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

Abstract

The invention discloses a position-based multi-operator short message access priority selection system and method, and relates to the technical field of computers, and the method comprises the steps: collecting an operator information set and a user demand information set in advance, building a prediction model based on the operator information set and the user demand information set, and using the prediction model. Calculating a comprehensive score of each operator, selecting an optimal operator based on the comprehensive score, collecting feedback information in real time, and dynamically adjusting a selection strategy based on the real-time feedback information; according to the invention, the success rate and reliability of short message sending can be obviously improved, and the operation cost is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a location-based multi-operator short message access priority selection system and method. Background Art

[0002] The platform needs to send a large number of text messages to users around the world, such as account verification, transaction notifications, and promotional information. The sending of these text messages is critical in terms of time and reliability, and choosing the right operator is a key factor in ensuring efficient information delivery. Traditional SMS sending usually relies on pre-set operators and lacks the ability to respond to real-time network conditions. This method has several significant shortcomings: first, it cannot cope with changes in signal strength and network congestion in different regions and time periods; second, cost control is poor, and operating costs increase due to the lack of application of dynamic pricing strategies; finally, service quality is uncontrollable, and user experience is easily affected by delays or transmission failures.

[0003] The Chinese patent with the authorization announcement number CN114727241B proposes a method to improve the efficiency of SMS group sending, including the following steps: S1, filter out the operator to which the number belongs by number segment rules; S2, filter out the operator supported by each channel; S3, obtain the detailed parameters of each channel and score each channel according to the scoring criteria; S4, set the priority sending channel according to the channel score value from high to low. However, this method fails to take into account the balance between communication quality, cost and sending success rate.

[0004] To this end, the present invention proposes a location-based multi-operator SMS access priority selection system and method. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a location-based multi-operator SMS access priority selection system and method, which can significantly improve the success rate and reliability of SMS sending, while effectively reducing operating costs.

[0006] To achieve the above purpose, a location-based multi-operator SMS access priority selection method is proposed, comprising the following steps:

[0007] Step 1: Collect operator information set and user demand information set in advance;

[0008] Step 2: Build a prediction model based on the operator information set and the user demand information set;

[0009] Step 3: Use the prediction model to calculate the comprehensive score of each operator;

[0010] Step 4: Select the best operator based on the comprehensive score and collect feedback information in real time;

[0011] Step 5: Dynamically adjust the selection strategy based on real-time feedback information.

[0012] The operator information set is collected in the following manner:

[0013] Through the operator platform, the signal coverage data, cost data and network congestion data generated in the process of sending text messages to users through the operator platform in the past are collected to form an operator information set.

[0014] The user demand information set is collected in the following manner:

[0015] The user location data, historical sending data and user device information related to the user's terminal device when sending text messages to users through the operator platform in the past are collected to form a user demand information set.

[0016] The method of constructing the prediction model based on the operator information set and the user demand information set is:

[0017] Step 21: quantify the parameters in the operator information set and the user demand information set and unify the data format;

[0018] Step 22: Based on the operator information set and the user demand information set in a unified data format, respectively train a success rate prediction model for predicting a transmission success rate and a delay prediction model for predicting a transmission delay;

[0019] Step 23: Construct a cost calculation formula as a cost prediction model based on the output of the success rate prediction model, the number of SMS messages sent each time, and the cost data.

[0020] The training of the success rate prediction model for training the prediction of the sending success rate comprises the following steps:

[0021] Step 2211: Selecting parameters related to the sending success rate from the operator information set and the user demand information set as success rate characteristic parameters, and using the success rate characteristic parameters each time a text message is sent through an operator as the first input sample of the success rate prediction model;

[0022] Step 2212: Select the random forest regression model as the success rate prediction model, and set the basic parameters of the random forest regression model;

[0023] Step 2213: Divide all first input samples into a training set, a validation set, and a test set. Use k-fold cross validation to evaluate the performance of each parameter combination in the validation set. Cross validation ensures the generalization ability of the model by evaluating the model multiple times on different data subsets. According to the evaluation results of cross validation, select the parameter combination with the best performance on the validation set to train the final success rate prediction model.

[0024] Step 2214: the success rate prediction model uses the success rate prediction value as output, the success rate prediction value is between [0,1], the mean square error is selected as the performance indicator, the success rate prediction model is trained with the best parameter combination, the random forest model is reinitialized using the best hyperparameter configuration selected by grid search, and the success rate prediction model is trained on the training set;

[0025] Step 2215: The success rate prediction model outputs each performance indicator to quantify the prediction ability of the model, and performs iterative training until the performance indicator converges.

[0026] The training of the delay prediction model for predicting the transmission delay comprises the following steps:

[0027] Step 2221: Select a parameter related to the sending success rate from the operator information set and the user demand information set as a delay characteristic parameter, and use the time series of the delay characteristic parameter when sending a text message through the operator each time as a second input sample of the delay prediction model;

[0028] Step 2222: Select the LSTM model as the delay prediction model, and construct an input-output pair (X, y) for the time window corresponding to each time series in the second input sample, where X is the model input, representing the time series of each delay feature parameter in the time window, and y is the prediction target, representing the delay at the next moment;

[0029] Step 2223: Use the model input X in each input-output pair as the input of the delay prediction model, use the prediction target y as the prediction target of the delay prediction model, use the predicted value of the delay as the output of the delay prediction model, and use the mean square error between the prediction target and the output as the error function to train the delay prediction model.

[0030] The method of constructing the cost calculation formula through the output of the success rate prediction model, the number of SMS messages sent each time and the cost data is as follows:

[0031] Each time the operator sends a text message, a first input sample is collected and input into the success rate prediction model to obtain a predicted success rate output by the success rate prediction model;

[0032] The cost calculation formula is calculated as follows:

[0033] C = p × z × r × n + w × (1-r);

[0034] Among them, C represents the cost of sending SMS, p is the unit price of sending SMS, z is the discount negotiated with the operator platform, r is the predicted success rate, n is the number of SMS sent, and w is the penalty coefficient for failed sending, which is used to balance the relationship between the sending success rate and the sending cost.

[0035] The method of using the prediction model to calculate the comprehensive score of each operator is as follows:

[0036] Collecting predicted first input samples and predicted second input samples of each operator platform within a preset sending cycle;

[0037] The predicted first input sample and the predicted second input sample are respectively input into the success rate prediction model and the delay prediction model to obtain the predicted success rate of SMS sending and the predicted delay of SMS sending, and the predicted cost is calculated by the cost calculation formula;

[0038] By calculating the weighted sum of the sending success rate, sending delay and cost, the comprehensive score of each operator platform is obtained, and the operator platform with the highest comprehensive score is selected as the best operator.

[0039] The method of collecting feedback information in real time is:

[0040] In a sending cycle, each time a text message is sent through a determined operator platform, corresponding signal coverage data, cost data, and network congestion data are collected as feedback information.

[0041] The method of dynamically adjusting the selection strategy based on real-time feedback information is as follows:

[0042] Pre-set warning limit ranges for parameters in signal coverage data, cost data, and network congestion data. When the parameters do not fall within the corresponding warning limit ranges, select operator adjustments;

[0043] The first input sample and the second input sample generated after each operator platform sends a text message each time during the current sending cycle are collected, and then the comprehensive score of each operator is recalculated through the first input sample, the second input sample and the prediction model, and the operator with the highest comprehensive score is selected as the target operator for adjustment.

[0044] A location-based multi-operator SMS access priority selection system is proposed, which includes a sample collection module, a model training module, an operator selection module, and a real-time feedback module; wherein each module is electrically connected to another;

[0045] A sample collection module collects operator information sets and user demand information sets in advance, and sends the operator information sets and user demand information sets to the model training module;

[0046] The model training module builds a prediction model based on the operator information set and the user demand information set, and sends the prediction model to the operator selection module;

[0047] The operator selection module uses the prediction model to calculate the comprehensive score of each operator, selects the best operator based on the comprehensive score, collects feedback information in real time, and sends the feedback information to the real-time feedback module;

[0048] The real-time feedback module dynamically adjusts the selection strategy based on real-time feedback information.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention collects operator information sets and user demand information sets in advance, builds a prediction model based on the operator information sets and the user demand information sets, uses the prediction model to calculate the comprehensive scores of each operator, selects the best operator based on the comprehensive score, collects feedback information in real time, and dynamically adjusts the selection strategy based on the real-time feedback information. By real-time monitoring of signal coverage data, network congestion data and cost data, the Internet platform can make more targeted operator selections when sending text messages, which not only improves the success rate of sending, but also optimizes costs while ensuring service quality. The intelligent selection strategy needs to be combined with a machine learning model to predict the service performance of the operator, and the dynamic control mechanism ensures that when unfavorable conditions are detected, the platform can quickly adjust the strategy and select a better operator, which can significantly improve the success rate and reliability of text message sending, while effectively reducing operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flowchart of a location-based multi-operator SMS access priority selection method in Embodiment 1 of the present invention;

[0052] Figure 2 This is a module connection relationship diagram of the location-based multi-operator SMS access priority selection system in Example 2 of the present invention. DETAILED DESCRIPTION

[0053] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] like Figure 1 As shown, the location-based multi-operator SMS access priority selection method includes the following steps:

[0056] Step 1: Collect operator information set and user demand information set in advance;

[0057] Step 2: Build a prediction model based on the operator information set and the user demand information set;

[0058] Step 3: Use the prediction model to calculate the comprehensive score of each operator;

[0059] Step 4: Select the best operator based on the comprehensive score and collect feedback information in real time;

[0060] Step 5: Dynamically adjust the selection strategy based on real-time feedback information.

[0061] The operator information set is collected in the following manner:

[0062] Through long-term cooperation with operator platforms, we collect signal coverage data, cost data, and network congestion data generated in the process of sending SMS messages to users through operator platforms in the past, and form an operator information collection;

[0063] Specifically, the signal coverage data includes signal strength, network type and signal stability:

[0064] The signal strength is the signal strength in dBm at different geographical locations; the network type is the coverage of communication networks such as 2G, 3G, 4G, and 5G; and the signal stability includes the drop rate and fluctuation frequency;

[0065] Preferably, historical signal data records may be used in combination with signal strength information uploaded by user devices to form a historical signal strength database, and a crowdsourcing platform may be used to collect signal data of user devices.

[0066] Further, the cost data includes a basic fee for each SMS;

[0067] The basic fee can be obtained by downloading past pricing plans and fee structures from the operator's official website, or by reading historical data recording contracts and volume discounts to learn specific discount information.

[0068] Furthermore, the network congestion data includes real-time delay, throughput and historical congestion trends:

[0069] Specifically, the real-time delay is network delay data in different time periods, the throughput represents the load level of the network at a certain moment in the past, and the historical congestion trend represents the congestion situation of the communication network during historical peak periods and historical low peak periods;

[0070] The network congestion data can be obtained by using a network monitoring tool to record historical network latency and throughput, or by directly extracting data from historical congestion reports and traffic analysis APIs provided by operators.

[0071] Furthermore, the user demand information set is collected in the following manner:

[0072] Collect user location data, historical sending data, and user device information related to user terminal devices when sending text messages to users through the operator platform in the past, and form a user demand information collection;

[0073] Specifically, the user location data includes geographic coordinates and location type;

[0074] The geographic coordinates are the latitude and longitude information of the user's location, and the location type is the classification of the pre-divided location area, including urban, suburban, rural and other classifications;

[0075] The user location data may be set to the location corresponding to the user's mobile phone number, or may be set to record the base station location where each mobile phone number was located when it last accessed the network.

[0076] Furthermore, the historical sending data includes the success rate, sending delay, and sending volume of each SMS sending;

[0077] The success rate refers to the proportion of text messages successfully received by users after text messages are sent to a batch of users through the operator platform each time. The sending delay is the average sending and receiving time delay of each text message. The sending volume is the total amount of text messages sent each time. The historical sending data can be read from the background data of the operator platform.

[0078] Specifically, the user device information includes device type, operating system, network connection type, etc.; the user device information may be provided in advance by the user.

[0079] Furthermore, the method of constructing the prediction model based on the operator information set and the user demand information set is:

[0080] Step 21: quantify the parameters in the operator information set and the user demand information set and unify the data format;

[0081] Specifically, the quantization process refers to converting a non-digital parameter into a digital form; specifically, the quantization process includes:

[0082] Signal strength: quantify signal strength data in different areas as a scalar such as average dBm value or a time series of signal strength changes at a specific location;

[0083] Signal stability: Calculate the drop rate or fluctuation frequency as a scalar;

[0084] Sending cost: a scalar quantity quantified as the average cost per SMS message;

[0085] Discount information: quantify the historical cost change trend as a time series;

[0086] Real-time delay: Use scalars to represent the current average delay or time series, and analyze the delay changes in different time periods;

[0087] Throughput: recorded as time series data, reflecting the network load at different times;

[0088] Geographic coordinates: Each location is encoded as a scalar or categorical quantity, e.g., by k-means clustering.

[0089] Success rate, latency, and send volume can be directly used as scalars, or composed into time series to capture trends and seasonal changes;

[0090] Device type and operating system can be quantified by encoding the different types as categorical variables.

[0091] The unified data format refers to converting all scalar data into the same format. For example, for prediction models whose input is a single scalar, such as CNN models, DNN models, etc., the scalar form of each input parameter can be read, and for prediction models whose input is a time series, the time series form of each input parameter can be read, that is, the form of a time series composed of temporally continuous input parameters can be read.

[0092] Step 22: Based on the operator information set and the user demand information set in a unified data format, respectively train a success rate prediction model for predicting a transmission success rate and a delay prediction model for predicting a transmission delay;

[0093] Step 23: Construct a cost calculation formula as a cost prediction model through the output of the success rate prediction model, the number of SMS messages sent each time, and the cost data;

[0094] Specifically, the training of the success rate prediction model for training the transmission success rate includes the following steps:

[0095] Step 2211: Selecting parameters related to the sending success rate from the operator information set and the user demand information set as success rate characteristic parameters, and using the success rate characteristic parameters each time a text message is sent through an operator as the first input sample of the success rate prediction model;

[0096] Specifically, the success rate characteristic parameters include signal strength (using average signal strength), network congestion, device type distribution, and geographic location distribution; wherein the device type distribution refers to the proportion of different device types used by target users who send text messages through the operator each time, such as the proportion of tablets, the proportion of mobile phones, etc., and the geographic location distribution refers to the distribution proportion of target users in various location areas, such as the proportion of first-tier cities, the proportion of second-tier cities, the proportion of rural areas, etc., so as to measure the availability of communication base stations.

[0097] Step 2212: Select the random forest regression model as the success rate prediction model, and set the basic parameters of the random forest regression model;

[0098] Preferably, the method of setting the basic parameters of the random forest model is:

[0099] Set the following parameters for the random forest regression model:

[0100] The number of trees is usually set to an integer in [100,1000], which is determined according to actual needs;

[0101] The maximum depth is set to an integer in [5,10] to control the depth of each tree to avoid overfitting;

[0102] The minimum number of sample splits is set to an integer in [2,10] to avoid overfitting, set to 2 or higher.

[0103] The minimum number of sample leaf nodes is set to an integer in [1,5);

[0104] Step 2213: Divide all first input samples into a training set, a validation set, and a test set. Use k-fold cross validation in the validation set to evaluate the performance of each parameter combination. Cross validation ensures the generalization ability of the model by evaluating the model multiple times on different data subsets. According to the evaluation results of cross validation, select the parameter combination with the best performance on the validation set to train the final success rate prediction model. k is a preset quantity parameter, and the value is generally 5.

[0105] Step 2214: the success rate prediction model uses the success rate prediction value as output, the success rate prediction value is a decimal between [0, 1], the mean square error is selected as the performance indicator, the success rate prediction model is trained with the best parameter combination, the random forest regression model is reinitialized using the best hyperparameter configuration selected by grid search, and the success rate prediction model is trained on the training set;

[0106] Step 2215: The success rate prediction model outputs each performance indicator to quantify the prediction ability of the model, and performs iterative training until the performance indicator converges.

[0107] Furthermore, the training of the delay prediction model for predicting the sending delay includes the following steps:

[0108] Step 2221: Select a parameter related to the sending success rate from the operator information set and the user demand information set as a delay characteristic parameter, and use the time series of the delay characteristic parameter when sending a text message through the operator each time as a second input sample of the delay prediction model;

[0109] Specifically, the delay characteristic parameters include network delay time series, signal strength change time series, and historical network congestion level time series;

[0110] Preferably, device type, geographical location distribution and sending time characteristics may be further selected as auxiliary characteristics;

[0111] Step 2222: Select the LSTM model as the delay prediction model, and construct an input-output pair (X, y) for the time window corresponding to each time series in the second input sample, where X is the model input, representing the time series of each delay feature parameter in the time window, and y is the prediction target, representing the delay at the next moment;

[0112] Step 2223: Use the model input X in each input-output pair as the input of the delay prediction model, use the prediction target y as the prediction target of the delay prediction model, use the predicted value of the delay as the output of the delay prediction model, and use the mean square error between the prediction target and the output as the error function to train the delay prediction model.

[0113] Furthermore, the cost calculation formula is constructed by using the output of the success rate prediction model, the amount of SMS sent each time, and the cost data as follows:

[0114] Each time the operator sends a text message, a first input sample is collected and input into the success rate prediction model to obtain a predicted success rate output by the success rate prediction model;

[0115] The cost calculation formula is calculated as follows:

[0116] C = p × z × r × n + w × (1-r);

[0117] Among them, C represents the cost of sending SMS, p is the unit price of sending SMS, z is the discount negotiated with the operator platform, r is the predicted success rate, n is the number of SMS sent, and w is the penalty coefficient for failed sending, which is used to balance the relationship between the sending success rate and the sending cost.

[0118] Furthermore, the method of using the prediction model to calculate the comprehensive score of each operator is as follows:

[0119] It is understandable that when sending SMS to users through the operator platform, it is extremely wasteful to adjust the operator every time a SMS is sent. Therefore, it can be set to use a fixed operator within a sending cycle.

[0120] In a preset sending cycle, the predicted first input sample and the predicted second input sample of each operator platform are collected; the sending cycle may be one week; specifically, the predicted first input sample and the predicted second input sample may be obtained by averaging the first input samples and the second input samples in several past sending cycles, for example, the average signal strength of the number of short messages sent by each operator in the past week is taken as the average signal strength in the first input sample;

[0121] The predicted first input sample and the predicted second input sample are respectively input into the success rate prediction model and the delay prediction model to obtain the predicted success rate of SMS sending and the predicted delay of SMS sending, and the predicted cost is calculated by the cost calculation formula;

[0122] By calculating the weighted sum of the sending success rate, sending delay and cost, the comprehensive score of each operator platform is obtained, and the operator platform with the highest comprehensive score is selected as the best operator.

[0123] Furthermore, the method of collecting feedback information in real time is:

[0124] In a sending cycle, each time a text message is sent through a determined operator platform, corresponding signal coverage data, cost data, and network congestion data are collected as feedback information.

[0125] Furthermore, the method of dynamically adjusting the selection strategy based on real-time feedback information is as follows:

[0126] Pre-set warning limit ranges for several parameters in signal coverage data, cost data, and network congestion data. When the parameters do not fall within the corresponding warning limit ranges, select operator adjustments;

[0127] For example, a signal strength threshold is set for signal strength: when the signal strength is lower than the set signal strength threshold, such as -90dBm, consider adjusting the operator;

[0128] The first input sample and the second input sample generated each time each operator platform sends a text message during the current sending cycle are collected, and then the comprehensive score of each operator is recalculated through the first input sample, the second input sample and the prediction model, and the operator with the highest comprehensive score is selected as the target operator for adjustment.

[0129] Example 2

[0130] like Figure 2 As shown, the location-based multi-operator SMS access priority selection system includes a sample collection module, a model training module, an operator selection module and a real-time feedback module; wherein each module is electrically connected;

[0131] A sample collection module collects operator information sets and user demand information sets in advance, and sends the operator information sets and user demand information sets to the model training module;

[0132] The model training module builds a prediction model based on the operator information set and the user demand information set, and sends the prediction model to the operator selection module;

[0133] The operator selection module uses the prediction model to calculate the comprehensive score of each operator, selects the best operator based on the comprehensive score, collects feedback information in real time, and sends the feedback information to the real-time feedback module;

[0134] The real-time feedback module dynamically adjusts the selection strategy based on real-time feedback information.

[0135] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0136] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.

[0137] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A location-based multi-operator SMS access priority selection method, characterized in that: The following steps are involved: Step 1: Collect operator information and user demand information in advance; Step 2: Build a prediction model based on the operator information set and the user demand information set; Step 3: Use the prediction model to calculate the comprehensive score of each operator; Step 4: Select the best operator based on the comprehensive score and collect feedback information in real time; Step 5: Dynamically adjust the selection strategy based on real-time feedback information.

2. The location-based multi-operator SMS access priority selection method according to claim 1, characterized in that: The collection method of operator information collection is: Through the operator platform, the signal coverage data, cost data and network congestion data generated in the process of sending text messages to users through the operator platform in the past are collected to form an operator information set.

3. The location-based multi-operator SMS access priority selection method according to claim 2, characterized in that: The user demand information set is collected in the following manner: The user location data, historical sending data and user device information related to the user's terminal device when sending text messages to users through the operator platform in the past are collected to form a user demand information set.

4. The location-based multi-operator SMS access priority selection method according to claim 3, characterized in that: The method of constructing the prediction model based on the operator information set and the user demand information set is: Step 21: quantify the parameters in the operator information set and the user demand information set and unify the data format; Step 22: Based on the operator information set and the user demand information set in a unified data format, respectively train a success rate prediction model for predicting a transmission success rate and a delay prediction model for predicting a transmission delay; Step 23: Construct a cost calculation formula as a cost prediction model based on the output of the success rate prediction model, the number of SMS messages sent each time, and the cost data.

5. The location-based multi-operator SMS access priority selection method according to claim 4, characterized in that: The training of the success rate prediction model for training the prediction of the sending success rate comprises the following steps: Step 2211: Selecting parameters related to the sending success rate from the operator information set and the user demand information set as success rate characteristic parameters, and using the success rate characteristic parameters each time a text message is sent through an operator as the first input sample of the success rate prediction model; Step 2212: Select the random forest regression model as the success rate prediction model, and set the basic parameters of the random forest regression model; Step 2213: Divide all first input samples into a training set, a validation set, and a test set, and use k-fold cross validation to evaluate the performance of each parameter combination in the validation set; based on the evaluation results of the cross validation, select the parameter combination with the best performance in the validation set to train the final success rate prediction model; Step 2214: the success rate prediction model uses the success rate prediction value as output, the success rate prediction value is between [0,1], the mean square error is selected as the performance indicator, the success rate prediction model is trained with the best parameter combination, the random forest regression model is reinitialized using the best hyperparameter configuration selected by grid search, and the success rate prediction model is trained on the training set; Step 2215: The success rate prediction model outputs the performance index each time to quantify the prediction ability of the model, and performs iterative training until the performance index converges.

6. The location-based multi-operator SMS access priority selection method according to claim 5, characterized in that: The training of the delay prediction model for predicting the transmission delay comprises the following steps: Step 2221: Select a parameter related to the sending success rate from the operator information set and the user demand information set as a delay characteristic parameter, and use the time series of the delay characteristic parameter when sending a text message through the operator each time as a second input sample of the delay prediction model; Step 2222: Select the LSTM model as the delay prediction model, and construct an input-output pair (X, y) for the time window corresponding to each time series in the second input sample, where X is the model input, representing the time series of each delay feature parameter in the time window, and y is the prediction target, representing the delay at the next moment; Step 2223: Use the model input X in each input-output pair as the input of the delay prediction model, use the prediction target y as the prediction target of the delay prediction model, use the predicted value of the delay as the output of the delay prediction model, and use the mean square error between the prediction target and the output as the error function to train the delay prediction model.

7. The location-based multi-operator SMS access priority selection method according to claim 6, characterized in that: The method of using the prediction model to calculate the comprehensive score of each operator is as follows: Collecting predicted first input samples and predicted second input samples of each operator platform within a preset sending cycle; The predicted first input sample and the predicted second input sample are respectively input into the success rate prediction model and the delay prediction model to obtain the predicted success rate of SMS sending and the predicted delay of SMS sending, and the predicted cost is calculated by the cost calculation formula; By calculating the weighted sum of the sending success rate, sending delay and cost, the comprehensive score of each operator platform is obtained, and the operator platform with the highest comprehensive score is selected as the best operator.

8. The location-based multi-operator SMS access priority selection method according to claim 7, characterized in that: The method of collecting feedback information in real time is: In a sending cycle, each time a text message is sent through a determined operator platform, corresponding signal coverage data, cost data, and network congestion data are collected as feedback information.

9. The location-based multi-operator SMS access priority selection method according to claim 8, characterized in that: The method of dynamically adjusting the selection strategy based on real-time feedback information is as follows: Pre-set warning limit ranges for parameters in signal coverage data, cost data, and network congestion data. When the parameters do not fall within the corresponding warning limit ranges, select operator adjustments; The first input sample and the second input sample generated after each operator platform sends a text message each time during the current sending cycle are collected, and then the comprehensive score of each operator is recalculated through the first input sample, the second input sample and the prediction model, and the operator with the highest comprehensive score is selected as the target operator for adjustment.

10. A location-based multi-operator SMS access priority selection system, which is used to implement the location-based multi-operator SMS access priority selection method according to any one of claims 1 to 9, characterized in that: It includes a sample collection module, a model training module, an operator selection module and a real-time feedback module; wherein each module is electrically connected to another; A sample collection module collects operator information sets and user demand information sets in advance, and sends the operator information sets and user demand information sets to the model training module; The model training module builds a prediction model based on the operator information set and the user demand information set, and sends the prediction model to the operator selection module; The operator selection module uses the prediction model to calculate the comprehensive score of each operator, selects the best operator based on the comprehensive score, collects feedback information in real time, and sends the feedback information to the real-time feedback module; The real-time feedback module dynamically adjusts the selection strategy based on real-time feedback information.

Images