Marketing short message route selection and flow control method and system

Abstract

The invention provides a marketing short message routing selection and flow control method and system, and the method comprises the steps: collecting historical sending records and supplementary data source data, carrying out the preprocessing and feature engineering of the collected data, and forming a unified data set; constructing a route evaluation model based on the unified data set, and selecting an optimal route based on the route evaluation model; outputting an optimal route selection result, and performing flow distribution adjustment according to the sending time delay, the delivery rate and the load rate data of each route acquired in real time; and establishing a closed-loop optimization mechanism, adding latest short message sending data, service provider management and control data and short message feature data into training data, updating a route evaluation model, ensuring that the optimal route evaluation model adapts to the change of the latest service provider service level, and forming linkage with dynamic flow control strategy optimization. According to the method, the routing evaluation model is constructed through multi-dimensional data fusion, dynamic flow distribution is realized in combination with sending time delay constraint, and the success rate of marketing short message delivery is improved.

Description

Technical Field

[0001] This invention relates to the field of SMS communication technology, and more specifically, to a method and system for routing and traffic control of marketing SMS messages. Background Technology

[0002] With the popularization of digital marketing, marketing SMS has become one of the core channels for enterprises to reach users. SMS dispatch systems need to distribute massive amounts of marketing SMS to different SMS service providers (routers) to complete the delivery. According to industry statistics, the daily average volume of marketing SMS processed by mainstream domestic SMS dispatch systems can reach tens of millions of messages. However, there are significant differences in the list management and sensitive word control strategies of different SMS service providers, and the service level of service providers (such as delivery rate, latency, and stability) fluctuates dynamically with time and traffic changes.

[0003] Current SMS dispatch systems lack scientific and precise solutions for routing and traffic control. Most rely on manual experience, fixed rules, or simple historical sending record statistics, failing to fully utilize multi-dimensional data to achieve optimal decision-making. This results in low success rate of marketing SMS delivery and excessive sending latency, affecting marketing effectiveness while increasing marketing and communication costs for enterprises.

[0004] In existing technologies, there are two main solutions for routing and traffic control of marketing SMS messages: Patent CN 11668970 A discloses a method for controlling SMS traffic, a method for generating SMS messages, and related devices. It achieves control by collecting historical delivery rate data from SMS service providers and combining it with real-time traffic and sending delay, but it does not consider the differences in service providers' list management and sensitive word management.

[0005] Some existing SMS dispatch systems use a fixed routing allocation strategy, which distributes marketing SMS messages to different service providers at a fixed ratio or randomly selects routes. They do not have a dynamic traffic control mechanism, which cannot adapt to the dynamic fluctuations in the service level of service providers. Furthermore, when a service provider experiences congestion or upgrades its management, it cannot adjust the traffic allocation in a timely manner, resulting in excessive SMS sending latency and a sharp drop in delivery rate.

[0006] Furthermore, existing technologies have not fully explored effective data sources other than historical transmission records. The utilization of historical transmission records is limited to simple statistics and does not involve in-depth analysis based on multi-dimensional features. This results in routing and traffic control schemes being close to "black boxes," making it impossible to explain the decision-making logic and achieve continuous optimization. Summary of the Invention

[0007] This invention addresses the technical problems existing in the prior art by providing a method and system for routing and traffic control of marketing SMS messages. It is applicable to SMS scheduling systems for accurate selection of routes from multiple service providers and dynamic traffic management, thereby improving the delivery success rate of marketing SMS messages and adapting to various scenarios that require large-scale sending of marketing SMS messages.

[0008] According to a first aspect of the present invention, a method for routing and traffic control of marketing text messages is provided, comprising: Collect historical transmission records and supplementary data sources, preprocess and feature-engineer the collected data to form a unified dataset; Based on the preprocessed unified dataset, a route evaluation model is constructed, and the optimal route is selected based on the route evaluation model; Establish a dynamic flow control mechanism based on latency constraints, output the optimal routing selection result, and adjust the flow allocation according to the real-time collected data on the sending latency, delivery rate and load rate of each route; Establish a closed-loop optimization mechanism by adding the latest SMS sending data, service provider management data, and SMS characteristic data to the training data, updating the routing evaluation model, ensuring that the optimal routing evaluation model adapts to the latest changes in service provider service levels, and forming a linkage with dynamic traffic control strategy optimization.

[0009] Based on the above technical solution, the present invention can also be improved as follows.

[0010] Optionally, the collection of the historical sending record data includes: collecting marketing SMS sending data within the past 3 months, including but not limited to: SMS content, sending time, target user number, route, delivery result, sending delay, failure reason and traffic allocation ratio; The supplementary data sources collected include: the list control rules, sensitive word control rules, service commitments and real-time service status of each SMS service provider, as well as the content type and target user characteristics of marketing SMS messages.

[0011] Optionally, the preprocessing of the collected data includes: removing invalid data, segmenting text data into words, constructing a bag-of-words model for documents, and normalizing numerical data; the feature engineering includes: establishing SMS features, target user features, association features, and indicator features.

[0012] Optionally, constructing the route evaluation model based on the preprocessed unified dataset includes: Using historical transmission records and service provider management data as training samples, the optimal route label is labeled, and a random forest-based route evaluation model is trained. The model performance is verified through a test set to ensure that the route selection accuracy is greater than the threshold. Construct a routing evaluation index system, identify core evaluation indicators including delivery performance indicators, management and control adaptation indicators, service stability indicators, and cost indicators, and dynamically adjust the weights according to business needs.

[0013] Optionally, the selection of the optimal route based on the route evaluation model includes: For marketing SMS messages to be sent, input SMS characteristics, target user characteristics, and relationship characteristics data. Calculate the evaluation score of each available route through the routing evaluation model, and output the optimal route selection result according to the score.

[0014] Optionally, the dynamic flow control mechanism includes: Based on the service provider's service commitments, real-time load rate, and historical latency data for each route, calculate the maximum traffic threshold that each route can carry. Initial traffic is allocated proportionally based on the evaluation scores of each route; routes with higher scores receive a higher proportion of initial traffic. Real-time collection of transmission latency, delivery rate, and load rate data for each route enables dynamic adjustment of route traffic; Ensure that the total traffic allocation for all routes equals the total traffic of the marketing SMS messages to be sent, and that the traffic allocation ratio for a single route is not lower than the minimum allocation ratio.

[0015] Optionally, the real-time collection of transmission latency, delivery rate, and load rate data for each route, and the dynamic adjustment of route traffic, includes: If the real-time transmission latency of a certain route exceeds the preset maximum latency or the load rate exceeds the threshold configured at the time of access, the traffic allocation ratio of that route will be automatically reduced, and the reduced traffic will be allocated to routes with higher evaluation scores, normal latency, and lower load. If the delivery rate of a route drops below the minimum allocation ratio for several consecutive periods, traffic allocation for that route will be suspended until its service status is restored, at which point traffic will be gradually restored.

[0016] Optionally, the establishment of the closed-loop optimization mechanism includes: Collect the actual results of each transmission and the routing service status data, compare the prediction results of the routing evaluation model with the actual results, and calculate the deviation. Based on the deviation data, the indicator weights and model parameters of the route evaluation model are dynamically adjusted to optimize the route selection logic; Based on historical traffic adjustment records and sending results, optimize the initial traffic allocation ratio and adjust thresholds to adapt to dynamic changes in the service provider's service level. We update service provider management rules and environmental data daily, and update historical sending records weekly to support model and strategy optimization.

[0017] Optionally, the latest SMS sending data includes: routing features and sending results, wherein the sending results include: delivery rate, latency, and failure reasons; actual sending results and routing service status data are collected, and the prediction results of the optimal routing evaluation model are compared with the actual results to provide sample data support for continuous iteration of routing selection and optimization of model parameters.

[0018] According to a second aspect of the present invention, a system for routing and traffic control of marketing text messages is provided, comprising: The data acquisition module is used to collect historical transmission records and supplementary data sources, and to preprocess the collected data to form a unified dataset. The route evaluation model building module is used to build a route evaluation model based on a preprocessed unified dataset. The optimal route selection module is used to select the optimal route based on the route evaluation model; The dynamic flow control module is used to output the optimal routing selection result and adjust the flow allocation based on the real-time collected data on the transmission latency, delivery rate and load rate of each route. The closed-loop optimization module is used to add the latest SMS sending data, service provider management data, and SMS feature data to the training data, update the routing evaluation model, and ensure that the optimal routing evaluation model adapts to the latest changes in service provider service levels, thus forming a linkage with the dynamic traffic control strategy optimization.

[0019] The technical effects and advantages of this invention are as follows: This invention provides a method and system for routing and traffic control of marketing SMS messages. It constructs a routing evaluation model through multi-dimensional data fusion, deeply mines multi-dimensional features of historical sending records, and combines this with other supplementary data sources to improve the accuracy of routing and traffic control, enabling continuous optimization of the solution. By incorporating sending latency constraints, it achieves dynamic traffic allocation, improving routing accuracy and delivery success rate, controlling sending latency within limits, reducing marketing costs, and ensuring explainable decisions and optimizable strategies. A closed-loop optimization mechanism is established to continuously improve the accuracy of routing and traffic control. This improves the delivery success rate of marketing SMS messages and adapts to various scenarios requiring large-scale marketing SMS sending. Attached Figure Description

[0020] Figure 1 A flowchart illustrating the steps of a method for routing and traffic control of marketing SMS messages provided in an embodiment of the present invention; Figure 2 A flowchart of the routing evaluation model provided in an embodiment of the present invention; Figure 3 A flowchart of dynamic flow control provided in an embodiment of the present invention; Figure 4A flowchart illustrating the closed-loop optimization mechanism provided in this embodiment of the invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that the embodiments of the present invention address the core deficiencies existing in the background art and solve the following technical problems: 1. Solve the "black box" problem of routing selection in SMS dispatch systems, break through the limitations of relying solely on human experience or single historical data, identify effective data sources other than historical sending records, establish a multi-dimensional data fusion routing evaluation system, and achieve interpretable and traceable selection of the optimal route; 2. Address the issue of insufficient routing adaptability caused by differences in list management and sensitive word control strategies among different service providers, and improve the delivery success rate of marketing SMS by matching routes based on the characteristics of management strategies; 3. To address the issues of lack of dynamism in traffic control and inability to adapt to fluctuations in service provider service levels, a dynamic traffic allocation mechanism based on sending latency constraints is established to achieve precise matching between traffic and routing service capabilities without exceeding the maximum sending latency. 4. To address the issue of insufficient utilization of historical sending records, by deeply mining the multi-dimensional characteristics of historical sending records (such as time period, user group, and SMS content), and combining them with other supplementary data sources, the accuracy of routing selection and traffic control can be improved, enabling continuous optimization of the solution.

[0023] Understandably, given the deficiencies in the background technology, this invention proposes a method for routing and traffic control of marketing SMS messages, specifically as follows: Figure 1 As shown, it includes the following steps: Step S1: Collect historical transmission records and supplementary data sources, preprocess and feature-engineer the collected data to form a unified dataset; In this embodiment, collecting the historical sending record data includes: collecting marketing SMS sending data within the past 3 months, including but not limited to: SMS content, sending time (accurate to the minute), target user number (encrypted), route (SMS service provider), delivery result (success / failure / delay), sending delay, reason for failure (list management blocking / sensitive word blocking / network congestion, etc.), and traffic allocation ratio.

[0024] In addition to historical transmission records, the supplementary data sources collected include: Service provider management data: Collect list management rules (such as blacklist scope, user location restrictions), sensitive word management rules (sensitive word library version, interception threshold), service commitments (maximum sending delay, minimum delivery rate), and real-time service status (online / offline, current load rate) of each SMS service provider. SMS feature data: Collect the content type of marketing SMS messages (such as promotional notices, event invitations, and benefit reminders) and target user characteristics (location, historical reception feedback).

[0025] Preprocessing of the collected data includes: Remove invalid data (such as empty numbers and duplicate sending records), perform word segmentation and construct a bag-of-words model for text data (SMS content and control rules), and normalize numerical data (latency and load rate).

[0026] Feature engineering includes: establishing SMS features, target user features, relationship features (such as the matching relationship between SMS features and service provider control rules), and indicator features. Among these, SMS Features: Construct SMS content type feature vectors, determine vector dimensions (content type, such as promotional notifications, event invitations, benefit reminders), and count the number of messages sent for each dimension (example data, such as 1200 promotional notifications, 300 event invitations, and 500 benefit reminders). Construct feature vectors (in a fixed dimension order, using the number of messages sent as the corresponding dimension value to form a feature vector of [1200, 300, 500] or according to the percentage of messages sent, forming a feature vector of [0.6, 0.15, 0.25]). Target user characteristics: Construct a target user location feature vector, determine the vector dimensions (provincial administrative divisions, such as Beijing, Shanghai, Guangdong, Jiangsu, Zhejiang, Shandong, etc.), and count the number of messages sent for each dimension (example data, such as Beijing: 1200 messages, Shanghai: 950 messages, Guangdong: 1800 messages, Jiangsu: 1100 messages, Zhejiang: 850 messages, Shandong: 700 messages). Construct feature vectors (by using the number of messages sent as the corresponding dimension values ​​in a fixed dimension order to form feature vectors, or by forming feature vectors according to the proportion of messages sent). Construct a feature vector of historical received feedback from the target user, determine the vector dimensions (received feedback, such as successful delivery, failed delivery, unknown), and count the number of messages sent for each dimension (example data, such as 8000 successful deliveries, 1500 failed deliveries, and 500 unknown messages). Construct a feature vector (in a fixed dimension order, using the number of messages sent as the corresponding dimension value to form a feature vector of [8000, 1500, 500] or according to the percentage of messages sent, forming a feature vector of [0.8, 0.15, 0.05]). The correlation characteristics include: the matching degree between SMS content and the service provider's sensitive word control rules; determining the vector dimensions (sensitive word hit status, such as service provider 1 - sensitive word hit, service provider 2 - sensitive word hit, service provider 3 - sensitive word hit, etc.); counting the number of messages sent in each dimension (example data, such as service provider 1 - sensitive word hit 120 messages, service provider 2 - sensitive word hit 90 messages, service provider 3 - sensitive word hit 150 messages, etc.); constructing the feature vector (in a fixed dimension order, the number of sensitive word hits is used as the corresponding dimension value to form a feature vector of [120, 90, 150, ... ...], which is then normalized to form the feature vector; The matching degree between target users and service provider list control rules is used to determine vector dimensions (list hit status, such as service provider 1 - list hit, service provider 2 - list hit, service provider 3 - list hit, etc.). The sending volume of each dimension is counted (example data, such as service provider 1 - 120 list hits, service provider 2 - 140 list hits, service provider 3 - 70 list hits, etc.). A feature vector is constructed (in a fixed dimension order, the list hit count is used as the corresponding dimension value to form a feature vector of [120, 140, 70, ..., etc.], which is then normalized to form the feature vector. The key performance indicators include: delivery performance indicators, including historical delivery rate, failure rate (categorized by failure cause), and average transmission latency; Service stability metrics: online rate, load rate fluctuation, and latency fluctuation of the service provider over the past 30 days; Cost metrics: The service provider's pricing standard for SMS sending.

[0027] Step S2: Based on the preprocessed unified dataset, construct a route evaluation model and select the optimal route based on the route evaluation model; like Figure 2 The routing evaluation model shown is divided into two stages: offline training and online evaluation. Both stages require data acquisition, data preprocessing, and feature engineering in step S1.

[0028] The online evaluation includes: collecting SMS data to be sent from the SMS dispatch system within the current time window; collecting service provider management data, such as list management rules, sensitive word management rules, service commitments, and real-time service status; generating SMS feature data, collecting content type, target user's collection location, historical received feedback, and associating it with customer profiles to obtain activity levels; performing word segmentation on text types, constructing a bag-of-words model, and normalizing numerical data; establishing SMS features, target user features, relationship features, and indicator features; and inputting the data into a multi-label regression random forest model to obtain routing regression scores for each service provider.

[0029] Offline training includes: collecting marketing SMS sending data from the past three months, dividing it into different sample sets by hour; collecting service provider management data, such as: list management rules, sensitive word management rules, service commitments, and real-time service status; generating SMS feature data, collecting content type, target user collection location, historical received feedback, and associating it with customer profiles to obtain activity levels; removing invalid data, performing word segmentation on text types, constructing a bag-of-words model, and normalizing numerical data; generating SMS sending plans, SMS features, and target user features within the current time window; and obtaining the delivery metrics (standardized values ​​for delivery rate and average sending latency) of each service provider within the corresponding time window, weighted and summed to calculate the Y-value for model training, and training a multi-label regression random forest model.

[0030] Note: Standardized delivery rate = Number of successfully delivered items on this route / Total number of items sent on this route The normalized average transmission delay value = (maximum delay threshold of the route - actual average transmission delay) / maximum delay threshold.

[0031] The constructed routing evaluation model includes: Construct a routing evaluation index system and determine the following four core evaluation indexes, and dynamically adjust the weights according to business needs.

[0032] Delivery performance metrics (weight 40%) include historical delivery rate, failure rate (categorized by failure cause), and average transmission latency. Control and adaptation indicators (weight 30%): the degree of matching between SMS content and the service provider's sensitive word control rules, and the degree of matching between target users and the service provider's list control rules; Service stability metrics (weight 20%): Service provider's online rate, load rate fluctuation, and latency fluctuation over the past 30 days; Cost metrics (weight 10%): Service provider's SMS sending unit price and data usage overage billing standards.

[0033] The optimal route selection based on the route evaluation model includes: For marketing SMS messages to be sent, input data such as SMS characteristics, target user characteristics, and relationship characteristics. The routing evaluation model calculates the evaluation score of each available route and outputs the optimal route selection result according to the score.

[0034] Step S3: Establish a dynamic flow control mechanism, including: outputting the optimal route selection result and adjusting the flow allocation based on the real-time collected data on the sending delay, delivery rate and load rate of each route; Dynamic flow control mechanisms such as Figure 3 As shown, it is divided into: determining the preset traffic threshold for routing, initial traffic allocation, initial traffic adjustment, traffic distribution, and real-time traffic adjustment; among which, Determining the traffic threshold for routes: Based on the service provider's service commitment (maximum sending capacity), real-time load rate, and historical latency data for each route, calculate the maximum traffic threshold that each route can carry to ensure that traffic allocation does not exceed the threshold and avoid network congestion and excessive latency. Initial traffic allocation: Based on the evaluation scores of each route output by the optimal route selection, the initial traffic is allocated according to preset rules. The higher the score of the route, the higher the initial traffic allocation ratio (e.g., the highest-scoring route is allocated 40% of the traffic, the second highest-scoring route is allocated 30%, and so on). Initial traffic adjustment: Adjustments are made in conjunction with SMS sending plans and route traffic thresholds. The traffic allocation is checked from high to low to see if it exceeds the threshold. If it exceeds 80% of the route threshold, the traffic allocation ratio of that route is automatically reduced to 80% of the threshold, and the reduced traffic is allocated to the next route. Real-time traffic adjustment: Real-time collection of transmission latency, delivery rate, and load rate data for each route. If the real-time transmission latency of a route exceeds the preset maximum latency, or the load rate exceeds 80%, the traffic allocation ratio of that route will be automatically reduced (5%-10% each time), and the reduced traffic will be allocated to routes with higher evaluation scores, normal latency, and lower load. If the delivery rate of a route continues to decline (decreases by more than 5% for 10 consecutive minutes), traffic allocation for that route will be suspended, and traffic will be gradually restored after its service status is restored. Traffic allocation constraints: Ensure that the total traffic allocation for all routes equals the total traffic of the marketing SMS messages to be sent, and that the traffic allocation ratio for a single route is not less than 5% (to avoid excessive traffic concentration leading to risks).

[0035] Step S4: Establish a closed-loop optimization mechanism by adding the latest SMS sending data, service provider management data, and SMS feature data to the training data, updating the routing evaluation model, and ensuring that the optimal routing evaluation model adapts to the latest changes in service provider service levels, thus linking with the optimization of dynamic traffic control strategies.

[0036] The closed-loop optimization mechanism is as follows: Figure 4 As shown, to adapt to dynamic changes in service provider service levels, a closed-loop optimization mechanism is established to continuously improve the accuracy of routing selection and traffic control. The closed-loop optimization mechanism includes: Real-time data feedback: Collect the actual results of each transmission (delivery rate, latency, failure reason) and routing service status data, compare the prediction results of the routing evaluation model with the actual results, and calculate the deviation; Model parameter adjustment: Based on the deviation data, dynamically adjust the indicator weights and model parameters of the route evaluation model to optimize the route selection logic; Traffic control strategy optimization: Based on historical traffic adjustment records and sending effects, optimize the initial traffic allocation ratio and adjustment threshold (such as the traffic adjustment range after latency exceeds the standard) to adapt to the dynamic changes in the service provider's service level. Data source updates: Daily updates of service provider management rules and environmental data, and weekly updates of historical transmission records to ensure data timeliness and provide support for model and strategy optimization.

[0037] It should be noted that the actual sending results (delivery rate, latency, and reasons for failure) and routing service status data are collected and compared with the prediction results of the optimal routing evaluation model to provide basic data support for the iteration and subsequent optimization of the routing selection algorithm.

[0038] In this embodiment, the continuous iteration of the routing algorithm is divided into three core stages: iterative training, iterative verification, and iterative deployment, thus forming a complete iterative closed loop. Iterative training: Performed by the model training module, it adopts a combination of "full retraining + incremental supplementation" to conduct iterative training periodically. The specific operation is as follows: Iteration cycle: The basic iteration cycle is once a week. If the model prediction error is ≥15% (quantization threshold), an emergency iteration will be triggered (started within 24 hours) to ensure that the algorithm can quickly adapt to sudden changes in the service level of the service provider. Training sample construction: Based on the updated historical data of the past 90 days, the latest collected samples (from the past 7 days) are added to construct an iterative training sample set; Iterative Validation: After iterative training is completed, the model validation module executes the validation process to ensure that the new iteration of the random forest routing algorithm is superior to the original algorithm. Validation is divided into two stages: offline validation and online gray-scale validation. The specific operations are as follows: Offline validation: Use the updated samples that have not participated in training in the last 7 days as the validation set to validate the core metrics of the new algorithm; if the above criteria are not met, return to step 3.1, readjust the parameters and iteratively train; if the criteria are met, proceed to the online gray-scale validation stage. Online gray-scale verification: Deploy the newly iterated optimal routing algorithm to the test environment, allocate 10% of the online SMS sending traffic to use this algorithm for routing, while the original algorithm is still used for 90% of the traffic, and continuously monitor for 24 hours; monitoring indicators include: routing accuracy, SMS delivery rate, average latency, and failure rate. If the new algorithm improves the delivery rate by ≥5%, reduces the average latency by ≥100ms, and reduces the failure rate by ≥3%, and there are no abnormal fluctuations (such as frequent switching of routing), the verification is successful; if the standard is not met, iterate again; Iterative Deployment: After successful online gray-scale validation, the iterative algorithm deployment process is initiated, with the specific steps as follows: Full deployment: Replace the original model with the new iteration of the optimal routing algorithm model and deploy it to the production environment. Link the routing algorithm with the traffic control module through the API interface to ensure that the routing algorithm and traffic control strategy work together. Production monitoring: After going live, continuous monitoring for 72 hours, real-time tracking of the algorithm's routing selection effect and various core indicators (delivery rate, latency, failure rate). If any indicator is abnormal, the rollback mechanism will be triggered immediately to switch back to the original algorithm, and the cause of the abnormality will be investigated (such as data drift or unreasonable parameter settings). Optimization of flow control strategy: Initialize relevant data: Collect traffic initialization allocation records for the past 7 days (including the initial traffic allocation ratio and allocation time of each service provider), as well as real-time traffic adjustment records within the target time window after initialization; the size of the target time window is determined according to the service fluctuation frequency of the service provider, and the selection criterion is "the higher the service fluctuation frequency of the service provider, the smaller the time window". In specific implementation cases, the time window can be selected as 30 minutes, 60 minutes or 120 minutes. Post-adjustment verification data: Collect historical traffic adjustment records within the past 7 days (including adjustment time, adjustment range, and adjustment targets), as well as SMS sending results within the post-adjustment verification time window (including delivery rate, sending delay, and failure reasons); the size of the verification time window is determined based on the stable duration of the traffic adjustment effect, and in specific implementation cases, 5 minutes, 10 minutes, or 30 minutes can be selected. In some implementation cases, the criteria for determining whether the initialization strategy needs to be adjusted are as follows: if the number of real-time traffic adjustments is ≥3 times within the target time window after initialization (quantitative standard), then the initialization allocation result is deemed inapplicable and the initialization strategy needs to be adjusted; if the number of adjustments is <3 times, then the original initialization strategy is maintained. The initialization strategy adjustment and verification are performed as follows: Adjustment Operation: Based on the initial allocation record and the real-time adjustment reasons (such as multiple adjustments due to excessive latency of a certain service provider), adjust the initial traffic allocation ratio of each service provider. The adjustment range is ±10%-20% of the original ratio (quantitative parameter). Prioritize reducing the initial traffic allocation ratio of service providers with unstable services (such as service providers with excessive latency, the initial traffic ratio is reduced by 15%). Verification process: Deploy the adjusted initialization strategy to the test environment and monitor the number of traffic adjustments within a target time window. If the number of adjustments is less than 3, and the delivery rate increases by ≥5% and the sending latency decreases by ≥100ms (quantitative verification standard), then the initialization strategy adjustment is effective. If the standard is not met, repeat the adjustment operation until the verification requirements are met. In some implementation cases, the criteria for determining whether to adjust the real-time adjustment strategy are as follows: if the SMS delivery rate increases by less than 5% and the sending latency decreases by less than 100ms (quantitative standard) within the verification time window after traffic adjustment, then the real-time adjustment strategy is deemed to be insufficiently adaptable and needs to be adjusted; if the above criteria are met, the original real-time adjustment strategy is maintained. Real-time strategy adjustment and verification are performed as follows: Adjustment procedures: For scenarios with excessive latency, the traffic adjustment range is adjusted. In some implementation cases, the adjustment rules are as follows: for latency exceeding the limit by ≤50ms, the traffic adjustment range is 10%-15%; for latency exceeding the limit by >50ms and ≤100ms, the traffic adjustment range is 15%-20%; for latency exceeding the limit by >100ms, the traffic adjustment range is 20%-25%. Verification process: Deploy the adjusted real-time adjustment strategy to the test environment and monitor the SMS sending results within three verification time windows. If the delivery rate increases by ≥5% and the sending latency decreases by ≥100ms, the real-time adjustment strategy takes effect. If the standard is not met, re-analyze the historical adjustment records and failure reasons, and repeat the adjustment operation until the verification requirements are met.

[0039] In summary, this invention, through the technical means of "multi-dimensional data source fusion + multi-index routing evaluation model + latency-constrained dynamic traffic control + closed-loop optimization," solves the technical problems of existing technologies such as black-box routing selection, single data source, lack of dynamic traffic control, and inability to adapt to differences in service provider management. It achieves improved delivery success rate and controllable sending latency for marketing SMS messages. Specific technical advantages are as follows: Efficiency optimization: By automating routing and dynamic traffic control, manual experience-based decision-making is replaced, reducing the cost of manual intervention. The routing response time is ≤1 second, and the traffic adjustment delay is ≤30 seconds, significantly improving SMS dispatch efficiency. At the same time, precise routing matching and traffic allocation increase the delivery rate of marketing SMS by more than 15%, reducing the waste of resources caused by duplicate sending.

[0040] Cost savings; by selecting the best route, priority is given to service providers with high delivery rates and low costs, while avoiding the waste of marketing costs due to delivery failures and excessive latency; dynamic traffic control avoids excessive billing caused by excessive traffic concentration, and is expected to reduce SMS sending costs by 8%-12%.

[0041] Risks are controllable; with the maximum sending latency as the core constraint, real-time traffic adjustment avoids routing congestion and latency exceeding the limit, ensuring that marketing SMS messages reach users on time; at the same time, by matching service provider management rules, delivery failures caused by list control and sensitive word control are reduced, thereby reducing the risk of marketing results not meeting the target; the closed-loop optimization mechanism enables the strategy to continuously adapt to fluctuations in service provider service levels, improving system stability.

[0042] Scalability; adopts a modular design, supports the addition of new SMS service providers, and allows flexible adjustment of parameters such as routing evaluation indicator weights, maximum sending latency, and traffic adjustment thresholds to adapt to the marketing SMS sending needs of different industries and scales; the data source acquisition module supports the addition of more supplementary data sources (such as user feedback data) to further improve the accuracy of routing selection and traffic control.

[0043] Additionally, embodiments of the present invention provide a system for routing and traffic control of marketing text messages, comprising: The data acquisition module is used to collect historical transmission records and supplementary data sources, and to preprocess the collected data to form a unified dataset. The route evaluation model building module is used to build a route evaluation model based on preprocessed multi-dimensional data. The optimal route selection module is used to select the optimal route based on the route evaluation model; The dynamic flow control module is used to output the optimal routing selection result and adjust the flow allocation based on the real-time collected data on the transmission latency, delivery rate and load rate of each route. The closed-loop optimization module is used to add the latest SMS sending data, service provider management data, and SMS feature data to the training data, update the routing evaluation model, and ensure that the optimal routing evaluation model adapts to the latest changes in service provider service levels, thus linking with the dynamic traffic control strategy optimization.

[0044] It is understood that the system for routing and traffic control of marketing SMS provided by the present invention corresponds to the method for routing and traffic control of marketing SMS provided in the foregoing embodiments. The relevant technical features of the system for routing and traffic control of marketing SMS can be referred to the relevant technical features of the method for routing and traffic control of marketing SMS, and will not be repeated here.

[0045] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0046] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0047] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for routing and traffic control of marketing SMS messages, characterized in that, Includes the following steps: Collect historical transmission records and supplementary data sources, preprocess and feature-engineer the collected data to form a unified dataset; Based on the preprocessed unified dataset, a route evaluation model is constructed, and the optimal route is selected based on the route evaluation model; Establish a dynamic flow control mechanism based on latency constraints, output the optimal routing selection result, and adjust the flow allocation according to the real-time collected data on the sending latency, delivery rate and load rate of each route; Establish a closed-loop optimization mechanism by adding the latest SMS sending data, service provider management data, and SMS characteristic data to the training data, updating the routing evaluation model, ensuring that the optimal routing evaluation model adapts to the latest changes in service provider service levels, and forming a linkage with dynamic traffic control strategy optimization.

2. The method for routing and traffic control of marketing SMS messages according to claim 1, characterized in that, The collection of historical sending record data includes: collecting marketing SMS sending data within the past 3 months, including: SMS content, sending time, target user number, route, delivery result, sending delay, failure reason and traffic allocation ratio; The supplementary data sources collected include: the list control rules, sensitive word control rules, service commitments and real-time service status of each SMS service provider, as well as the content type and target user characteristics of marketing SMS messages.

3. The method for routing and traffic control of marketing SMS messages according to claim 1, characterized in that, The preprocessing of the collected data includes: removing invalid data, segmenting text data into words, constructing a bag-of-words model for documents, and normalizing numerical data; the feature engineering includes: establishing SMS features, target user features, association features, and indicator features.

4. The method for routing and traffic control of marketing SMS messages according to claim 1, characterized in that, The construction of the route evaluation model based on the preprocessed unified dataset includes: Using historical transmission records and service provider management data as training samples, the optimal route label is labeled, and a random forest-based route evaluation model is trained. The model performance is verified through a test set to ensure that the route selection accuracy is greater than the threshold. Construct a routing evaluation index system, identify core evaluation indicators including delivery performance indicators, management and control adaptation indicators, service stability indicators, and cost indicators, and dynamically adjust the weights according to business needs.

5. The method for routing and traffic control of marketing SMS messages according to claim 1, characterized in that, The optimal route selection based on the route evaluation model includes: For marketing SMS messages to be sent, input SMS characteristics, target user characteristics, and relationship characteristics data. Calculate the evaluation score of each available route through the routing evaluation model, and output the optimal route selection result according to the score.

6. The method for routing and traffic control of marketing SMS messages according to claim 1, characterized in that, The dynamic flow control mechanism includes: Based on the service provider's service commitments, real-time load rate, and historical latency data for each route, calculate the maximum traffic threshold that each route can carry. Initial traffic is allocated proportionally based on the evaluation scores of each route; routes with higher scores receive a higher proportion of initial traffic. Real-time collection of transmission latency, delivery rate, and load rate data for each route enables dynamic adjustment of route traffic; Ensure that the total traffic allocation for all routes equals the total traffic of the marketing SMS messages to be sent, and that the traffic allocation ratio for a single route is not lower than the minimum allocation ratio.

7. The method for routing and traffic control of marketing SMS messages according to claim 6, characterized in that, The real-time collection of transmission latency, delivery rate, and load rate data for each route, and the dynamic adjustment of route traffic, includes: If the real-time transmission latency of a certain route exceeds the preset maximum latency or the load rate exceeds the threshold configured at the time of access, the traffic allocation ratio of that route will be automatically reduced, and the reduced traffic will be allocated to routes with higher evaluation scores, normal latency, and lower load. If the delivery rate of a route drops below the minimum allocation ratio for several consecutive periods, traffic allocation for that route will be suspended until its service status is restored, at which point traffic will be gradually restored.

8. The method for routing and traffic control of marketing SMS messages according to claim 1, characterized in that, The establishment of the closed-loop optimization mechanism includes: Collect the actual results of each transmission and the routing service status data, compare the prediction results of the routing evaluation model with the actual results, and calculate the deviation. Based on the deviation data, the indicator weights and model parameters of the route evaluation model are dynamically adjusted to optimize the route selection logic; Based on historical traffic adjustment records and sending results, optimize the initial traffic allocation ratio and adjust thresholds to adapt to dynamic changes in the service provider's service level. We update service provider management rules and environmental data daily, and update historical sending records weekly to support model and strategy optimization.

9. A method for routing and traffic control of marketing SMS messages according to claim 8, characterized in that, The latest SMS sending data includes: routing characteristics and sending results. The sending results include: delivery rate, latency, and reasons for failure. The actual sending results and routing service status data are collected, and the prediction results of the optimal routing evaluation model are compared with the actual results to provide sample data support for continuous iteration of routing selection and optimization of model parameters.

10. A system for routing and flow control of marketing text messages, characterized in that, include: The data acquisition module is used to collect historical transmission records and supplementary data sources, and to preprocess the collected data to form a unified dataset. The route evaluation model building module is used to build a route evaluation model based on a preprocessed, unified dataset. The optimal route selection module is used to select the optimal route based on the route evaluation model; The dynamic flow control module is used to output the optimal routing selection result and adjust the flow allocation based on the real-time collected data on the transmission latency, delivery rate and load rate of each route. The closed-loop optimization module is used to add the latest SMS sending data, service provider management data, and SMS feature data to the training data, update the routing evaluation model, and ensure that the optimal routing evaluation model adapts to the latest changes in service provider service levels, thus forming a linkage with the dynamic traffic control strategy optimization.

Images