Network optimization task execution method and device, electronic equipment and storage medium

Abstract

The invention provides a network optimization task execution method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining node type information and node level information of a first task node, obtaining initial task data according to the node type information and the node level information, and obtaining an initial network optimization model, the initial network optimization model has corresponding historical rule data, training the initial network optimization model according to the initial task data and the historical rule data to obtain a target network optimization model, and executing a second network optimization task in the network optimization system according to the target network optimization model. According to the method and the device, full-process data marking can be carried out on a plurality of task nodes of network optimization, data marking in a thinking analysis process of network optimization is realized, the operation behavior marking rate is effectively improved, effective information after loss is avoided, the forming rate of reusable knowledge is effectively improved, and the data marking efficiency is improved. And the execution effect and the execution efficiency of the network optimization task of the network optimization system are ensured.

Description

Technical Field

[0001] The present disclosure relates to the technical field of network optimization, and in particular to a method, device, electronic device and storage medium for executing a network optimization task. Background Art

[0002] In the related art, in the process of processing network optimization tasks by the network optimization system, network optimization experts usually re-identify, analyze and formulate solutions based on the problems described in the network optimization tasks and the data provided, combined with network optimization experience, and using various professional auxiliary tools provided in the network optimization system.

[0003] In this way, the thinking and analysis process of network optimization experts is invisible, the operational behavior trace rate is low and the effective information loss rate is high, the formation rate of reusable knowledge is low, and the conversion of expert experience into solidified network optimization processing rules is time-consuming and has a low conversion rate, which affects the execution effect and efficiency of the network optimization system for network optimization tasks. Summary of the invention

[0004] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0005] To this end, the purpose of the present disclosure is to propose a network optimization task execution method, device, electronic device, storage medium and computer program product.

[0006] The first aspect of the present disclosure provides a method for executing a network optimization task, including: obtaining node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by a network optimization system, and the node level information indicates the node level at which the first task node is located in the first network optimization task; obtaining initial task data based on the node type information and the node level information; obtaining an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data indicates data related to network optimization rules generated by the initial network optimization model; training the initial network optimization model based on the initial task data and the historical rule data to obtain a target network optimization model; and executing a second network optimization task in the network optimization system based on the target network optimization model, wherein the second network optimization task is a newly added network optimization task in the network optimization system.

[0007] The second aspect of the present disclosure proposes a network optimization task execution device, including: a first acquisition module, used to acquire node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by a network optimization system, and the node level information indicates the node level at which the first task node is located in the first network optimization task; a second acquisition module, used to acquire initial task data based on the node type information and the node level information; a third acquisition module, used to acquire an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data indicates data related to network optimization rules generated by the initial network optimization model; a training module, used to train the initial network optimization model based on the initial task data and the historical rule data to obtain a target network optimization model; a processing module, used to execute a second network optimization task in the network optimization system based on the target network optimization model, wherein the second network optimization task is a newly added network optimization task in the network optimization system.

[0008] The network optimization task execution method, device, electronic device, storage medium and computer program product proposed in the embodiments of the present disclosure at least include the following beneficial effects: by obtaining node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by a network optimization system, and the node level information indicates the node level at which the first task node is located in the first network optimization task, initial task data are obtained according to the node type information and the node level information, and an initial network optimization model is obtained, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data indicates data related to the network optimization rules generated by the initial network optimization model, the initial network optimization model is trained according to the initial task data and the historical rule data to obtain a target network optimization model, and according to the target network optimization model, a second network optimization task in the network optimization system is executed, wherein the second network optimization task is a newly added network optimization task in the network optimization system, and data traces of the entire process of multiple task nodes of network optimization can be recorded, so as to realize data traces of the thinking and analysis process of network optimization, effectively improve the low trace rate of operation behavior, avoid the loss of effective information, effectively improve the formation rate of reusable knowledge, and ensure the execution effect and efficiency of the network optimization tasks of the network optimization system.

[0009] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0011] Figure 1 It is a flowchart of a method for executing a network optimization task proposed in an embodiment of the present disclosure;

[0012] Figure 2 A schematic diagram of a network optimization scenario service process design in an embodiment of the present disclosure;

[0013] Figure 3 is a flowchart of a method for executing a network optimization task proposed in another embodiment of the present disclosure;

[0014] Figure 4 It is a structural diagram of a three-layer five-dimensional data trace model in an embodiment of the present disclosure;

[0015] Figure 5 is a schematic diagram of the model training and learning process in the embodiment of the present disclosure;

[0016] Figure 6 is a flowchart of a method for executing a network optimization task proposed in another embodiment of the present disclosure;

[0017] Figure 7 is a schematic diagram of model effectiveness control in an embodiment of the present disclosure;

[0018] Figure 8 is a schematic diagram of a network optimization collaborative execution process in an embodiment of the present disclosure;

[0019] Fig. 9 Schematic diagram of the intelligent evolution scheme of the network optimization system in the embodiment of the present disclosure

[0020] Fig.10 It is a structural diagram of a network optimization task execution device proposed in an embodiment of the present disclosure;

[0021] Fig.11 is a schematic diagram of the structure of a network optimization task execution device proposed in another embodiment of the present disclosure;

[0022] Fig.12 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0023] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present disclosure, and are not to be construed as limitations of the present disclosure. On the contrary, the embodiments of the present disclosure include all changes, modifications, and equivalents that fall within the spirit and connotation of the appended claims.

[0024] Figure 1 It is a flowchart of a network optimization task execution method proposed in an embodiment of the present disclosure.

[0025] It should be noted that the executor of the network optimization task execution method of this embodiment is a network optimization task execution device, which can be implemented by software and / or hardware, and can be configured in an electronic device, without limitation.

[0026] like Figure 1 As shown, the network optimization task execution method includes:

[0027] S101: Obtain node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by a network optimization system, and the node level information indicates a node level of the first task node in the first network optimization task.

[0028] Among them, the network optimization system refers to a working platform used to perform network optimization tasks, and the network optimization system can be used to identify, analyze and formulate execution plans for network optimization tasks.

[0029] Among them, the first network optimization task refers to the network optimization task currently executed by the network optimization system, which is to leave data traces for the entire task execution process. The network optimization task refers to monitoring the operating status of the network, including network failures, network performance, and network quality, and triggering the corresponding processing flow to optimize the network problem after a problem is found.

[0030] The first task node refers to the current task node of the first network optimization task currently executed by the network optimization system.

[0031] Among them, the node type information refers to the data information used to indicate the process task type corresponding to the first task node. The process task type is determined during the design of the network optimization processing scenario process, which are five process task types: perception, analysis, decision-making, execution, and evaluation. Correspondingly, the node type information can indicate the perception node, analysis node, decision node, execution node, and evaluation node.

[0032] The node level information indicates the node level at which the first task node is located in the first network optimization task, and the node level may include: user level, service level, and network level.

[0033] In the disclosed embodiment, when obtaining the node type information and node level information of the first task node, a process design for the network optimization processing scenario can be first performed, and the design corresponds to five process task types: perception, analysis, decision-making, execution, and evaluation. After the process design is completed, the node type information can indicate a perception node, an analysis node, a decision node, an execution node, and an evaluation node. When the process is running, the first task node will complete the corresponding process tasks of these five process task types according to the process logic. When the first task node is executed, the process task type of the first task node is obtained to obtain the node type information of the first task node, the node level of the first task node in the network optimization system is obtained, and the node level corresponding to the first task node is determined from the user layer, business layer, and network layer as the node level information.

[0034] S102: Acquire initial task data according to the node type information and the node level information.

[0035] After acquiring the node type information and node level information of the first task node as described above, the embodiment of the present disclosure may acquire initial task data according to the node type information and node level information.

[0036] Among them, the initial task data refers to the trace data generated on the network optimization system during the execution of the first network optimization task, when the first task node executes each task process logic in sequence. The initial task data is the task data directly obtained when the trace data is not standardized.

[0037] In the disclosed embodiment, when the initial task data is obtained according to the node type information and the node level information, the five task process logics indicated by the node type information and the three node levels indicated by the node level information obtained from the process design of the network optimization processing scenario can be used to establish a three-layer, five-dimensional trace data model, and data is stored according to the three-layer, five-dimensional trace data model, wherein the three layers are: user layer, business layer, and network layer, and the five dimensions are: perception, analysis data and analysis behavior, decision-making, execution, and evaluation. Under this design method, when the first task node runs to the corresponding dimension, that is, the perception, analysis, decision-making, execution, and evaluation task type, all operation behaviors of the expert on the network optimization system and the operation behaviors automatically performed by the network optimization system will be recorded as corresponding trace data, and the trace data will be stored according to the node level indicated by the node level information to form a three-layer, five-dimensional trace data model to obtain the initial task data.

[0038] For example, Figure 2 As shown, Figure 2 The schematic diagram of the business process design of the network optimization scenario in the embodiment of the present disclosure is that the network optimization system can be operated by an expert to complete the first network optimization task by performing different tasks in perception, cognition (analysis task), decision-making, execution and evaluation. In these task type stages, the task can be implemented manually or automatically by the system. When the first task node runs to the corresponding process node, that is, the perception, analysis, decision-making, execution, and evaluation task type, all the operation behaviors of the expert on the network optimization system and the operation behaviors automatically performed by the network optimization system will be recorded as corresponding trace data, thereby obtaining the initial task data.

[0039] S103: Acquire an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data represents data related to the network optimization rules generated by the initial network optimization model.

[0040] Among them, the initial network optimization model refers to a network optimization model preset in the network optimization system for executing the first network optimization task, or it can also be an artificial intelligence processing model that has been obtained with a task execution plan corresponding to a determined network optimization task.

[0041] Among them, historical rule data represents data related to the network optimization rules generated by the initial network optimization model. The historical rule data can be the network optimization rules generated by the historical network optimization tasks completed by the initial network optimization model, as well as task information data related to the network optimization rules, etc., or it can also be data related to the execution process of task one and the historical network optimization task, without limitation.

[0042] In the disclosed embodiment, a network optimization model preset in the network optimization system for executing a first network optimization task can be obtained as an initial network optimization model, and an artificial intelligence processing model having a task execution plan corresponding to a determined network optimization task can also be obtained as the initial network optimization model. Accordingly, network optimization rules generated by historical network optimization tasks executed by the initial network optimization model and task information data related to the network optimization rules are obtained as historical rule data.

[0043] S104: Train an initial network optimization model according to the initial task data and the historical rule data to obtain a target network optimization model.

[0044] After obtaining the initial task data according to the node type information and the node level information and obtaining the initial network optimization model, the embodiment of the present disclosure can train the initial network optimization model according to the initial task data and the historical rule data to obtain the target network optimization model.

[0045] The target network optimization model refers to a network optimization model obtained by training an initial network optimization model based on initial task data and historical rule data, after the model performance is optimized, which can be used to determine the task execution rules of the network optimization task.

[0046] In the disclosed embodiment, when the initial network optimization model is trained according to the initial task data and the historical rule data to obtain the target network optimization model, the conditions for triggering the training of the initial network optimization model and the number of model training times can be set first. For example, the training conditions of the initial network optimization model can be set to be triggered according to a certain number of work orders (network optimization tasks) completed every day, or when a message sent by the following model production participation control stage is received to automatically initiate a self-learning task, or the training conditions can also be manually initiated model training, and then when the initial network optimization model meets the training conditions, it can be trained according to the initial task data and the historical rule data according to a preset number of training times. The initial task data can be standardized and converted into data that can be used for model training, and then multiple model trainings are performed based on the converted data and the historical rule data to obtain multiple candidate network optimization models. After each training, the corresponding candidate network optimization model and its corresponding version number are obtained, and then the model performance of the multiple candidate network optimization models is compared, and the candidate network optimization model with the best model performance is selected as the target network optimization model.

[0047] In other embodiments, a comparison process of the model training learning results can also be performed, and the candidate network optimization models of each complete model training are compared in terms of model testing indicators, and manual confirmation is provided to manually select the optimal model. At the same time, automatic comparison of learning results is also supported, and each complete training process of each candidate network optimization model is compared and ranked according to the test accuracy on the test set.

[0048] Optionally, in some embodiments, an initial network optimization model is trained according to initial task data and historical rule data to obtain a target network optimization model, a preset number of model training times and model training conditions are obtained, and it is determined whether the initial network optimization model meets the model training conditions. When the initial network optimization model meets the model training conditions and the number of training times of the initial network optimization model does not reach the preset number of model training times, the initial network optimization model is trained according to the initial task data and historical rule data to obtain the target network optimization model, wherein the model training conditions include at least any one of the following: a model training trigger message has been generated in the network optimization system, and the number of first network optimization tasks completed by the initial network optimization model is greater than a task number threshold.

[0049] Among them, the model training condition refers to a pre-set condition for determining whether to train and learn the initial network optimization model, and the model training condition includes at least any one of the following: a model training trigger message has been generated in the network optimization system, and the number of the first network optimization tasks completed by the initial network optimization model is greater than the task number threshold.

[0050] Among them, the model training trigger message refers to a data message used to trigger the training and learning of the initial network optimization model. The model training trigger message can be generated by the network optimization system to automatically trigger the training of the initial network optimization model, or it can be generated manually, without limitation.

[0051] The task quantity threshold refers to a preset task quantity threshold value for determining the initial network optimization model training based on the number of completed first network optimization tasks.

[0052] In an embodiment of the present disclosure, an initial network optimization model is trained according to initial task data and historical rule data to obtain a target network optimization model, a preset model training number and model training conditions are obtained, and it is determined whether the initial network optimization model meets the model training conditions. When the initial network optimization model meets the model training conditions and the training number of the initial network optimization model does not reach the preset model training number, for example, when the number of first network optimization tasks completed by the initial network optimization model is greater than a task number threshold, or a model training trigger message has been generated in the network optimization system, it is determined to start training the initial network optimization model, and the initial network optimization model is trained according to the initial task data and historical rule data to obtain the target network optimization model.

[0053] For example, the initial task data can be combined as the optimization update for the parameters of the historical rule data after machine learning, and then the updated historical rule data can be used to train and learn the initial network optimization model. The entire training and learning process follows the standard machine learning training process. The algorithm used may include a collection of publicly available open source algorithms or a self-developed algorithm, without any restriction on this. The learning and training method also follows the training method of the algorithm collection used. After training, the corresponding candidate network optimization model is obtained. Each time it is generated, the data will increase. Alternatively, the initial task data and all the data in the historical rule data are used to learn the initial network optimization model. The entire learning process follows the standard machine learning training process. The algorithm used may include a collection of publicly available open source algorithms or a self-developed algorithm, without any restriction on this. The learning and training method also follows the training method of the algorithm collection used. After training, the corresponding candidate network optimization model is obtained. Each time it is generated, the data will increase. Then, the best performance can be selected from the multiple candidate network optimization models obtained through training as the target network optimization model.

[0054] Among them, the learning process of model training can be visualized in the network optimization system, and the test results and indicator parameters of the initial network optimization model on the training test set at each epoch in the training process (a complete training of the model using the training data is an epoch), including indicator data such as loss value loss and accuracy rate acc, are stored and visualized in charts through the specific pages of the network optimization system.

[0055] S105: Execute a second network optimization task in the network optimization system according to the target network optimization model, wherein the second network optimization task is a newly added network optimization task in the network optimization system.

[0056] The second network optimization task refers to a newly added network optimization task in the network optimization system, and the second network optimization task in the network optimization system can be executed according to the target network optimization model.

[0057] After the initial network optimization model is trained according to the initial task data and the historical rule data to obtain the target network optimization model, the embodiment of the present disclosure can execute the second network optimization task in the network optimization system according to the target network optimization model.

[0058] In the implementation of the present disclosure, when executing the second network optimization task in the network optimization system according to the target network optimization model, the optimization rules corresponding to the second network optimization task can be determined according to the task work order of the second network optimization task, the problems described in the work order and the data provided, using the target network optimization model, and the second network optimization task can be executed based on the optimization rules.

[0059] In other embodiments, a production method that combines manual work, expert rules, and a target network optimization model can be used simultaneously to formulate respective network optimization rules, and set a rule decision model to perform network optimization rules, thereby obtaining the network optimization rules with the best final performance to execute the second network optimization task.

[0060] In this embodiment, by obtaining the node type information and node level information of the first task node, wherein the first task node is the task node of the first network optimization task currently executed by the network optimization system, and the node level information indicates the node level of the first task node in the first network optimization task, initial task data is obtained according to the node type information and the node level information, and an initial network optimization model is obtained, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data indicates data related to the network optimization rules generated by the initial network optimization model, and the initial network optimization model is trained according to the initial task data and the historical rule data to obtain a target network optimization model, and according to the target network optimization model, a second network optimization task in the network optimization system is executed, wherein the second network optimization task is a newly added network optimization task in the network optimization system, and the data trace of the whole process of multiple task nodes of network optimization can be recorded, so as to realize the data trace of the thinking and analysis process of network optimization, effectively improve the low trace rate of operation behavior, avoid the loss of effective information, effectively improve the formation rate of reusable knowledge, and ensure the execution effect and efficiency of the network optimization tasks of the network optimization system.

[0061] Figure 3 It is a flowchart of a network optimization task execution method proposed in an embodiment of the present disclosure.

[0062] like Figure 3 As shown, the network optimization task execution method includes:

[0063] S301: Obtain node type information and node level information of the first task node.

[0064] For the detailed description of S301 , please refer to the above embodiment, which will not be repeated here.

[0065] S302: Determine the node process execution logic corresponding to the first task node according to the node type information.

[0066] The node process execution logic refers to the task steps that need to be completed during the execution of the first task node.

[0067] In the embodiment of the present disclosure, when determining the node process execution logic corresponding to the first task node based on the node type information, the task steps that need to be completed during the execution of the first task node can be obtained from the network optimization based on the node type, and the obtained task steps can be used as the node process execution logic.

[0068] S303: According to the node level information, the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level is obtained.

[0069] Among them, execution behavior data refers to the trace data generated by all operation behaviors when completing the node process execution logic on the network optimization system.

[0070] After the node process execution logic corresponding to the first task node is determined according to the node type information, the present disclosure can respectively obtain the trace data generated by all operation behaviors of the system running to the business layer, user layer, and network layer when the node process execution logic is completed on the network optimization system, and use the obtained trace data as execution behavior data.

[0071] Optionally, in some embodiments, the task execution behavior includes: a first type of task execution behavior and a second type of task execution behavior, wherein the first type of task execution behavior is the task execution behavior automatically generated by the network optimization system, and the second type of task execution behavior is the task execution behavior generated by the network optimization system when it receives the task execution instruction of the target object, wherein according to the node level information, the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level is obtained, the first execution behavior data generated by the first type of task execution behavior at the current node level and the second execution behavior data generated by the second type of task execution behavior at the current node level are obtained, and the first execution behavior data and the second execution behavior data are collectively used as the execution behavior data.

[0072] Among them, the first type of task execution behavior refers to the task execution behavior automatically generated by the network optimization system when executing the node process execution logic.

[0073] Among them, the first execution behavior data refers to the trace data generated during the execution of the first type of task execution behavior.

[0074] Among them, the second type of task execution behavior refers to the task execution behavior generated when manipulating the network optimization system to receive the task execution instruction of the target object, that is, it is the task execution behavior generated after the network optimization system is manually operated.

[0075] The target object refers to the network optimization experts or other personnel who operate the network optimization system.

[0076] Among them, the second execution behavior data refers to the trace data generated during the execution of the second type of task execution behavior.

[0077] In the embodiments of the present disclosure, when obtaining the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level according to the node level information, the first execution behavior data generated by the first type of task execution behavior and the second execution behavior data generated by the second type of task execution behavior can be obtained respectively, and the trace data automatically generated at each node level when the network optimization system executes the node process execution logic can be obtained as the first execution behavior data, and the trace data automatically generated at each node level after the network optimization system is manually operated can be obtained as the second execution behavior data, and then the first execution behavior data and the second execution behavior data can be used together as the execution behavior data.

[0078] For example, in the network optimization system, two attributes are set for each first task node. The attribute node_task_type identifies which of the five task types (dimensions) in the process the first task node belongs to, which can be: perception, analysis, decision-making, execution, and evaluation. The node_dimension_type indicates which node level the dimension of the first task node belongs to, which can be: user layer, business layer, and network. When the process of the first network optimization task is running, when the first task node runs a specific task node, data traces of the task execution behavior will be left. Two methods of active traces and passive traces are designed to leave a full amount of relevant data for the entire life cycle of self-intelligent production, and through feature engineering, data standardization is established for use in the next stage of self-learning, and the efficiency and quality of self-learning are laid.

[0079] Among them, in the process of leaving traces of behavioral data, a three-layer and five-dimensional trace data model is designed accordingly. The three layers are user layer, business layer, and network layer. Each layer contains five dimensions, namely perception, analysis data and analysis behavior, decision-making, execution, and evaluation dimensions. When the process runs to the corresponding dimensions, that is, perception, analysis, decision-making, execution, and evaluation task types, all operations performed by experts on the system and operations performed automatically by the system will be recorded as corresponding trace data.

[0080] The constructed data model can be expressed as: Op_Mark_Data = {UserLayerData, BusinessLayerData, NetworkLayerData}, where Op_Mark_Data is the data set of the trace data model, which contains the following three dimensional data sets, user layer UserLayerData, business layer BusinessLayerData, and network layer NetworkLayerData. For these three dimensional data sets, each data dimension data set is composed of the following data, They are perception, analysis of data and analysis of behavior, decision-making, execution, and evaluation of five-dimensional data, such as Figure 4 As shown, Figure 4 It is a structural diagram of the three-layer five-dimensional data trace model in the embodiment of the present disclosure. Based on this trace data model, when each process on the network optimization system is running, whether it is an expert manual processing process, an automatic processing process, or a mixed processing process, the behavior and operation data generated will be mapped to the three-layer five-dimensional data model of Op_Mark_Data and Dimension_Data according to the aforementioned node attributes node_task_type and node_dimension_type when the behavior occurs. Among them, when each task is processed on the web page of the network optimization self-intelligent workbench of the network optimization system, when the process runs to each dimension, the business expert's full amount of behavior data on the web page of the network optimization system, including confirmation, correction, use of auxiliary tools, order and trajectory of using auxiliary tools, input and output operations of the use process, and analysis process of decision-making and active input, etc., when the problem and plan are formulated, will be recorded in the system background database and correspond to each dimension.

[0081] S304: Using the execution behavior data as initial task data.

[0082] In the embodiment of the present disclosure, after acquiring the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level according to the node level information, the execution behavior data can be used as the initial task data.

[0083] S305: Acquire an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data.

[0084] For the specific description of S305, please refer to the above embodiment, which will not be repeated here.

[0085] S306: Perform data format standardization processing on the initial task data to obtain target task data.

[0086] The target task data refers to the task data that can participate in model training after the initial task data is converted into a format table.

[0087] In the disclosed embodiment, when the data format of the initial task data is standardized to obtain the target task data, a feature engineering mechanism can be designed for the initial task data to convert the initial task data into standardized data. A three-step method is set in the feature engineering mechanism to construct a conversion method for constructing behavioral specification data with unified dimensions, including formatting, unification and normalization. 270 feature fields of common attributes are also preset to eliminate ambiguous data in the initial task data, thereby obtaining the target task data after conversion.

[0088] S307: Train the initial network optimization model according to the target task data and the historical rule data to obtain the target network optimization model.

[0089] In the disclosed embodiment, when the initial network optimization model is trained according to the target task data and the historical rule data to obtain the target network optimization model, the initial network optimization model can be trained and learned in combination with the historical rule data Al_knowledge and the target task data Op_Mark_Data. During the training process, the rules or schemes corresponding to the network optimization tasks are automatically generated. At the same time, the model self-learning process can be visualized, and the triggering and frequency of learning can be controlled. The input data sources of the model self-learning include two data sources: historical rule data and target task data. The target task data is the data that can be input into the model after the initial task data is processed by the feature mechanism. The historical rule data is the result of production transformation, which has been generated after meeting the feature engineering requirements and can be directly used as the learning data input. Among them, the existing knowledge Al_knowledge, It can be expressed as, Al_knowledge_Data = {RuleAttData, RuleThreshold, RuleContent} are respectively the field of the RuleAttData rule, the setting threshold of the RuleThreshold rule, and the content of the RuleContent rule. Then, self-learning model training can be initiated on the network optimization system according to the historical rule data and the target task data. A learning controller for model training can be designed in the network optimization system. The learning controller includes three parts: learning task management, learning process visualization, and learning result comparison. The learning task management completes multiple trainings to obtain multiple candidate network optimization models. The learning process visualization is used to control the model training process. The learning result comparison is used to determine the target network optimization model with the best performance from multiple candidate network optimization models.

[0090] Optionally, in some embodiments, when training the initial network optimization model according to the target task data and the historical rule data to obtain the target network optimization model, the historical rule data can be updated according to the target task data to obtain the processed rule data, the initial network optimization model can be trained according to the processed rule data to obtain the first candidate network optimization model, the initial network optimization model can be trained with the historical rule data and the target task data to obtain the second candidate network optimization model, and the target network optimization model can be determined from the first candidate network optimization model and the second candidate network optimization model.

[0091] The processed rule data refers to the data obtained after updating part of the data content in the historical rule data according to the target task data.

[0092] The first candidate network optimization model refers to a network optimization model obtained after training the initial network optimization model according to the processed rule data.

[0093] The second candidate network optimization model refers to a network optimization model obtained by training the initial network optimization model with the combined historical rule data and the target task data.

[0094] In the disclosed embodiment, when the initial network optimization model is trained according to the target task data and the historical rule data to obtain the target network optimization model, the historical rule data can be updated according to the target task data, and the task data can be used as the result label of the evaluation of the historical rule data, and then the parameters of the historical rule data can be optimized and updated after machine learning in combination with a supervised machine learning algorithm or an unsupervised machine learning algorithm to obtain processed rule data, and then the processed rule data can be input into the initial network optimization model for training to obtain a first candidate network optimization model, and the training can be performed multiple times to obtain multiple first candidate network optimization models, and the target task data and the historical rule data can be input into the initial network optimization model, and the initial network optimization model can be trained with the target task data and the historical rule data to obtain a second candidate network optimization model, and then the network optimization model with the best performance can be selected from the first candidate network optimization model and the second candidate network optimization model as the target network optimization model.

[0095] Optionally, in some embodiments, when determining a target network optimization model from a first candidate network optimization model and a second candidate network optimization model, a model test indicator can be obtained, wherein the model test indicator is used to test and process the model training results of the first candidate network optimization model and the second candidate network optimization model, the model training results of the first candidate network optimization model are tested and processed according to the model test indicator to obtain a first model indicator parameter, the model training results of the second candidate network optimization model are tested and processed according to the model test indicator to obtain a second model indicator parameter, and the target network optimization model is selected from the first candidate network optimization model and the second candidate network optimization model according to the first model indicator parameter and the second model indicator parameter.

[0096] The model testing index may be an index such as the loss value or accuracy of the model, or may be any other data index that can evaluate the performance of the model, without limitation.

[0097] Optionally, in some embodiments, when selecting a target network optimization model from a first candidate network optimization model and a second candidate network optimization model based on the first model indicator parameter and the second model indicator parameter, an indicator parameter threshold can be obtained, a target model indicator parameter whose value is closest to the indicator parameter threshold is determined from the first model indicator parameter and the second model indicator parameter, and a candidate network optimization model that matches the target model indicator parameter is selected from the first candidate network optimization model and the second candidate network optimization model as the target network optimization model.

[0098] Optionally, in some embodiments, the target task data includes: execution node behavior data, wherein the execution node behavior data is data generated when the first task node is a task execution node; the historical rule data includes: rule threshold data and rule content data, wherein the historical rule data is updated and processed according to the target task data to obtain processed rule data, and the rule threshold data and rule content data are updated and processed according to the execution node behavior data to obtain processed rule data.

[0099] In the disclosed embodiment, when the rule threshold data and the rule content data are updated and processed according to the execution node behavior data to obtain the processed rule data, the execution node behavior data evaluationData in the target rule data Op_Mark_Data can be combined as the evaluation result label for the historical rule data Al_knowledge_Data, and then the rule threshold data RuleThreshold and the rule content data in the historical rule data Al_knowledge_Data can be optimized and updated after machine learning by combining a supervised machine learning algorithm or an unsupervised machine learning algorithm. For example, after the learning is completed, the value of the rule threshold data RuleContent will be changed, or the rule content data RuleContent will be updated to see whether it continues to be valid, thereby obtaining the processed rule data.

[0100] For example, Figure 5 As shown, Figure 5 It is a schematic diagram of the model training and learning process in the embodiment of the present disclosure. The initial task data is converted into target task data that can be directly input into the initial network optimization model through feature engineering. The historical rule data is updated and processed according to the target task data to obtain processed rule data. The processed rule data, target task data and historical rule data jointly participate in the training process of the initial network optimization model. The entire learning process of training the initial network optimization model for the processed rule data follows the standard machine learning training process. The algorithm used may include a collection of publicly available open source algorithms or a self-developed algorithm, and there is no restriction on this. The learning and training method also follows the training method of the algorithm collection used. The result of the learning is new knowledge Al_knowledge_Vn, where Vn represents the version of knowledge. Each time it is generated, The data will be incremented to obtain multiple first candidate network optimization models. The initial network optimization model is trained for the joint target task data and historical rule data to generate a learning process for new rules. All data in Op_Mark_Data and data in Al_knowledge_Data are used for training and learning. The entire learning process follows the standard machine learning training process. The algorithms used may include a collection of publicly available open source algorithms or self-developed algorithms, without any restrictions. The learning and training method also follows the training method of the algorithm collection used. The result of the learning is the new knowledge Al_knowledge_Create_Vn, where create_Vn represents the nth version of the new knowledge. Each time it is generated, the data will be incremented to obtain multiple second candidate network optimization models.

[0101] S308: Execute a second network optimization task in the network optimization system according to the target network optimization model.

[0102] For the specific description of S308, please refer to the above embodiment, which will not be repeated here.

[0103] In this embodiment, by obtaining the node type information and node level information of the first task node, wherein the first task node is the task node of the first network optimization task currently executed by the network optimization system, and the node level information indicates the node level of the first task node in the first network optimization task, initial task data is obtained according to the node type information and the node level information, and an initial network optimization model is obtained, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data indicates data related to the network optimization rules generated by the initial network optimization model, and the initial network optimization model is trained according to the initial task data and the historical rule data to obtain a target network optimization model, and according to the target network optimization model, a second network optimization task in the network optimization system is executed, wherein the second network optimization task is a newly added network optimization task in the network optimization system, and the data trace of the whole process of multiple task nodes of network optimization can be recorded, so as to realize the data trace of the thinking and analysis process of network optimization, effectively improve the low trace rate of operation behavior, avoid the loss of effective information, effectively improve the formation rate of reusable knowledge, and ensure the execution effect and efficiency of the network optimization tasks of the network optimization system.

[0104] Figure 6 It is a flowchart of a network optimization task execution method proposed in an embodiment of the present disclosure.

[0105] like Figure 6 As shown, the network optimization task execution method includes:

[0106] S601: Obtain a model business effectiveness index and a model production effectiveness index of a target network optimization model, wherein the model business effectiveness index is used to describe the model performance of the target network optimization model, and the model production effectiveness index is used to describe the production efficiency of the target network optimization model.

[0107] Among them, the model business effectiveness index refers to the data index used to describe the model performance of the target network optimization model.

[0108] Among them, the model production effectiveness index refers to the data indicator used to describe the production efficiency of the target network optimization model.

[0109] In the disclosed embodiment, after the target network optimization model is obtained through training, the target network optimization model will enter production to provide model services. A mechanism for putting the model into production and evaluating the model production effectiveness can be designed, which are respectively a business effectiveness controller and a production participation effectiveness controller. The business effectiveness controller has corresponding model business effectiveness indicators, and the production participation effectiveness controller has corresponding model production effectiveness indicators. In the business effectiveness controller, the effectiveness parameters of each task / work order processing will be dynamically counted to obtain the ki indicator set (model business effectiveness indicators). The default model business effectiveness indicators include: completion effectiveness of tasks / work orders, completion time of tasks / work orders, completion path of tasks / work orders, completion automation rate of tasks / work orders, completion self-intelligence level of tasks / work orders, and trajectory distribution deviation of tasks / work orders. Model business effectiveness indicators can be dynamically added, that is, model business effectiveness indicators can be dynamically expanded and added on this basis.

[0110] In the disclosed embodiment, in the production participation effectiveness controller, a Te indicator set (model production effectiveness indicator) is preset. The model production effectiveness indicator includes: model historical participation effectiveness, offline knowledge update rate, new knowledge generation rate, model online service time, model online service call volume and model online service output deviation index. Correspondingly, the model production effectiveness index can be dynamically added and expanded. For each task / work order processing, the ki indicator will be updated through the business effectiveness controller, and the Te indicator will be updated through the production participation effectiveness controller. The ki indicator and the Te indicator are time-related. When the process is designed for the first time, the ki indicator and the Te indicator corresponding to the process are empty. When the process runs for the first time, the indicators in the two indicator sets will be updated to form an indicator set related to the process running time, such as the completion time of the task / work order in the ki indicator. After the process runs, it will calculate the value of the process end time-process start time. It should be noted that the indicators in the indicator set can be dynamically expanded, and the calculation method of the indicator is also specified during the expansion and provided in the form of executable code during implementation.

[0111] S602: Determine a first indicator parameter and a second indicator parameter, wherein the first indicator parameter and the second indicator parameter are used to participate in determining whether to execute a second network optimization task.

[0112] The first indicator parameter and the second indicator parameter are used to participate in determining whether to execute the second network optimization task, and the values ​​of the first indicator parameter and the second indicator parameter can be adaptively set according to the actual application scenario.

[0113] In the disclosed embodiment, the first indicator parameter α and the second indicator parameter β can be determined as open parameters, that is, they can be set manually. The update rate of the model can be controlled by the first indicator parameter α and the second indicator parameter β, that is, they participate in determining whether to execute the second network optimization task.

[0114] S603: Determine the model effectiveness parameter of the target network optimization model according to the model business effectiveness indicator, the model production effectiveness indicator, the first indicator parameter, and the second indicator parameter.

[0115] The model effectiveness parameters refer to data parameters that can be used to determine the online and offline status of the target network optimization model in the network optimization system.

[0116] In the embodiment of the present disclosure, the model business effectiveness index and the model production effectiveness index of the target network optimization model are obtained, and the first indicator parameter and the second indicator parameter are determined. After that, the upper and lower lines of the target network optimization model can be collaboratively controlled according to the model business effectiveness index ki, the model production effectiveness index Te, the first indicator parameter α and the second indicator parameter β. The calculation expression of the model effectiveness parameter can be introduced. Substitute the model business effectiveness index ki, the model production effectiveness index Te, the first indicator parameter α and the second indicator parameter β into the expression for calculation and processing to obtain the corresponding model effectiveness parameters, that is, through the temporal correlation of the ki index and the Te index, calculate their change rates in business effectiveness and production participation effectiveness respectively, and use the two to obtain the kite' factor of effectiveness control. Finally, use the switch function for the kite' factor to obtain the calculation result f(kite') of the upper and lower lines of the control model as the model effectiveness parameter. For example, if Figure 7 As shown, Figure 7 It is a schematic diagram of model effectiveness control in an embodiment of the present disclosure.

[0117] S604: Determine working status information of the target network optimization model according to the model effectiveness parameter, wherein the working status information is used to indicate whether the target network optimization model performs the second network optimization task.

[0118] The working status information is used to indicate whether the target network optimization model performs the second network optimization task, that is, the working status information is used to indicate whether the target network optimization model continues to remain online in the network optimization system.

[0119] In the disclosed embodiment, when determining the working status information of the target network optimization model according to the model effectiveness parameter, the parameter threshold and the model effectiveness parameter may be set to determine the working status information of the target network optimization model.

[0120] Optionally, in some embodiments, the working status information of the target network optimization model is determined based on the model effectiveness parameter, and the effectiveness parameter threshold can be obtained. If the model effectiveness parameter is greater than or equal to the effectiveness parameter threshold, the working status information is determined to be that the target network optimization model performs the second network optimization task; if the model effectiveness parameter is less than the effectiveness parameter threshold, the working status information is determined to be that the target network optimization model does not perform the second network optimization task.

[0121] The effectiveness parameter threshold refers to a parameter threshold value pre-set for the model effectiveness parameter, and the effectiveness parameter threshold can be set to 0.5.

[0122] In the disclosed embodiment, the working status information of the target network optimization model is determined according to the model effectiveness parameter, and the effectiveness parameter threshold can be obtained, and then the effectiveness parameter threshold and the model effectiveness parameter can be numerically compared. If the model effectiveness parameter is greater than or equal to the effectiveness parameter threshold, that is, when f(kite')≥0.5, the model can be put online, and the working status information is determined to be that the target network optimization model performs the second network optimization task. If the model effectiveness parameter is less than the effectiveness parameter threshold, that is, when f(kite')<0.5, the working status information is determined to be that the target network optimization model does not perform the second network optimization task, and the model can be temporarily offline.

[0123] In this embodiment, through two sets of indicators related to time (model business effectiveness indicators and model production effectiveness indicators), the model can be quickly switched online and offline in the time dimension. At the same time, because the two sets of indicators used include business effectiveness and participation effectiveness, the control of the model will be flexible and can be dynamically and flexibly adjusted as the indicators expand.

[0124] Optionally, in some embodiments, after determining that the working status information indicates that the target network optimization model does not perform the second network optimization task, a model training trigger message is generated, wherein the model training trigger message is used to trigger training processing of the target network optimization model to optimize the model performance of the target network optimization model, and perform the second network optimization task based on the optimized target network optimization model.

[0125] In the embodiment of the present disclosure, after determining that the working status information indicates that the target network optimization model does not execute the second network optimization task, a model training trigger message can be generated in the network optimization system to trigger the training of the target network optimization model to optimize the model performance of the target network optimization model, that is, when the model effectiveness guarantee f(kite') is in an effective state (less than 0.5), a message will be sent to the learning task trigger in the trace learning stage to trigger the model learning task to obtain the updated target network optimization model, and the second network optimization task will be executed based on the optimized target network optimization model.

[0126] In the implementation of the present disclosure, the scheme described in the model production participation control stage uses the result of f(kite') to evolve and update the model and rules. Because it is time-related and also related to the number of times the task / work order is processed, it can be controlled by the open α and β parameters. Therefore, the speed of evolution and update can be controlled more flexibly. At the same time, the Te indicator set includes two basic indicators: the offline knowledge update rate and the new knowledge generation rate, and corresponds to the knowledge data of the behavior trace stage. Therefore, new knowledge can be continuously generated and tested through learning from the data generated in the tasks that are constantly carried out under the self-intelligence production mode.

[0127] S605: Determine a first candidate network optimization rule corresponding to the second network optimization task according to the target network optimization model.

[0128] The first candidate network optimization rule refers to a network optimization rule automatically determined according to a target network optimization model.

[0129] In the implementation of the present disclosure, when executing the second network optimization task according to the target network optimization model, the target network optimization model can be used to determine the first candidate network optimization rule corresponding to the second network optimization task, and the problem description of the second network optimization task described in the work order corresponding to the second network optimization task can be input into the target network optimization model to obtain the network optimization rule output by the target network optimization model as the first candidate network optimization rule.

[0130] S606: Determine a second candidate network optimization rule corresponding to the second network optimization task according to the historical rule data.

[0131] The second candidate network optimization rule refers to a network optimization rule corresponding to the second network optimization task that is formulated by experts based on historical rule data and historical experience.

[0132] In the disclosed embodiment, when determining the second candidate network optimization rule corresponding to the second network optimization task based on the historical rule data, experts can formulate a network optimization rule for solving the second network optimization task based on the historical rule data and historical experience, and use the network optimization rule as the second candidate network optimization rule.

[0133] S607: Determine a target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule.

[0134] In the embodiment of the present disclosure, after determining the first candidate network optimization rule corresponding to the second network optimization task according to the target network optimization model and determining the second candidate network optimization rule corresponding to the second network optimization task according to the historical rule data, the target network optimization rule can be determined from the first candidate network optimization rule and the second candidate network optimization rule.

[0135] Optionally, in some embodiments, a target network optimization rule is determined from a first candidate network optimization rule and a second candidate network optimization rule, and a rule weight coefficient of the first candidate network optimization rule can be determined based on a model business effectiveness indicator and a model production effectiveness indicator. If the rule weight coefficient is greater than a weight coefficient threshold, the first candidate network optimization rule is used as the target network optimization rule; if the rule weight coefficient is less than or equal to the weight coefficient threshold, the second candidate network optimization rule is used as the target network optimization rule.

[0136] In the embodiment of the present disclosure, when determining the target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule, a rule decision model can be established according to the model business effectiveness index and the model production effectiveness index. The design method of rule execution decision is as follows: obtain the model business effectiveness index ki and the model production effectiveness index Te in the description of the model production participation control stage, and update the historical ki index and Te index based on the time trend of the update rate kite t , to determine whether to reverse the scheme selection. When the scheme selection is confirmed to be reversed, the second candidate network optimization rule is determined as the target network optimization rule. When it is determined that the scheme selection is not reversed, the first candidate network optimization rule is determined as the target network lake rule. The method for confirming whether to reverse is as follows: Establish the rule weight coefficient calculation expression And set the weight coefficient threshold to 0, and then perform numerical comparison on the rule weight coefficient and the weight coefficient threshold. If the rule weight coefficient is greater than the weight coefficient threshold, the first candidate network optimization rule is used as the target network optimization rule. If the rule weight coefficient is less than or equal to the weight coefficient threshold, the second candidate network optimization rule is used as the target network optimization rule. When the scheme inversion control function f(kite t ) is greater than 0, the scheme selection will be reversed, where δ is the reversal control rate, which can be set manually to control the strength of the reversal. When the scheme is determined to be reversed, the second candidate network optimization rule is determined as the target network rule. When the scheme is determined not to be reversed, the first candidate network optimization rule is determined as the target network rule.

[0137] For example, Figure 8 As shown, Figure 81 is a schematic diagram of the network optimization collaborative execution process in the embodiment of the present disclosure. In the process, after the second network optimization task problem is identified, the expert solution and the Turing solution are used to formulate the solution for the identified problem. When the process is instantiated for the first time, because no trace learning has been generated, the expert solution expert_plan and the Turing solution Turing_plan are the same, and no solution is selected at this time. After the first instantiation of the process is completed, trace data is generated, and trace learning is performed. After the Turing solution is updated, the expert solution and the Turing solution will be selected during the second and subsequent instantiations of the process. The expert solution expert_plan uses the existing knowledge Al _knowledge is used to formulate a plan, while the Turing plan Turing_plan is Al_knowledge_Vn or Al_knowledge_Create_Vn after trace learning, that is, the nth version of the existing knowledge update or the nth version of the new knowledge. The plan formulated by Turing is not always more effective than the plan formulated by the expert. Therefore, when using the existing knowledge Al_knowledge and the nth version of Al_knowledge_Vn or Al_knowledge_Create_Vn generated based on trace learning to formulate and generate a plan, it is necessary to select a plan, that is, to make a selection through the "plan execution decision" described above.

[0138] S608: Execute a second network optimization task according to the target network optimization rule.

[0139] In the embodiment of the present disclosure, after determining the target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule, the second network optimization task can be executed according to the target network optimization rule.

[0140] Optionally, in some embodiments, after executing the second network optimization task according to the target network optimization rule, the execution result of the target network optimization rule can be obtained. If the execution result indicates that the target network optimization rule is not executed successfully, the target network optimization rule is updated according to the first candidate network optimization rule and the second candidate network optimization rule, and the second network optimization task is executed based on the updated target network optimization rule.

[0141] That is to say, when the expert solution or the Turing solution fails during execution, another solution will be executed immediately. When both the expert solution and the Turing solution of the self-intelligent model fail, manual analysis and execution will be performed. This can ensure the effectiveness of production and allow the machine learning model to participate in online production when its effectiveness is low, but its low effectiveness will not affect the overall production effectiveness.

[0142] For example, Fig. 9 As shown, Fig. 9 It is a flowchart of the intelligent evolution scheme of the network optimization system in the embodiment of the present disclosure. The "three-layer five-dimensional trace data model" in the behavior trace establishes the standardization of trace data, provides a data foundation for self-learning and iterative learning evolution in the subsequent stages, and lays the foundation for the efficiency and quality of self-learning. The two data sources in the trace learning stage, namely the trace data and the existing knowledge data source, enable the trained model to be re-learned and updated based on the existing knowledge. At the same time, among the triggering conditions of the learning task in this stage, one of the triggering conditions is to receive a message from the model production participation control stage to initiate the learning task, which enables the learning to be self-intelligent and autonomous. Instead of manual observation in the past, manual decision-making to initiate learning, or only simple incremental learning in time. In the model production participation control stage, the mechanism of model production and model production effectiveness evaluation is the basis for the model to be dynamic and flexible and self-intelligent and autonomous production and participate in regulation, and it is also the source of messages for the model to carry out evolutionary update control. At the same time, in the self-intelligent production mode, the way of program execution decision-making is based on the data and programs in the model production participation control stage, making the selection of programs more reasonable and more autonomous, and the "collaborative execution model" at this stage allows the execution of programs to ensure effectiveness while making self-intelligent execution possible. Finally, in the evolution update control stage, because it is time-related and also related to the number of tasks / work orders processed, and can be controlled through the open α and β parameters, the speed of evolution update can be controlled more flexibly.

[0143] In the disclosed embodiment, the "three-layer five-dimensional trace data model" adopted is more complete than the existing trace methods. While leaving traces of conclusions, the thinking and analysis behaviors of the process are also retained, and this process is more standardized. In the related technology, there is no complete and standard standard data source in the trace stage as in the disclosed embodiment of this scheme, which requires heavy data cleaning. At the same time, because it is not systematic and has little integration with existing knowledge, it is difficult to learn very useful knowledge, resulting in the offline test of the model barely meeting the online standards, but quickly failing after going online.

[0144] In the implementation of the present disclosure, the triggering method of the learning task of the network optimization model is more self-intelligent and autonomous. Compared with the traditional method of simple manual or indifferent automatic triggering, the online availability of the model learning is higher and more reasonable. It is completely different from the traditional model online and offline control method. In the traditional method, the model is put online immediately after training is completed, and the gray release and A / B test methods are generally adopted to control the model to participate in production. However, although this method can smoothly transition the model of the online service to the model with good online service performance and ensure the stability of the overall system, this method is only suitable for scenarios that will not cause fatal impact on the business, that is, even if the worst solutions are given in the gray release, the effectiveness of the business will not be irreversibly affected. For example, the product recommendation of the shopping website, if the product recommended by the online model makes the user feel uninterested or not highly relevant in the local time range, it will not greatly affect the user's willingness to finally buy the product he wants to buy, and will not immediately directly bring fatal revenue impact to the business of the shopping website. At the same time, there is also more time to slowly optimize the model offline, that is, in this scenario, the effectiveness of the model will not have more negative impact on the business. The situation is completely opposite for network optimization services. If the effectiveness of the solution generated by the online model service is actually very low, it will affect the execution closed-loop cost of the work order using the model solution in a short period of time. If the low effectiveness is maintained for a long time without intervention, it will cause a fatal blow to the network quality and quickly affect the operator's network operation and business revenue.

[0145] In the embodiments of the present disclosure, compared with the traditional methods of relying solely on manual labor, or relying solely on expert rules, or using only AI capabilities in network optimization production, which have high costs, long time, and other problems such as improving efficiency, effectiveness, and performance, there is a production method that simultaneously uses manual labor, expert rules, and machine learning models to collaborate, and the method of executing decisions through plans can ensure the effectiveness of production while allowing the machine learning model to participate in online production when its effectiveness is low, but will not affect the overall production effectiveness due to its low effectiveness.

[0146] In the disclosed embodiment, the evolutionary update is combined with the trace learning stage, and one of the triggering conditions is to receive a message from the model production participation control stage to initiate the learning task. This enables the learning to be self-intelligent and complete the learning autonomously, rather than the previous manual observation and manual decision-making to initiate learning, or only simple and indifferent incremental learning in time, to ensure the training effect of the network optimization model.

[0147] Fig.10 It is a structural diagram of a network optimization task execution device proposed in an embodiment of the present disclosure.

[0148] like Fig.10 As shown, the network optimization task execution device 100 includes:

[0149] A first acquisition module 1001 is used to acquire node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by the network optimization system, and the node level information indicates the node level of the first task node in the first network optimization task;

[0150] The second acquisition module 1002 is used to acquire initial task data according to the node type information and the node level information;

[0151] A third acquisition module 1003 is used to acquire an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data represents data related to the network optimization rules generated by the initial network optimization model;

[0152] A training module 1004 is used to train an initial network optimization model according to initial task data and historical rule data to obtain a target network optimization model; and

[0153] The processing module 1005 is used to execute a second network optimization task in the network optimization system according to the target network optimization model, wherein the second network optimization task is a newly added network optimization task in the network optimization system.

[0154] In some embodiments of the present disclosure, the second acquisition module 1002 is specifically configured to:

[0155] Determine the node process execution logic corresponding to the first task node according to the node type information;

[0156] According to the node level information, the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level is obtained;

[0157] The execution behavior data is used as the initial task data.

[0158] In some embodiments of the present disclosure, the task execution behavior includes: a first type of task execution behavior and a second type of task execution behavior, the first type of task execution behavior is a task execution behavior automatically generated by the network optimization system, and the second type of task execution behavior is a task execution behavior generated by the network optimization system receiving a task execution instruction of a target object;

[0159] The second acquisition module 1002 is further used for:

[0160] Acquire first execution behavior data generated by the first type of task execution behavior at the current node level and second execution behavior data generated by the second type of task execution behavior at the current node level;

[0161] The first execution behavior data and the second execution behavior data are collectively regarded as execution behavior data.

[0162] In some embodiments of the present disclosure, the training module 1004 is specifically configured to:

[0163] Get the preset model training times and model training conditions;

[0164] Determine whether the initial network optimization model meets the model training conditions;

[0165] When the initial network optimization model meets the model training conditions and the training times of the initial network optimization model does not reach the preset model training times, the initial network optimization model is trained according to the initial task data and the historical rule data to obtain the target network optimization model.

[0166] In some embodiments of the present disclosure, the model training condition includes at least any one of the following:

[0167] A model training trigger message has been generated in the network optimization system;

[0168] The number of first network optimization tasks completed by the initial network optimization model is greater than a task number threshold.

[0169] In some embodiments of the present disclosure, the training module 1004 is further configured to:

[0170] Perform data format standardization on the initial task data to obtain the target task data;

[0171] The initial network optimization model is trained according to the target task data and historical rule data to obtain the target network optimization model.

[0172] In some embodiments of the present disclosure, the training module 1004 is further configured to:

[0173] The historical rule data is updated according to the target task data to obtain the processed rule data;

[0174] Training the initial network optimization model according to the processed rule data to obtain a first candidate network optimization model;

[0175] The initial network optimization model is trained by combining historical rule data and target task data to obtain a second candidate network optimization model;

[0176] A target network optimization model is determined from the first candidate network optimization model and the second candidate network optimization model.

[0177] In some embodiments of the present disclosure, the training module 1004 is further configured to:

[0178] Obtaining a model test indicator, wherein the model test indicator is used to test the model training results of the first candidate network optimization model and the second candidate network optimization model;

[0179] Performing test processing on the model training result of the first candidate network optimization model according to the model test index to obtain a first model index parameter;

[0180] Performing test processing on the model training result of the second candidate network optimization model according to the model test index to obtain a second model index parameter;

[0181] A target network optimization model is selected from the first candidate network optimization model and the second candidate network optimization model according to the first model index parameter and the second model index parameter.

[0182] In some embodiments of the present disclosure, the training module 1004 is further configured to:

[0183] Get indicator parameter thresholds;

[0184] Determine a target model index parameter whose value is closest to the index parameter threshold value from the first model index parameter and the second model index parameter;

[0185] A candidate network optimization model that matches the target model indicator parameters is selected from the first candidate network optimization model and the second candidate network optimization model as the target network optimization model.

[0186] In some embodiments of the present disclosure, the target task data includes: execution node behavior data, wherein the execution node behavior data is data generated when the first task node is a task execution node; the historical rule data includes: rule threshold data and rule content data;

[0187] The training module 1004 is also used for:

[0188] The rule threshold data and the rule content data are updated and processed according to the execution node behavior data to obtain processed rule data.

[0189] In some embodiments of the present disclosure, Fig.11 As shown, Fig.11 is a schematic diagram of the structure of a network optimization task execution device proposed in another embodiment of the present disclosure, and further includes:

[0190] A fourth acquisition module 1006 is used to acquire a model business effectiveness index and a model production effectiveness index of the target network optimization model before executing the second network optimization task in the network optimization system according to the target network optimization model, wherein the model business effectiveness index is used to describe the model performance of the target network optimization model, and the model production effectiveness index is used to describe the production efficiency of the target network optimization model;

[0191] A first determination module 1007, used to determine a first indicator parameter and a second indicator parameter, wherein the first indicator parameter and the second indicator parameter are used to participate in determining whether to perform a second network optimization task;

[0192] A second determination module 1008, configured to determine a model effectiveness parameter of the target network optimization model according to the model business effectiveness index, the model production effectiveness index, the first indicator parameter, and the second indicator parameter;

[0193] The third determination module 1009 is used to determine the working state information of the target network optimization model according to the model effectiveness parameter, wherein the working state information is used to indicate whether the target network optimization model performs the second network optimization task.

[0194] In some embodiments of the present disclosure, the third determining module 1009 is specifically configured to:

[0195] Get the performance parameter threshold;

[0196] If the model effectiveness parameter is greater than or equal to the effectiveness parameter threshold, determining the working state information to execute the second network optimization task for the target network optimization model;

[0197] If the model effectiveness parameter is less than the effectiveness parameter threshold, it is determined that the working state information is that the target network optimization model does not perform the second network optimization task.

[0198] In some embodiments of the present disclosure, the processing module 1005 is specifically configured to:

[0199] Determining, according to the target network optimization model, a first candidate network optimization rule corresponding to the second network optimization task;

[0200] Determining, based on the historical rule data, a second candidate network optimization rule corresponding to the second network optimization task;

[0201] Determining a target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule;

[0202] According to the target network optimization rule, a second network optimization task is performed.

[0203] In some embodiments of the present disclosure, the processing module 1005 is further configured to:

[0204] Determine the rule weight coefficient of the first candidate network optimization rule according to the model business effectiveness index and the model production effectiveness index;

[0205] If the rule weight coefficient is greater than the weight coefficient threshold, the first candidate network optimization rule is used as the target network optimization rule;

[0206] If the rule weight coefficient is less than or equal to the weight coefficient threshold, the second candidate network optimization rule is used as the target network optimization rule.

[0207] In some embodiments of the present disclosure, the processing module 1005 is further configured to:

[0208] After executing the second network optimization task according to the target network optimization rule, obtaining the execution result of the target network optimization rule;

[0209] If the execution result indicates that the target network optimization rule is not executed successfully, updating the target network optimization rule according to the first candidate network optimization rule and the second candidate network optimization rule;

[0210] Based on the updated target network optimization rule, a second network optimization task is executed.

[0211] In some embodiments of the present disclosure, the third determining module 1009 is further configured to:

[0212] After determining that the working status information indicates that the target network optimization model does not perform the second network optimization task, generating a model training trigger message, wherein the model training trigger message is used to trigger training processing of the target network optimization model to optimize the model performance of the target network optimization model;

[0213] The second network optimization task is performed based on the target network optimization model after the optimization process.

[0214] In this embodiment, by obtaining the node type information and node level information of the first task node, wherein the first task node is the task node of the first network optimization task currently executed by the network optimization system, and the node level information indicates the node level of the first task node in the first network optimization task, initial task data is obtained according to the node type information and the node level information, and an initial network optimization model is obtained, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data indicates data related to the network optimization rules generated by the initial network optimization model, and the initial network optimization model is trained according to the initial task data and the historical rule data to obtain a target network optimization model, and according to the target network optimization model, a second network optimization task in the network optimization system is executed, wherein the second network optimization task is a newly added network optimization task in the network optimization system, and the data trace of the whole process of multiple task nodes of network optimization can be recorded, so as to realize the data trace of the thinking and analysis process of network optimization, effectively improve the low trace rate of operation behavior, avoid the loss of effective information, effectively improve the formation rate of reusable knowledge, and ensure the execution effect and efficiency of the network optimization tasks of the network optimization system.

[0215] Fig.12 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present disclosure is shown.

[0216] Fig.12 The electronic device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0217] like Fig.12 As shown, the electronic device 12 is in the form of a general purpose computing device. The components of the electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a memory 28, and a bus 18 that connects various system components (including the memory 28 and the processing unit 16).

[0218] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0219] The electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0220] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Fig.12 Not shown, often called a "hard drive").

[0221] although Fig.12 Not shown in the figure, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (hereinafter referred to as: CD-ROM), a digital versatile disc read only memory (hereinafter referred to as: DVD-ROM) or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present disclosure.

[0222] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described in the present disclosure.

[0223] The electronic device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), may also communicate with one or more devices that enable the human body to interact with the electronic device 12, and / or communicate with any device that enables the electronic device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). Such communication may be performed through an input / output (I / O) interface 22. In addition, the electronic device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 20. As shown, the network adapter 20 communicates with other modules of the electronic device 12 through a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0224] The processing unit 16 executes various functional applications and parameter information determination by running the program stored in the memory 28, such as implementing the network optimization task execution method mentioned in the above embodiment.

[0225] It should be noted that, in the description of the present disclosure, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present disclosure, unless otherwise specified, the meaning of "plurality" is two or more.

[0226] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present disclosure includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present disclosure belong.

[0227] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0228] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0229] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0230] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0231] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0232] Although the embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present disclosure. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present disclosure.

Claims

1. A network optimization task execution method, characterized in that: The method comprises: Acquire node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by the network optimization system, and the node level information indicates a node level of the first task node in the first network optimization task; Acquire initial task data according to the node type information and the node level information; Acquire an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data represents data related to the network optimization rules generated by the initial network optimization model; Training the initial network optimization model according to the initial task data and the historical rule data to obtain a target network optimization model; and According to the target network optimization model, a second network optimization task in the network optimization system is executed, wherein the second network optimization task is a newly added network optimization task in the network optimization system.

2. The method according to claim 1, characterized in that The acquiring of initial task data according to the node type information and the node level information includes: Determine, according to the node type information, a node process execution logic corresponding to the first task node; According to the node level information, acquiring the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level; The execution behavior data is used as the initial task data.

3. The method according to claim 2, characterized in that The task execution behavior includes: a first type of task execution behavior and a second type of task execution behavior, wherein the first type of task execution behavior is a task execution behavior automatically generated by the network optimization system, and the second type of task execution behavior is a task execution behavior generated by the network optimization system receiving a task execution instruction of a target object; Wherein, acquiring the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level according to the node level information includes: Acquire first execution behavior data generated by the first-type task execution behavior at the current node level and second execution behavior data generated by the second-type task execution behavior at the current node level; The first execution behavior data and the second execution behavior data are used together as the execution behavior data.

4. The method according to claim 1, characterized in that The training of the initial network optimization model according to the initial task data and the historical rule data to obtain a target network optimization model includes: Get the preset model training times and model training conditions; Determining whether the initial network optimization model meets the model training conditions; When the initial network optimization model meets the model training condition and the training times of the initial network optimization model does not reach the preset model training times, the initial network optimization model is trained according to the initial task data and the historical rule data to obtain the target network optimization model.

5. The method according to claim 4, characterized in that The model training conditions include at least one of the following: A model training trigger message has been generated in the network optimization system; The number of the first network optimization tasks completed by the initial network optimization model is greater than a task number threshold.

6. The method according to claim 4, characterized in that The training of the initial network optimization model according to the initial task data and the historical rule data to obtain the target network optimization model includes: Performing data format standardization processing on the initial task data to obtain target task data; The initial network optimization model is trained according to the target task data and the historical rule data to obtain the target network optimization model.

7. The method according to claim 6, characterized in that The training of the initial network optimization model according to the target task data and the historical rule data to obtain the target network optimization model includes: Update the historical rule data according to the target task data to obtain processed rule data; Training the initial network optimization model according to the processed rule data to obtain a first candidate network optimization model; Combine the historical rule data and the target task data to train the initial network optimization model to obtain a second candidate network optimization model; The target network optimization model is determined from the first candidate network optimization model and the second candidate network optimization model.

8. The method according to claim 7, characterized in that The determining the target network optimization model from the first candidate network optimization model and the second candidate network optimization model includes: Acquire a model test indicator, wherein the model test indicator is used to test the model training results of the first candidate network optimization model and the second candidate network optimization model; Performing test processing on the model training result of the first candidate network optimization model according to the model test index to obtain a first model index parameter; Performing a test process on the model training result of the second candidate network optimization model according to the model test index to obtain a second model index parameter; The target network optimization model is selected from the first candidate network optimization model and the second candidate network optimization model according to the first model indicator parameter and the second model indicator parameter.

9. The method according to claim 8, characterized in that The selecting the target network optimization model from the first candidate network optimization model and the second candidate network optimization model according to the first model indicator parameter and the second model indicator parameter includes: Get indicator parameter thresholds; Determine a target model index parameter whose value is closest to the index parameter threshold value from the first model index parameter and the second model index parameter; A candidate network optimization model that matches the target model indicator parameters is selected from the first candidate network optimization model and the second candidate network optimization model as the target network optimization model.

10. The method according to claim 7, characterized in that The target task data includes: execution node behavior data, wherein the execution node behavior data is data generated when the first task node is a task execution node; the historical rule data includes: rule threshold data and rule content data; The updating process of the historical rule data according to the target task data to obtain the processed rule data includes: The rule threshold data and the rule content data are updated according to the execution node behavior data to obtain the processed rule data.

11. The method according to claim 1, characterized in that Before executing the second network optimization task in the network optimization system according to the target network optimization model, the method further includes: Acquire a model business effectiveness index and a model production effectiveness index of the target network optimization model, wherein the model business effectiveness index is used to describe the model performance of the target network optimization model, and the model production effectiveness index is used to describe the production efficiency of the target network optimization model; Determine a first indicator parameter and a second indicator parameter, wherein the first indicator parameter and the second indicator parameter are used to participate in determining whether to perform the second network optimization task; Determining a model effectiveness parameter of the target network optimization model according to the model business effectiveness indicator, the model production effectiveness indicator, the first indicator parameter, and the second indicator parameter; According to the model effectiveness parameter, working status information of the target network optimization model is determined, wherein the working status information is used to indicate whether the target network optimization model performs the second network optimization task.

12. The method according to claim 11, characterized in that Determining the working status information of the target network optimization model according to the model effectiveness parameter includes: Get the performance parameter threshold; If the model effectiveness parameter is greater than or equal to the effectiveness parameter threshold, determining that the working status information is for the target network optimization model to perform the second network optimization task; If the model effectiveness parameter is less than the effectiveness parameter threshold, it is determined that the working status information indicates that the target network optimization model does not perform the second network optimization task.

13. The method according to claim 1, characterized in that The performing a second network optimization task in the network optimization system according to the target network optimization model includes: Determining, according to the target network optimization model, a first candidate network optimization rule corresponding to the second network optimization task; Determining, according to the historical rule data, a second candidate network optimization rule corresponding to the second network optimization task; Determining the target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule; The second network optimization task is performed according to the target network optimization rule.

14. The method according to claim 13, characterized in that The determining the target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule includes: Determining a rule weight coefficient of the first candidate network optimization rule according to the model business effectiveness indicator and the model production effectiveness indicator; If the rule weight coefficient is greater than the weight coefficient threshold, taking the first candidate network optimization rule as the target network optimization rule; If the rule weight coefficient is less than or equal to the weight coefficient threshold, the second candidate network optimization rule is used as the target network optimization rule.

15. The method according to claim 13, characterized in that After performing the second network optimization task according to the target network optimization rule, the method further includes: Obtaining the execution result of the target network optimization rule; If the execution result indicates that the target network optimization rule is not executed successfully, updating the target network optimization rule according to the first candidate network optimization rule and the second candidate network optimization rule; Based on the updated target network optimization rule, the second network optimization task is executed.

16. The method according to claim 12, characterized in that After determining that the working status information indicates that the target network optimization model does not perform the second network optimization task, the method further includes: Generate a model training trigger message, wherein the model training trigger message is used to trigger training processing of the target network optimization model to optimize the model performance of the target network optimization model; The second network optimization task is performed based on the optimized target network optimization model.

17. A network optimization task execution device, characterized in that: The device comprises: A first acquisition module, configured to acquire node type information and node level information of a first task node, wherein the first task node is a task node of a first network optimization task currently executed by the network optimization system, and the node level information indicates a node level of the first task node in the first network optimization task; A second acquisition module, used for acquiring initial task data according to the node type information and the node level information; A third acquisition module is used to acquire an initial network optimization model, wherein the initial network optimization model has corresponding historical rule data, and the historical rule data represents data related to the network optimization rules generated by the initial network optimization model; a training module, configured to train the initial network optimization model according to the initial task data and the historical rule data to obtain a target network optimization model; and A processing module is used to execute a second network optimization task in the network optimization system according to the target network optimization model, wherein the second network optimization task is a newly added network optimization task in the network optimization system.

18. The device according to claim 17, characterized in that The second acquisition module is specifically used for: Determine, according to the node type information, a node process execution logic corresponding to the first task node; According to the node level information, acquiring the execution behavior data generated by the task execution behavior in the node process execution logic at the current node level; The execution behavior data is used as the initial task data.

19. The device according to claim 18, characterized in that The task execution behavior includes: a first type of task execution behavior and a second type of task execution behavior, wherein the first type of task execution behavior is a task execution behavior automatically generated by the network optimization system, and the second type of task execution behavior is a task execution behavior generated by the network optimization system receiving a task execution instruction of a target object; Wherein, the second acquisition module is further used for: Acquire first execution behavior data generated by the first-type task execution behavior at the current node level and second execution behavior data generated by the second-type task execution behavior at the current node level; The first execution behavior data and the second execution behavior data are used together as the execution behavior data.

20. The device according to claim 17, characterized in that The training module is specifically used for: Get the preset model training times and model training conditions; Determining whether the initial network optimization model meets the model training conditions; When the initial network optimization model meets the model training condition and the training times of the initial network optimization model does not reach the preset model training times, the initial network optimization model is trained according to the initial task data and the historical rule data to obtain the target network optimization model.

21. The device according to claim 20, characterized in that The model training conditions include at least one of the following: A model training trigger message has been generated in the network optimization system; The number of the first network optimization tasks completed by the initial network optimization model is greater than a task number threshold.

22. The device according to claim 21, characterized in that The training module is also used for: Performing data format standardization processing on the initial task data to obtain target task data; The initial network optimization model is trained according to the target task data and the historical rule data to obtain the target network optimization model.

23. The device according to claim 22, characterized in that The training module is also used for: Update the historical rule data according to the target task data to obtain processed rule data; Training the initial network optimization model according to the processed rule data to obtain a first candidate network optimization model; Combine the historical rule data and the target task data to train the initial network optimization model to obtain a second candidate network optimization model; The target network optimization model is determined from the first candidate network optimization model and the second candidate network optimization model.

24. The device according to claim 23, characterized in that The training module is also used for: Acquire a model test indicator, wherein the model test indicator is used to test the model training results of the first candidate network optimization model and the second candidate network optimization model; Performing test processing on the model training result of the first candidate network optimization model according to the model test index to obtain a first model index parameter; Performing a test process on the model training result of the second candidate network optimization model according to the model test index to obtain a second model index parameter; The target network optimization model is selected from the first candidate network optimization model and the second candidate network optimization model according to the first model indicator parameter and the second model indicator parameter.

25. The device according to claim 24, characterized in that The training module is also used for: Get indicator parameter thresholds; Determine a target model index parameter whose value is closest to the index parameter threshold value from the first model index parameter and the second model index parameter; A candidate network optimization model that matches the target model indicator parameters is selected from the first candidate network optimization model and the second candidate network optimization model as the target network optimization model.

26. The device according to claim 23, characterized in that The target task data includes: execution node behavior data, wherein the execution node behavior data is data generated when the first task node is a task execution node; the historical rule data includes: rule threshold data and rule content data; Wherein, the training module is also used for: The rule threshold data and the rule content data are updated according to the execution node behavior data to obtain the processed rule data.

27. The device according to claim 17, characterized in that Also includes: A fourth acquisition module is used to acquire a model business effectiveness index and a model production effectiveness index of the target network optimization model before executing the second network optimization task in the network optimization system according to the target network optimization model, wherein the model business effectiveness index is used to describe the model performance of the target network optimization model, and the model production effectiveness index is used to describe the production efficiency of the target network optimization model; A first determination module, used to determine a first indicator parameter and a second indicator parameter, wherein the first indicator parameter and the second indicator parameter are used to participate in determining whether to perform the second network optimization task; A second determination module, configured to determine a model effectiveness parameter of the target network optimization model according to the model business effectiveness indicator, the model production effectiveness indicator, the first indicator parameter, and the second indicator parameter; The third determination module is used to determine the working status information of the target network optimization model according to the model effectiveness parameter, wherein the working status information is used to indicate whether the target network optimization model performs the second network optimization task.

28. The device according to claim 27, characterized in that The third determination module is specifically used to: Get the performance parameter threshold; If the model effectiveness parameter is greater than or equal to the effectiveness parameter threshold, determining that the working status information is for the target network optimization model to perform the second network optimization task; If the model effectiveness parameter is less than the effectiveness parameter threshold, it is determined that the working status information indicates that the target network optimization model does not perform the second network optimization task.

29. The device according to claim 17, characterized in that The processing module is specifically used for: Determining, according to the target network optimization model, a first candidate network optimization rule corresponding to the second network optimization task; Determining, according to the historical rule data, a second candidate network optimization rule corresponding to the second network optimization task; Determining the target network optimization rule from the first candidate network optimization rule and the second candidate network optimization rule; The second network optimization task is performed according to the target network optimization rule.

30. The device according to claim 29, characterized in that The processing module is further used for: Determining a rule weight coefficient of the first candidate network optimization rule according to the model business effectiveness indicator and the model production effectiveness indicator; If the rule weight coefficient is greater than the weight coefficient threshold, taking the first candidate network optimization rule as the target network optimization rule; If the rule weight coefficient is less than or equal to the weight coefficient threshold, the second candidate network optimization rule is used as the target network optimization rule.

31. The device according to claim 29, characterized in that The processing module is also used to: After executing the second network optimization task according to the target network optimization rule, obtaining an execution result of the target network optimization rule; If the execution result indicates that the target network optimization rule is not executed successfully, updating the target network optimization rule according to the first candidate network optimization rule and the second candidate network optimization rule; Based on the updated target network optimization rule, the second network optimization task is executed.

32. The device according to claim 28, characterized in that The third determination module is further used for: After determining that the working status information indicates that the target network optimization model does not perform the second network optimization task, generating a model training trigger message, wherein the model training trigger message is used to trigger training processing of the target network optimization model to optimize the model performance of the target network optimization model; The second network optimization task is performed based on the optimized target network optimization model.

33. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 16.

34. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: in, The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 16.

Images