieving a standard SMS channel selection model, configuring the weight of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generating a target SMS channel selection model for a target user; Using the target SMS channel selection model to score the N candidate SMS channels, and determining the candidate SMS channel with the highest score as the target SMS channel; When the target user sends a text message, the text message sent by the target user is sent through the target text message channel.

2. The method for intelligent and strict selection of short message channels according to claim 1, characterized in that: The method further comprises: Obtaining a historical SMS sending record, wherein the historical SMS sending record includes a plurality of historical SMS sending samples corresponding to N candidate SMS channels, wherein each historical SMS sending sample corresponds to a channel attribute parameter when a candidate SMS channel performs multiple SMS sending within a preset time window; Based on the channel attribute parameters of each historical SMS sending sample, the channel performance score corresponding to the historical SMS sending sample is marked; Summarize the channel attribute parameters and corresponding channel performance scores of all historical SMS sending samples to obtain a model training sample set, wherein the model training sample set includes multiple model training samples corresponding to multiple historical SMS sending samples one by one, and each model training sample includes a channel attribute parameter sample and a channel performance score sample; Based on the model training sample set, with the channel attribute parameter samples as input and the channel performance score samples as supervision targets, a standard SMS channel selection model is constructed.

3. The method for intelligent and strict selection of short message channels according to claim 1, characterized in that: According to the channel demand information, the weight of each channel attribute parameter in the standard SMS channel selection model is configured to generate a target SMS channel selection model for the target user, including: Acquire channel demand information of historical users, and classify the historical users into multiple user categories based on the similarity of preferences of the historical users for channel attributes; For each user category, a corresponding adaptive weight generation model is trained respectively, wherein the adaptive weight generation model takes channel demand information as input and outputs the weight of each channel attribute parameter; Determine the user category to which the target user belongs; Inputting the channel demand information of the target user into the corresponding adaptive weight generation model to generate a target weight coefficient combination, including the weight of each channel attribute parameter; The standard SMS channel selection model is configured according to the target weight coefficient combination to obtain a target SMS channel selection model.

4. The method for intelligent and strict selection of short message channels according to claim 3, characterized in that: The adaptive weight generation model is a generative adversarial network model, comprising: A generator network encodes the channel demand information of the target user into a latent vector and outputs an initial weight coefficient combination based on the latent vector; A discriminator network outputs a authenticity score based on the initial weight coefficient combination output by the generator network; According to the authenticity score output by the discriminator network, the parameters of the generator network are updated, and the operation is iteratively performed to generate a target weight coefficient combination.

5. The method for intelligent and strict selection of short message channels according to claim 1, characterized in that: The method further comprises: Obtaining the scheduled time for the target user to send a text message; For the scheduled time, predict the scheduled time channel attribute parameters corresponding to the N candidate SMS channels respectively; Using the target SMS channel selection model to score the N candidate SMS channels based on the predetermined time channel attribute parameters, and determining the candidate SMS channel with the highest score as the predetermined target SMS channel; At the predetermined time, the short message sent by the target user is sent through a predetermined target short message channel.

6. The method for intelligent and strict selection of short message channels according to claim 5, characterized in that: For the scheduled time, predicting the scheduled time channel attribute parameters corresponding to the N candidate SMS channels respectively includes: Acquire a channel attribute parameter record for a historical time period, wherein the channel attribute parameter record for a historical time period includes channel attribute parameters of the N candidate SMS channels in the historical time period; Based on the channel attribute parameter records of the historical time period, a time series prediction model is constructed, wherein the time series prediction model takes any given time point as input and outputs the predicted channel attribute parameters of the N candidate SMS channels at that time point; The scheduled time is input into the time series prediction model, the predicted channel attribute parameters of the N candidate short message channels at the scheduled time are obtained, and the predicted channel attribute parameters are determined as the scheduled time channel attribute parameters.

7. The method for intelligent and strict selection of short message channels according to claim 6, characterized in that: The method further comprises: Continuously monitoring the real-time channel attribute parameters of the N candidate SMS channels within a preset time period before the scheduled time; Comparing the real-time channel attribute parameters with the predicted predetermined time channel attribute parameters to obtain a channel attribute deviation; If the channel attribute deviation of a candidate SMS channel exceeds a preset threshold, the scheduled channel switching process is triggered, including: Inputting the real-time channel attribute parameters into the target SMS channel selection model, and re-scoring the N candidate SMS channels; If the candidate SMS channel with the highest score is different from the predetermined target SMS channel, the candidate SMS channel with the highest score is determined as a new predetermined target SMS channel; At the predetermined time, the short message sent by the target user is sent through the new predetermined target short message channel.

8. A device for intelligent and strict selection of SMS channels, characterized in that: The method for implementing the intelligent and strict selection of SMS channels according to any one of claims 1 to 7 comprises: An information acquisition module, the information acquisition module is used to receive channel demand information of the target user, and simultaneously collect channel attribute parameters corresponding to N candidate SMS channels, the channel attribute parameters including channel price, delay, arrival rate and channel status parameters; A model configuration module, the model configuration module is used to call a standard SMS channel selection model, configure the weight of each channel attribute parameter in the standard SMS channel selection model according to the channel demand information, and generate a target SMS channel selection model for a target user; A channel scoring module, the channel scoring module is used to score the N candidate SMS channels using the target SMS channel selection model, and determine the candidate SMS channel with the highest score as the target SMS channel; The short message sending module is used for sending the short message sent by the target user through the target short message channel when the target user sends a short message.

Citation Information

Cited By

  • Big data analysis method and platform for short message sending success rate of multi-region Internet of Vehicles

    CN119545393A

  • Data processing method and system, electronic equipment and storage medium

    CN119782364A

  • A data processing method and system, electronic equipment, and storage medium

    CN119782364B

  • Multi-operator short message access priority selection system and method based on position

    CN119946566A

  • Short message sending channel intelligent scheduling method based on multi-source data fusion

    CN120151781A