Scenario financial middle-end engine processing method

Abstract

The present application provides a scenario financial middle-end engine processing method, including: obtaining transmission sample data of multiple groups of data to be processed; training multiple DQNs according to the transmission sample data and a preset first training number of times to obtain multiple first time slot allocation networks; determining the first time slot allocation network with the largest target reward value obtained according to the time slot allocation and power control results of the transmission sample data as the second time slot allocation network; training the second time slot allocation network according to the transmission sample data and a preset second training number of times to obtain a target time slot allocation network; performing scenario recognition on the data to be processed sent by the financial front desk to the financial middle-end engine to obtain a target processing background and a target link corresponding to the target processing background; performing time slot allocation according to the data to be processed, the target link and the target time slot allocation network to control the transmission power of the data to be processed when it is transmitted to the target processing background. The present application can reduce the power consumption of transmitting the data to be processed to the processing background.

Description

Technical Field

[0001] The present application relates to the technical field of data transmission of a financial middle-office engine, and specifically to a scenario financial middle-office engine processing method. Background Art

[0002] Through the financial middle-office engine, the data to be processed sent by the financial front desk can be sent to the corresponding processing backend, so that the processing backend can process the data to be processed, thereby completing the distribution of the data to be processed. Specifically, since the processing backend will have different databases, processing logic, etc. for different scenarios, the distribution of the data to be processed through the financial middle-office engine can reduce the impact of the addition, editing or deletion of the processing backend on the financial front desk. However, when the financial middle-office engine sends the data to be processed to the processing backend, it needs to be transmitted through a link, and the power consumption caused by the transmission is not stable, which makes the financial middle-office engine prone to heating, increased power consumption, and low transmission efficiency when the power consumption corresponding to the transmitted data is relatively high. Summary of the invention

[0003] The purpose of this application is to overcome the shortcomings and deficiencies in the prior art and to provide a scenario financial middle-office engine processing method that can reduce the power consumption of the financial middle-office engine in transmitting the processed data to the processing background, reduce the possibility of heat generation in the financial middle-office engine, and reduce power consumption and improve transmission efficiency.

[0004] A first aspect of an embodiment of the present application provides a scenario finance middle-end engine processing method, including:

[0005] Acquire multiple sets of transmission sample data of data to be processed; the transmission sample data is a sample of the data to be processed sent by the financial front-end to the financial middle-end engine;

[0006] Training a plurality of DQNs according to the transmission sample data and a preset first training number to obtain a plurality of first time slot allocation networks;

[0007] Determine, from each of the first time slot allocation networks, a target reward value of each of the first time slot allocation networks according to the time slot allocation and power control results of the transmission sample data, and determine the first time slot allocation network with the largest target reward value as the second time slot allocation network;

[0008] Training the second time slot allocation network according to the transmission sample data and a preset second training number to obtain a target time slot allocation network;

[0009] Receive the data to be processed sent by the financial front-end to the financial middle-end engine, perform scene recognition on the data to be processed, obtain the target processing backend for processing the data to be processed, and the target link corresponding to the target processing backend;

[0010] Timeslot allocation is performed according to the data to be processed, the target link and the target timeslot allocation network to control the transmission power of the data to be processed to the target processing background.

[0011] Further, the step of training multiple DQNs according to the transmission sample data and the preset first training times to obtain multiple first time slot allocation networks includes:

[0012] Determine the temporary reward value of each first time slot allocation network training according to the time slot and transmission power allocated by each DQN when transmitting sample data;

[0013] Obtaining a distribution reward value according to each of the DQN network parameters and the temporary reward value;

[0014] Training the network weight parameters of each of the DQNs according to the first training times and the allocated reward value of each training to obtain a plurality of first time slot allocation networks;

[0015] The step of determining the target reward value of each of the first time slot allocation networks according to the time slot allocation and power control results of the transmission sample data includes:

[0016] The temporary reward value of the last training of each DQN is determined as the target reward value.

[0017] Furthermore, the step of obtaining the allocated reward value according to the network parameters of each DQN and the temporary reward value includes:

[0018] Obtaining a reward base value and a transmission power for data transmission in the time slots allocated by each of the DQNs; the reward base value is used to indicate whether congestion is encountered when data transmission is performed according to the time slots allocated by each of the DQNs;

[0019] Obtaining the transmission time required for each of the DQNs to transmit data;

[0020] Obtaining transmission energy consumed by transmitting data according to the transmission power and the transmission time;

[0021] The temporary reward value is obtained according to the reward basic value and the transmission energy.

[0022] Further, the step of obtaining the reward base value when each of the time slots allocated by the DQN is used for data transmission includes:

[0023] If data transmission according to the time slot allocated by the DQN encounters congestion, the reward base value is determined to be a negative value; if data transmission according to the time slot allocated by the DQN does not encounter congestion, the reward base value is determined to be a positive value.

[0024] Further, the step of obtaining transmission energy consumed by transmitting data according to the transmission power and the transmission time includes:

[0025] The transmission energy is obtained by the following formula:

[0026]

[0027] Where E is the transmission energy, is the transmission power when transmitting data in time slot k allocated to the nth DQN; τ is the transmission time.

[0028] Furthermore, the step of obtaining the temporary reward value according to the reward base value and the transmission energy includes:

[0029] The temporary reward value is obtained through the following formula:

[0030]

[0031] Among them, r t is the temporary reward value, R is the basic reward value, and E is the transmission energy.

[0032] Furthermore, the step of obtaining the allocated reward value according to the network parameters of each DQN and the temporary reward value includes:

[0033] The distribution reward value of each DQN is obtained by the following formula:

[0034] y t =r t +γmaxQ(S t+1 ,a';θ t )

[0035] Among them, y t is the distribution return value; r t is the temporary reward value; γ is the discount factor, γ∈[0,1); Q(S t+1 ,a';θ t ) is the time slot allocation result of DQN.

[0036] Further, the step of training the network weight parameters of each DQN according to the first training times and the allocated reward value of each training to obtain a plurality of first time slot allocation networks includes:

[0037] According to the first training number and the preset gradient direction, training the network weight parameters of each of the DQNs;

[0038] Obtaining the loss function after each DQN training;

[0039] The DQN corresponding to the loss function that is less than a preset function threshold is determined as the first time slot allocation network.

[0040] Further, the step of training the network weight parameters of each of the DQNs according to the first training times and the preset gradient direction includes:

[0041] The network weight parameters after DQN training are obtained through the following formula:

[0042]

[0043] Among them, θ t+1 is the network weight parameter after the tth training; θ t is the network weight parameter before the tth training; η is the learning rate; Q(S t+1 ,a';θ t ) is the time slot allocation result of DQN after the tth training; is the gradient direction.

[0044] Further, the step of obtaining the loss function of each DQN after training includes:

[0045] The loss function after each DQN training is obtained through the following formula:

[0046]

[0047] Among them, Loss(θ) is the loss function; targetQ(S t ,a;θ t ) is the target time slot allocation result, Q(S t ,a;θ t ) is the DQN time slot allocation result after training, N is the first training number

[0048] Compared with the prior art, the present application can train multiple DQNs according to multiple preset first training times of transmission sample data of multiple groups of data to be processed, obtain multiple first time slot allocation networks, and then determine the corresponding target reward value according to the time slot allocation and power control results of the multiple first time slot allocation networks, so as to train the second time slot allocation network (the first time slot allocation network with the largest target reward value) according to the transmission sample data and the preset second training times to obtain the target time slot allocation network; when applied, the scene recognition is performed on the data to be processed sent by the financial front desk to obtain the corresponding target processing background and target link, and then time slot allocation is performed according to the data to be processed, the target link and the target time slot allocation network to control the transmission power of the data to be processed to the target processing background, thereby achieving the technical effect of reducing the power consumption of the financial middle platform engine in transmitting the data to be processed to the processing background, reducing the possibility of heat generation of the financial middle platform engine, and reducing power consumption and improving transmission efficiency.

[0049] In order to provide a clearer understanding of the present application, the specific implementation of the present application will be described below in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a flowchart of a scenario finance middle-office engine processing method according to an embodiment of the present application.

[0051] Figure 2 This is a flowchart of step S21 of the scenario finance middle-office engine processing method of an embodiment of the present application.

[0052] Figure 3 This is a flowchart of step S23 of the scenario finance middle-office engine processing method of an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the objectives, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0054] It should be clear that the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the embodiments of the present application.

[0055] When the following description relates to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. In the description of the present application, it should be understood that the terms "first", "second", "third", etc. are only used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence, nor can they be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances. The singular forms of "a", "said" and "the" used in the present application and the appended claims are also intended to include the majority form, unless the context clearly indicates other meanings. The words "if" / "if" used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0056] In addition, in the description of this application, unless otherwise specified, "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0057] See also Figure 1 , which is a flow chart of a scenario finance middle office engine processing method according to an embodiment of the present application, including:

[0058] S1: Acquire multiple groups of transmission sample data of data to be processed; the transmission sample data is a sample of data to be processed sent by the financial front-end to the financial middle-end engine;

[0059] Data to be processed refers to data sent by the financial front desk and needs to be processed by the processing backend. Among them, the financial front desk first sends the data to be processed to the financial middle office engine, and then the financial middle office engine transmits the data to be processed to the processing backend, so this application can be applied to the financial middle office engine.

[0060] The transmission sample data may be historical data to be processed, for example, the data to be processed previously sent by the financial front desk to the financial middle office engine as a sample. The transmission sample data may also be data to be processed that is driven by the financial front desk to send to the financial middle office engine in order to train the time slot allocation model.

[0061] S2: training a plurality of DQNs according to the transmission sample data and a preset first training number to obtain a plurality of first time slot allocation networks;

[0062] The at least two DQNs trained according to the first training number are DQNs with the same initial network structure and the same initial network parameters. The first training number is the number of trainings for the first training preset by the user, for example, set to 50 times, 60 times, etc.

[0063] At the beginning of training, you can first create an experience replay pool of any size, such as an experience replay pool of 10,000, 15,000, 20,000, etc. In the initial state, the experience replay pool is empty. After step S1, multiple sets of transmission sample data can be stored in the experience replay pool. The data in the experience replay pool is used for subsequent neural network training. When 1,000 pieces are stored, training can begin. During training, 64 pieces of data can be taken from the experience replay pool and sent to each DQN for training, and the network parameters are updated and the training data is stored in the experience replay pool.

[0064] S3: Determine the target reward value of each of the first time slot allocation networks according to the time slot allocation and power control results of the transmission sample data, and determine the first time slot allocation network with the largest target reward value as the second time slot allocation network;

[0065] The target reward value refers to the last temporary reward value of each DQN after reaching the preset first training times, and during the training process, the temporary reward value obtained in each training is used to indicate the training direction of the DQN network parameters. The temporary reward value of each training is related to the time slot allocation and power control results. For example, it can be expressed that the smaller the power, the larger the corresponding target reward value.

[0066] It should be noted that, during the training process, although the target reward value plays an indicative role in the training direction of the DQN network parameters, the result of each time slot allocation of the DQN is random. Therefore, even if the time slot allocation training is also conducted for power, as the number of training times increases, the network parameters of each DQN will show different degrees of difference. Therefore, after reaching the first number of training times, the target reward values ​​of each DQN will also be different. The second time slot allocation network with the largest target reward value is the DQN with the best time slot allocation training effect for power and the most training potential. Continuing training here in the second time slot allocation network can obtain a DQN suitable for time slot allocation and power control.

[0067] S4: training the second time slot allocation network according to the transmission sample data and a preset second training number to obtain a target time slot allocation network;

[0068] The second training number and the first training number are independent training numbers, and optionally, the second training number is greater than the first training number. Performing secondary training on the second time slot allocation network according to the second training number can improve the accuracy of the second time slot allocation network in allocating time slots for power, so as to obtain a target time slot allocation network.

[0069] S5: receiving the data to be processed sent by the financial front-end to the financial middle-end engine, performing scene recognition on the data to be processed, and obtaining a target processing backend for processing the data to be processed, and a target link corresponding to the target processing backend;

[0070] Among them, different application scenarios correspond to different processing backends. When the financial middle-office engine receives the data to be processed, it will identify the application scenario corresponding to the data to be processed, and then determine the appropriate processing backend based on the application scenario, and then transmit the data to be processed to the processing backend.

[0071] S6: Perform time slot allocation according to the data to be processed, the target link and the target time slot allocation network to control the transmission power of the data to be processed to the target processing background.

[0072] Specifically, the data to be processed and the target link can be used as inputs of the target time slot allocation network, so that the target time slot allocation network allocates time slots to the data to be processed according to the link status of the target link, so as to control the transmission power of the data to be processed to the target processing background according to the allocated time slots.

[0073] Among them, by allocating time slots for the data transmission process to be processed through the target time slot allocation network, the transmission success probability and transmission power of the allocated time slots can be effectively controlled, making data transmission more efficient. It can also reduce the power consumption of the financial middle-office engine in transmitting the data to be processed to the processing background, reduce the possibility of heat generation of the financial middle-office engine, and reduce power consumption and improve transmission efficiency.

[0074] In a feasible embodiment, the step S2: training multiple DQNs according to the transmission sample data and the preset first training times to obtain multiple first time slot allocation networks includes:

[0075] S21: determining a temporary reward value for each of the first time slot allocation networks during training according to the time slot and transmission power allocated by each DQN for transmitting sample data;

[0076] S22: Obtaining a distribution reward value according to the network parameters of each of the DQNs and the temporary reward value;

[0077] The distribution reward value refers to the optimal Q table value of each DQN during the current training, which can be calculated based on the network parameters and the corresponding temporary reward value.

[0078] S23: training network weight parameters of each of the DQNs according to the first training times and the allocated reward value of each training, to obtain a plurality of first time slot allocation networks;

[0079] Since the allocation reward value is related to the temporary reward value, and the temporary reward value is related to whether the data transmission after the time slot allocation encounters congestion and the transmission power, the allocation reward value is also related to whether the data transmission after the time slot allocation is smooth and the transmission power. Therefore, the allocation reward value has the function of guiding the training of DQN.

[0080] The step S3: determining the target reward value of each of the first time slot allocation networks according to the time slot allocation and power control results of the transmission sample data, comprises:

[0081] S31: Determine the temporary reward value of the last training of each DQN as the target reward value.

[0082] It can be understood that the last training refers to the last training after reaching the preset first training times. At this time, the target reward value of each DQN can indicate the training effect of each DQN. Specifically, the higher the target reward value, the better the training effect of the corresponding DQN.

[0083] In this embodiment, the allocation reward value can be obtained according to the time slot allocation and power control results to train the DQN, so that the smoothness of the data transmission and the transmission power corresponding to the allocated time slot can guide the training, and then the training effect of each DQN in the initial training can be determined according to the size of the target reward value of the last training of each DQN.

[0084] See also Figure 2 In a feasible embodiment, the step of obtaining the allocated reward value according to the network parameters of each DQN and the temporary reward value in S21 includes:

[0085] S211: Obtaining a reward base value and a transmission power for data transmission in the time slots allocated by each of the DQNs; the reward base value is used to indicate whether congestion is encountered when data transmission is performed according to the time slots allocated by each of the DQNs;

[0086] Whether congestion is encountered refers to whether the allocated time slot is an occupied time slot. When the allocated time slot is an occupied time slot, data transmission can be realized only after the allocated time slot is idle. At this time, it is manifested as congestion during data transmission.

[0087] The transmission power can be detected by detection equipment, and can be obtained by monitoring the changes in power consumption of the financial middle-office engine.

[0088] S212: Obtaining the transmission time required for each of the DQNs to transmit data;

[0089] Transmission time refers to the time it takes for the financial middle-end engine to transmit data to the target processing backend after the time slots of each DQN are allocated.

[0090] S213: Obtaining transmission energy consumed by transmitting data according to the transmission power and the transmission time;

[0091] Specifically, the transmission energy can be obtained by the following formula:

[0092]

[0093] Where E is the transmission energy, is the transmission power when transmitting data in time slot k allocated to the nth DQN; τ is the transmission time.

[0094] S214: Obtaining the temporary reward value according to the reward basic value and the transmission energy.

[0095] Specifically, the temporary reward value can be obtained by the following formula:

[0096]

[0097] Among them, r t is the temporary reward value, R is the basic reward value, and E is the transmission energy.

[0098] In a feasible embodiment, the step of S211: obtaining the reward base value for data transmission in the time slots allocated by each DQN includes:

[0099] If data transmission according to the time slot allocated by the DQN encounters congestion, the reward base value is determined to be a negative value; if data transmission according to the time slot allocated by the DQN does not encounter congestion, the reward base value is determined to be a positive value.

[0100] The reward base value is an indicator value used to indicate the congestion encountered during data transmission, wherein negative values ​​may be -1, -2, -10, etc., and positive values ​​may be 1, 2, 10, etc. Specifically, the reward base value may be set by the user, and the reward base value may be used to measure whether the allocated time slot is idle and the transmission power when data is transmitted through the time slot. For example, when the user believes that the importance of whether the time slot is idle is greater than the transmission power when data is transmitted in the time slot, the user may set the reward base value so that the absolute value of the reward base value is greater than the inverse of the transmission energy corresponding to the transmission power; when the importance of the idle state is less than the transmission power when data is transmitted in the time slot, the user may set the reward base value so that the absolute value of the reward base value is less than the inverse of the transmission energy corresponding to the transmission power.

[0101] In a feasible embodiment, the step of obtaining a distribution reward value according to the network parameters of each DQN and the temporary reward value in S22 includes:

[0102] The distribution reward value of each DQN is obtained by the following formula:

[0103] y t =r t +γmaxQ(S t+1 ,a';θ t )

[0104] Among them, y t is the distribution return value; r t is the temporary reward value; γ is the discount factor, γ∈[0,1); Q(S t+1 ,a';θ t ) is the time slot allocation result of DQN.

[0105] See also Figure 3 In a feasible embodiment, the step S23: training the network weight parameters of each DQN according to the first training times and the allocated reward value of each training to obtain a plurality of first time slot allocation networks includes:

[0106] S231: training network weight parameters of each of the DQNs according to the first training times and a preset gradient direction;

[0107] Specifically, the network weight parameters after DQN training can be obtained by the following formula:

[0108]

[0109] Among them, θ t+1 is the network weight parameter after the tth training; θ t is the network weight parameter before the tth training; η is the learning rate; Q(S t+1 ,a';θ t ) is the time slot allocation result of DQN after the tth training; is the gradient direction.

[0110] S232: Obtain the loss function of each DQN after training;

[0111] Specifically, the loss function after each DQN training can be obtained by the following formula:

[0112]

[0113] Among them, Loss(θ) is the loss function; targetQ(S t ,a;θ t ) is the target time slot allocation result, Q(S t ,a;θ t) is the DQN time slot allocation result after training, and N is the first training number.

[0114] S233: Determine the DQN corresponding to the loss function that is less than a preset function threshold as the first time slot allocation network.

[0115] Among them, the preset function threshold is used to determine whether the training of the DQN meets the user's expected training requirements. For example, if the loss function corresponding to the DQN is less than the preset function threshold, it means that the training of the DQN meets the user's expected training requirements, and the trained DQN can be used as the first time slot allocation network; if the corresponding loss function is greater than or equal to the preset function threshold, it means that the training of the DQN does not meet the user's expected training requirements, and the DQN needs to continue training or needs to be retrained.

[0116] Compared with the prior art, the present application can train multiple DQNs according to multiple preset first training times of transmission sample data of multiple groups of data to be processed, obtain multiple first time slot allocation networks, and then determine the corresponding target reward value according to the time slot allocation and power control results of the multiple first time slot allocation networks, so as to train the second time slot allocation network (the first time slot allocation network with the largest target reward value) according to the transmission sample data and the preset second training times to obtain the target time slot allocation network; when applied, the scene recognition is performed on the data to be processed sent by the financial front desk to obtain the corresponding target processing background and target link, and then time slot allocation is performed according to the data to be processed, the target link and the target time slot allocation network to control the transmission power of the data to be processed to the target processing background, thereby achieving the technical effect of reducing the power consumption of the financial middle platform engine in transmitting the data to be processed to the processing background, reducing the possibility of heat generation of the financial middle platform engine, and reducing power consumption and improving transmission efficiency.

[0117] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0118] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the function selected in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 function selected in a box or multiple boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 steps for the function selected in a box or multiple boxes.

[0120] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0121] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0122] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0123] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0124] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A scenario financial middle-end engine processing method, characterized in that: include: Obtain transmission sample data of multiple groups of data to be processed; The transmission sample data is a sample of the data to be processed sent by the financial front-end to the financial middle-end engine; Training a plurality of DQNs according to the transmission sample data and a preset first training number to obtain a plurality of first time slot allocation networks; Determine, from each of the first time slot allocation networks, a target reward value of each of the first time slot allocation networks according to the time slot allocation and power control results of the transmission sample data, and determine the first time slot allocation network with the largest target reward value as the second time slot allocation network; Training the second time slot allocation network according to the transmission sample data and a preset second training number to obtain a target time slot allocation network; Receive the data to be processed sent by the financial front-end to the financial middle-end engine, perform scene recognition on the data to be processed, obtain the target processing backend for processing the data to be processed, and the target link corresponding to the target processing backend; Performing time slot allocation according to the data to be processed, the target link and the target time slot allocation network to control the transmission power of the data to be processed to the target processing background; The step of training a plurality of DQNs according to the transmission sample data and a preset first training number to obtain a plurality of first time slot allocation networks includes: Determine the temporary reward value of each first time slot allocation network training according to the time slot and transmission power allocated by each DQN when transmitting sample data; Obtaining a distribution reward value according to each of the DQN network parameters and the temporary reward value; Training the network weight parameters of each of the DQNs according to the first training times and the allocated reward value of each training to obtain a plurality of first time slot allocation networks; The step of determining the target reward value of each of the first time slot allocation networks according to the time slot allocation and power control results of the transmission sample data includes: The temporary reward value of the last training of each DQN is determined as the target reward value.

2. The scenario financial middle-end engine processing method according to claim 1 is characterized in that: The step of obtaining the allocated reward value according to the network parameters of each DQN and the temporary reward value comprises: Obtaining a reward base value and a transmission power for data transmission in the time slots allocated by each of the DQNs; the reward base value is used to indicate whether congestion is encountered when data transmission is performed according to the time slots allocated by each of the DQNs; Obtaining the transmission time required for each of the DQNs to transmit data; Obtaining transmission energy consumed by transmitting data according to the transmission power and the transmission time; The temporary reward value is obtained according to the reward basic value and the transmission energy.

3. The scenario financial middle-end engine processing method according to claim 2 is characterized in that: The step of obtaining the reward base value for each time slot allocated by the DQN for data transmission comprises: If data transmission according to the time slot allocated by the DQN encounters congestion, the reward base value is determined to be a negative value; if data transmission according to the time slot allocated by the DQN does not encounter congestion, the reward base value is determined to be a positive value.

4. The scenario financial middle-end engine processing method according to claim 2 is characterized in that: The step of obtaining transmission energy consumed by transmitting data according to the transmission power and the transmission time comprises: The transmission energy is obtained by the following formula: Where E is the transmission energy, is the transmission power when transmitting data in time slot k allocated to the nth DQN; τ is the transmission time.

5. The scenario financial middle platform engine processing method according to claim 2 is characterized in that: The step of obtaining the temporary reward value according to the reward base value and the transmission energy comprises: The temporary reward value is obtained through the following formula: Among them, r t is the temporary reward value, R is the basic reward value, and E is the transmission energy.

6. The scenario financial middle-end engine processing method according to claim 1 is characterized in that: The step of obtaining the allocated reward value according to the network parameters of each DQN and the temporary reward value comprises: The distribution reward value of each DQN is obtained by the following formula: y t =r t +γmaxQ(S t+1 ,a';θ t ) Among them, y t is the distribution return value; r t is the temporary reward value; γ is the discount factor, γ∈[0,1); Q(S t+1 ,a';θ t ) is the time slot allocation result of DQN.

7. The scenario financial middle-end engine processing method according to claim 1 is characterized in that: The step of training the network weight parameters of each DQN according to the first training times and the allocated reward value of each training to obtain a plurality of first time slot allocation networks includes: According to the first training times and the preset gradient direction, training the network weight parameters of each of the DQNs; Obtaining the loss function after each DQN training; The DQN corresponding to the loss function that is less than a preset function threshold is determined as the first time slot allocation network.

8. The scenario financial middle-end engine processing method according to claim 7 is characterized in that: The step of training the network weight parameters of each of the DQNs according to the first training times and the preset gradient direction includes: The network weight parameters after DQN training are obtained through the following formula: Among them, θ t+1 is the network weight parameter after the tth training; θ t is the network weight parameter before the tth training; η is the learning rate; Q(S t+1 ,a';θ t ) is the time slot allocation result of DQN after the tth training; is the gradient direction.

9. The scenario financial middle-end engine processing method according to claim 7 is characterized in that: The step of obtaining the loss function of each DQN after training includes: The loss function after each DQN training is obtained through the following formula: Among them, Loss(θ) is the loss function; targetQ(S t ,a;θ t ) is the target time slot allocation result, Q(S t ,a;θ t ) is the DQN time slot allocation result after training, and N is the first training number.

Images