Method and system for sending message matching channel

Abstract

The embodiment of the present invention discloses a method and system for sending a message through a matching channel, wherein the method comprises: obtaining business data; analyzing the business data to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between message factors and channel factors; constructing a neural network model, injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between message factors and channel factors into the neural network model as constraint conditions, training the neural network model to obtain an optimal neural network model; inputting a message to be sent into the optimal neural network model to obtain a matching channel for the message. The method solves the problem in the prior art that it is impossible to match an accurate sending channel for a message.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, system, electronic device and storage medium for sending a message matching channel. Background Art

[0002] In the field of communications, current information transmission is becoming more and more diversified, from the earlier traditional short message transmission to multimedia messages, 5G messages, and international short messages. Information attributes are constantly expanding, and the attributes of message sending channels are also expanding. The current information sending strategy is to configure the corresponding channel for sending information. This work can be manually configured or dynamically configured by the program, but corresponding to the huge amount of information transmission, it involves switching channels and enterprise information configuration channels, which is quite cumbersome and prone to errors; taking the program to configure the channel for the message as an example, considering that the information consists of n attributes and the channel consists of m attributes (such as the channel protocol CMPP, SMGP, SGIP and SMPP four attributes, and the channel cost high, medium and low three attributes), find the channel based on one attribute of the information. Whether each attribute of the channel is available (decision to use) is a Boolean function set {available, unavailable}, and then it appears The situation is an exponential data. Moreover, the information has n attributes, and some attributes of the channel are multiple enumerations. It is obviously inappropriate and unrealistic to use such a complex program to complete the entire automatic configuration and switching. It is necessary to introduce artificial intelligence to deal with such problems, so as to realize the automatic switching of message sending channels and channel expansion, and provide business operators with information sending reference data (such as: before sending information, give the type of channel to be used for sending, the estimated sending time, whether capacity expansion is needed, etc.).

[0003] Therefore, there is an urgent need for a message matching channel sending method that can automatically switch the message sending channel. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a method, system, electronic device and storage medium for matching a message with a sending channel, so as to solve the problem in the prior art that an accurate sending channel cannot be matched for a message.

[0005] To achieve the above object, an embodiment of the present invention provides a method for sending a message matching channel, the method specifically comprising:

[0006] Get business data;

[0007] Analyzing the business data to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor, including:

[0008] Determine the linear correlation between message factors and channel factors based on the changing rules of the strong correlation between message factors and channel factors;

[0009] Based on the changing rules of the strong correlation between message factors and channel factors, the nonlinear relationship between message factors and channel factors is determined;

[0010] Obtaining an influence relationship between a message factor and a message factor change rate on a matching result, and a differential equation based on the influence relationship;

[0011] Obtaining the influence relationship between multiple message factors and the change rates of multiple message factors on the matching results, and a partial differential equation based on the influence relationship;

[0012] Constructing a neural network model, injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as constraints or boundaries into the neural network of the neural network model, guiding and constraining the learning of the neural network, so as to obtain an optimal neural network model;

[0013] The message to be sent is input into the optimal neural network model to obtain a matching channel for the message.

[0014] On the basis of the above technical solution, the present invention can also be improved as follows:

[0015] Furthermore, the construction of the neural network model injects the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as constraints or boundaries into the neural network of the neural network model to guide and constrain the learning of the neural network to obtain the optimal neural network model, including:

[0016] Calculate expected numerical loss;

[0017] Calculate physical information loss;

[0018] Calculating a total loss based on the expected numerical loss and the physical information loss;

[0019] The optimal neural network model is obtained by minimizing the overall loss.

[0020] Further, the calculating the expected numerical loss includes:

[0021] Calculate the expected numerical loss using formula (1);

[0022] Formula (1);

[0023] In the formula, is the expected numerical loss, m is the number of constraint factors, is the vector norm squared, is the input parameter, θ is the weight and offset, is the function that approximates U through a neural network, It is the unknown function of channel factors on message factors. The unknown function includes linear relationship, nonlinear relationship, differential equation and partial differential equation.

[0024] Further, the calculating physical information loss includes:

[0025] The physical information loss is calculated by formula (2);

[0026] Formula (2);

[0027] In the formula, is the physical information loss, n is the number of constraint factors, F is the partial differential equation corresponding to the expected solution of U(x,θ), is the vector norm squared, An element of the message factor X element set.

[0028] A system for sending a message matching channel, comprising:

[0029] Acquisition module, used to obtain business data;

[0030] A constraint determination module, used to analyze the business data and determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor;

[0031] Determine the linear correlation between message factors and channel factors based on the changing rules of the strong correlation between message factors and channel factors;

[0032] Based on the changing rules of the strong correlation between message factors and channel factors, the nonlinear relationship between message factors and channel factors is determined;

[0033] Obtaining an influence relationship between a message factor and a message factor change rate on a matching result, and a differential equation based on the influence relationship;

[0034] Obtaining the influence relationship between multiple message factors and the change rates of multiple message factors on the matching results, and a partial differential equation based on the influence relationship;

[0035] Building blocks for building neural network models;

[0036] A training module, for injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as a constraint condition or boundary into the neural network of the neural network model, guiding and constraining the learning of the neural network to obtain an optimal neural network model;

[0037] The matching module is used to input the message to be sent into the optimal neural network model to obtain a matching channel for the message.

[0038] An electronic device comprises a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the method when executing the computer program.

[0039] A non-transitory computer-readable storage medium stores a computer program, which implements the steps of the method when executed by a processor.

[0040] The embodiments of the present invention have the following advantages:

[0041] The message matching channel sending method in the present invention uses an artificial neural network commonly used in artificial intelligence, and is based on the improvement of unsupervised learning of historical data. It can well solve the negative impact that noise data in historical data may have on learning results. Through supervised learning such as physical information neural network, it can not only improve the learning efficiency, but also improve the expected results of the output, solving the problem that the existing technology cannot accurately match the sending channel for the message. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0043] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0044] Figure 1 A flowchart of a method for sending a message matching channel according to the present invention;

[0045] Figure 2 The architecture diagram of the system for sending a message matching channel of the present invention;

[0046] Figure 3 This is a schematic diagram of the physical structure of the electronic device provided by the present invention.

[0047] The accompanying drawings are denoted as follows:

[0048] Acquisition module 10 , constraint determination module 20 , construction module 30 , training module 40 , matching module 50 , electronic device 60 , processor 601 , memory 602 , bus 603 . DETAILED DESCRIPTION

[0049] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not 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.

[0050] Figure 1 This is a flow chart of an embodiment of a method for sending a message matching channel according to the present invention, as shown in FIG. Figure 1 As shown, a method for sending a message matching channel provided by an embodiment of the present invention includes the following steps:

[0051] S101, obtaining business data;

[0052] Specifically: Business data includes message data and channel data;

[0053] Message data includes message type (SMS, QQ-MMS, 5G, international SMS), delivery strategy (immediate delivery, scheduled delivery), timeliness requirement, and arrival requirement;

[0054] Channel data includes channel protocol type (CMPP, SMGP, SGIP, SMPP), channel affiliation (national, provincial), channel rate, channel type (super trust dedicated, 5G dedicated, international SMS dedicated, rich media message dedicated, verification code dedicated, ordinary channel, etc.), channel fee, operator (China Mobile, China Unicom, China Telecom) and status (idle, not idle, available, disabled).

[0055] S102, analyzing the business data to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor;

[0056] Specifically, the linear correlation between the message factor and the channel factor is determined based on the changing law of the strong correlation between the message factor and the channel factor;

[0057] The linear correlation can be as simple as y=wx+b (w represents the linear slope, b represents the offset). For example, if the constraint is not to use the verification code channel to send ordinary short text messages, then the constraint can be y=0 (w=0, b=0); if the channel reaches a bottleneck, the constraint y=b is satisfied (x>=max, max represents the sending bottleneck critical value).

[0058] The nonlinear relationship between message factors and channel factors is determined based on the changing law of the strong correlation between message factors and channel factors.

[0059] A nonlinear relationship can be that y and x are inversely proportional to each other, y=1 / x, and y and x are parabolic, such as: and other curves.

[0060] Obtain an influence relationship between a message factor and a message factor change rate on a matching result, and a differential equation based on the influence relationship.

[0061] Differential Equations For example, the amount of transmission of a channel is associated with the rate of change of a certain message attribute value (or bottleneck association, or linear association or other). Of course, there are many types of differential equations, which will not be described in detail in the present invention.

[0062] Obtain the influence relationship between multiple message factors and the change rates of the multiple message factors on the matching results, and obtain a partial differential equation based on the influence relationship.

[0063] ‌, where F represents the partial differential equation function, U represents the unknown function, represents the independent variable (message factor), Represents U (channel) pair There are many forms of partial differential equations, which cover almost all types of relationships between message factors and channels. There are many types of partial differential equations. The previous formula only describes the general representation of first-order partial differential equations. The present invention does not elaborate on multi-order and various types of partial differential equations.

[0064] Sometimes it is not easy to analyze the correlation equation directly from the message factors and channels, but it is easy to obtain the differential (or partial differential) equation through data analysis.

[0065] S103, constructing a neural network model, injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as constraints or boundaries into the neural network of the neural network model, guiding and constraining the learning of the neural network, so as to obtain an optimal neural network model;

[0066] Specifically, on the basis of the general neural network model, known conditional constraints are used to guide and constrain neural network learning. The above conditional terms are obtained through analysis in the business. (1) Simple linear and nonlinear relationships (2) Differential equations (3) Partial differential equations are injected into the neural network as constraints or as boundaries, making learning more effective and increasing learning efficiency and the accuracy of learning results.

[0067] The sigmoid function is used as the activation function of the neural network neurons that are not injected. Specifically, it can be expressed as f(z) = ,z represents the output calculation of the neural network neuron z= +b( Input parameters, Representation parameters For the weight, b represents the offset, and the weight value w superscript i is constantly adjusted and changed through learning);

[0068] The entire neural network training process is based on the physical information constraints of partial differential equations (PDE) as an example. Let F(U(x),x) be a partial differential equation. To approximate the expected solution of U(x,θ) through neural network training, where x represents the independent variable and θ represents the weight and offset;

[0069] Calculate expected numerical loss;

[0070] The calculation of the expected numerical loss includes:

[0071] Calculate the expected numerical loss using formula (1);

[0072] Formula (1);

[0073] In the formula, is the expected numerical loss, m is the number of constraint factors, is the vector norm squared, is the input parameter, θ is the weight and offset, is the function that approximates U through a neural network, is the unknown function of the channel factor on the message factor, and the unknown function includes linear relationship, nonlinear relationship, differential equation and partial differential equation;

[0074] , where F represents the partial differential equation function, U represents the unknown function, represents the independent variable (message factor), Represents U (channel) pair The partial derivative of .

[0075] Calculate physical information loss;

[0076] The computation of physical information loss includes:

[0077] The physical information loss is calculated by formula (2);

[0078] Formula (2);

[0079] In the formula, is the physical information loss, n is the number of constraint factors, F is the partial differential equation corresponding to the expected solution of U(x,θ), is the vector norm squared, An element of the message factor X element set.

[0080] Calculating a total loss based on the expected numerical loss and the physical information loss;

[0081] Overall loss +λ (λ is a hyperparameter that balances data loss and information loss);

[0082] By minimizing the overall loss to obtain the optimal neural network model, the network parameters θ are usually updated using the gradient descent method or its variants;

[0083] S104, inputting the message to be sent into the optimal neural network model to obtain a matching channel for the message;

[0084] The method for matching a message with a channel for sending obtains business data; analyzes the business data to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between message factors and channel factors; constructs a neural network model, injects the linear relationship, nonlinear relationship, differential equation and partial differential equation between message factors and channel factors into the neural network model as constraint conditions, trains the neural network model to obtain an optimal neural network model; inputs the message to be sent into the optimal neural network model to obtain a matching channel for the message. The method solves the problem in the prior art that it is impossible to match an accurate sending channel for a message.

[0085] The message matching channel sending method uses artificial neural networks commonly used in artificial intelligence. Based on the improvement of unsupervised learning of historical data, it can effectively solve the negative impact that noise data in historical data may have on learning results. Through supervised learning such as physical information neural networks, it can not only improve learning efficiency, but also improve the expected output results.

[0086] Figure 2 This is a system architecture diagram for an embodiment of the message matching channel sending of the present invention; Figure 2 As shown, a system for sending a message matching channel provided by an embodiment of the present invention includes the following steps:

[0087] An acquisition module 10 is used to acquire business data;

[0088] A constraint determination module 20, for analyzing the business data to determine a linear relationship, a nonlinear relationship, a differential equation, and a partial differential equation between message factors and channel factors;

[0089] The constraint determination module 20 is also used for:

[0090] Determine the linear correlation between message factors and channel factors based on the changing rules of the strong correlation between message factors and channel factors;

[0091] The nonlinear relationship between message factors and channel factors is determined based on the changing law of the strong correlation between message factors and channel factors.

[0092] Obtain an influence relationship between a message factor and a message factor change rate on a matching result, and a differential equation based on the influence relationship.

[0093] Obtain the influence relationship between multiple message factors and the change rates of the multiple message factors on the matching results, and obtain a partial differential equation based on the influence relationship.

[0094] A construction module 30 is used to construct a neural network model;

[0095] A training module 40, for injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as a constraint condition or boundary into the neural network of the neural network model, guiding and constraining the neural network learning, so as to obtain an optimal neural network model;

[0096] The training module 40 is also used for:

[0097] Calculate expected numerical loss;

[0098] Calculate the expected numerical loss using formula (1);

[0099] Formula (1);

[0100] In the formula, is the expected numerical loss, m is the number of constraint factors, is the vector norm squared, is the input parameter, θ is the weight and offset, is the function that approximates U through a neural network, It is the unknown function of channel factors on message factors. The unknown function includes linear relationship, nonlinear relationship, differential equation and partial differential equation.

[0101] Calculate physical information loss;

[0102] The physical information loss is calculated by formula (2);

[0103] Formula (2);

[0104] In the formula, is the physical information loss, n is the number of constraint factors, F is the partial differential equation corresponding to the expected solution of U(x,θ), is the vector norm squared, An element of the message factor X element set.

[0105] Calculating a total loss based on the expected numerical loss and the physical information loss;

[0106] The optimal neural network model is obtained by minimizing the overall loss.

[0107] The matching module 50 is used to input the message to be sent into the optimal neural network model to obtain a matching channel for the message.

[0108] The system for sending a message through a matching channel of the present invention obtains business data through an acquisition module 10; analyzes the business data through a constraint determination module 20 to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between message factors and channel factors; constructs a neural network model through a construction module 30; injects the linear relationship, nonlinear relationship, differential equation and partial differential equation between message factors and channel factors into the neural network model as constraint conditions through a training module 40 for training to obtain an optimal neural network model; inputs the message to be sent into the optimal neural network model through a matching module 50 to obtain a matching channel for the message. The method for sending a message through a matching channel solves the problem in the prior art that it is impossible to match an accurate sending channel for a message.

[0109] Figure 3 A schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the electronic device 60 includes: a processor 601 (processor), a memory 602 (memory) and a bus 603;

[0110] The processor 601 and the memory 602 communicate with each other via the bus 603;

[0111] The processor 601 is used to call the program instructions in the memory 602 to execute the methods provided by the above-mentioned method embodiments, for example, including: obtaining business data; analyzing the business data to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor; constructing a neural network model, injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as constraints into the neural network model, training the neural network model to obtain an optimal neural network model; inputting the message to be sent into the optimal neural network model to obtain a matching channel for the message.

[0112] The present embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the methods provided by the above-mentioned method embodiments, for example, including: obtaining business data; analyzing the business data to determine the linear relationship, non-linear relationship, differential equation and partial differential equation between message factors and channel factors; constructing a neural network model, injecting the linear relationship, non-linear relationship, differential equation and partial differential equation between message factors and channel factors as constraints into the neural network model, and training the neural network model to obtain an optimal neural network model; inputting the message to be sent into the optimal neural network model to obtain a matching channel for the message.

[0113] A person skilled in the art can understand that all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various storage media that can store program codes.

[0114] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0115] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0116] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for sending a message matching channel, characterized in that: The method for sending the message matching channel specifically includes: Get business data; Analyzing the business data to determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor, including: Determine the linear relationship between message factors and channel factors based on the changing law of the strong correlation between message factors and channel factors; Based on the changing rules of the strong correlation between message factors and channel factors, the nonlinear relationship between message factors and channel factors is determined; Obtaining an influence relationship between a message factor and a message factor change rate on a matching result, and a differential equation based on the influence relationship; Obtaining the influence relationship between multiple message factors and the change rates of multiple message factors on the matching results, and a partial differential equation based on the influence relationship; Constructing a neural network model, injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as constraints or boundaries into the neural network of the neural network model, guiding and constraining the neural network learning, so as to obtain the optimal neural network model, including: Calculate expected numerical loss; Calculate physical information loss; Calculating a total loss based on the expected numerical loss and the physical information loss; By minimizing the overall loss, an optimal neural network model is obtained; The calculation of the expected numerical loss includes: Calculate the expected numerical loss using formula (1); Formula (1): In the formula, is the expected numerical loss, m is the number of constraint factors, is the vector norm squared, is the input parameter, θ is the weight and offset, is the function that approximates U through a neural network, It is the unknown function of channel factors on message factors, and the unknown function includes linear relationship, nonlinear relationship, differential equation and partial differential equation; Input the message to be sent into the optimal neural network model to obtain the matching channel of the message.

2. The method for sending a message matching channel according to claim 1, characterized in that: The computation of physical information loss includes: The physical information loss is calculated by formula (2); Formula (2): In the formula, is the physical information loss, n is the number of constraint factors, F is the partial differential equation corresponding to the expected solution of U(x,θ), is the vector norm squared, An element of the message factor X element set.

3. A system for sending messages through matching channels, characterized in that: include: Acquisition module, used to obtain business data; A constraint determination module, used to analyze the business data and determine the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor; Determine the linear relationship between message factors and channel factors based on the changing law of the strong correlation between message factors and channel factors; Based on the changing rules of the strong correlation between message factors and channel factors, the nonlinear relationship between message factors and channel factors is determined; Obtaining an influence relationship between a message factor and a message factor change rate on a matching result, and a differential equation based on the influence relationship; Obtaining the influence relationship between multiple message factors and the change rates of multiple message factors on the matching results, and a partial differential equation based on the influence relationship; Building blocks for building neural network models; A training module, for injecting the linear relationship, nonlinear relationship, differential equation and partial differential equation between the message factor and the channel factor as a constraint condition or boundary into the neural network of the neural network model, guiding and constraining the learning of the neural network to obtain an optimal neural network model, including: Calculate expected numerical loss; Calculate physical information loss; Calculating a total loss based on the expected numerical loss and the physical information loss; By minimizing the overall loss, an optimal neural network model is obtained; The calculation of the expected numerical loss includes: Calculate the expected numerical loss using formula (1); Formula (1): In the formula, is the expected numerical loss, m is the number of constraint factors, is the vector norm squared, is the input parameter, θ is the weight and offset, is the function that approximates U through a neural network, It is the unknown function of channel factors on message factors, and the unknown function includes linear relationship, nonlinear relationship, differential equation and partial differential equation; The matching module is used to input the message to be sent into the optimal neural network model to obtain a matching channel for the message.

4. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 2 are implemented.

5. A non-transitory computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 2 are implemented.

Images