Gray scale switch control method and device, network equipment and storage medium

Abstract

The invention relates to the technical field of computers, in particular to a gray scale switch control method and device, network equipment and a storage medium. The gray scale switch control method comprises the steps that a data set corresponding to a target node set is acquired according to service demand information, and the data set comprises a user behavior data subset and a system performance data subset; a first prediction model is adopted to carry out identification processing on the data set, and a state identification result corresponding to the gray level switch is obtained; and controlling the on-off state of the gray scale switch according to the state identification result. According to the invention, the invasion of gray scale switch control can be reduced, and the high flexibility and intelligence of gray scale switch control can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a grayscale switch control method, device, network equipment and storage medium. Background Art

[0002] With the development of science and technology, more and more services are provided to users, and improving service quality has become the focus of users. For example, in a large-scale distributed system, grayscale switches can be controlled. For example, grayscale switch control can be performed through hardware coding, or through configuration files, or through dynamic service discovery and routing, or grayscale switch control based on rule engine grayscale strategy, but these control methods are highly invasive and inflexible, making grayscale switch control inaccurate. Summary of the invention

[0003] The present disclosure provides a grayscale switch control method, device, network device and storage medium, which can reduce the intrusiveness of grayscale switch control and improve the flexibility and intelligence of grayscale switch control. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, a grayscale switch control method is provided, comprising:

[0005] According to the business demand information, a data set corresponding to the target node set is obtained, wherein the data set includes a user behavior data subset and a system performance data subset;

[0006] Using the first prediction model to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch;

[0007] According to the state recognition result, the on / off state of the grayscale switch is controlled.

[0008] Optionally, the method further includes:

[0009] According to the data type and the data attribute, a first training data set is collected through a node set, wherein the node set includes a network node subset and a server subset;

[0010] Iteratively training a second prediction model corresponding to the current application scenario on the first training data set to obtain a third prediction model;

[0011] Using a verification data set to verify the third prediction model, and obtaining a verification result corresponding to the third prediction model;

[0012] When the verification result indicates that the third prediction model does not meet the model requirements, the third prediction model is adjusted until the first prediction model that meets the model requirements is obtained.

[0013] Optionally, the using a verification data set to verify the third prediction model to obtain a verification result corresponding to the third prediction model includes:

[0014] Splitting the verification data set to obtain a preset number of verification data subsets;

[0015] A preset number of validation data subsets are used to cross-validate the third prediction model to obtain a cross-validation result corresponding to the third prediction model, wherein the cross-validation result includes at least one of classification accuracy, classification precision, classification recall and F1 score.

[0016] Optionally, the method further includes:

[0017] Acquire the second prediction model according to the data volume corresponding to the first training data set;

[0018] or,

[0019] Acquire the second prediction model according to the data characteristics corresponding to each data in the first training data set;

[0020] or,

[0021] According to the model interpretability requirement, obtaining the second prediction model;

[0022] or,

[0023] The second prediction model is obtained according to the category distribution information corresponding to the first training data set.

[0024] Optionally, before the first prediction model is used to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch, the method further includes:

[0025] According to the business requirements, feature engineering is performed on the data set to obtain a feature set corresponding to the data set;

[0026] According to the label generation rule, label identification is performed on each data in the data set to obtain label information corresponding to each data.

[0027] Optionally, the adopting the first prediction model to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch includes:

[0028] According to the label information and the attribute information, the data set is preprocessed to obtain a preprocessed data set, wherein the data preprocessing includes at least one of cleaning data processing, standardizing data processing, and normalizing data processing;

[0029] When the format information corresponding to the preprocessed data set does not meet the format requirement, performing format conversion processing on the cleaned data set to obtain a format-converted data set;

[0030] The first prediction model is used to perform recognition processing on the data set after the format conversion processing to obtain a state recognition result corresponding to the grayscale switch.

[0031] Optionally, the method further includes:

[0032] Updating the first prediction model once every preset time period;

[0033] or,

[0034] When the second training data set is obtained, the first prediction model is updated.

[0035] According to a second aspect of an embodiment of the present disclosure, there is provided a grayscale switch control device, comprising:

[0036] A set acquisition unit, used to acquire a data set corresponding to a target node set according to business demand information, wherein the data set includes a user behavior data subset and a system performance data subset;

[0037] A result acquisition unit, used to perform recognition processing on the data set using a first prediction model to obtain a state recognition result corresponding to the grayscale switch;

[0038] A switch control unit is used to control the on / off state of the grayscale switch according to the state recognition result.

[0039] According to a third aspect of an embodiment of the present disclosure, a network device is provided, including:

[0040] processor;

[0041] a memory for storing instructions executable by the processor;

[0042] The processor is configured to execute the instructions to implement the grayscale switch control method described in any one of the aforementioned aspects.

[0043] According to a fourth aspect of an embodiment of the present disclosure, a storage medium is provided. When instructions in the storage medium are executed by a processor of a network device, the network device can execute the grayscale switch control method described in any one of the preceding aspects.

[0044] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements any one of the methods described in the preceding aspects when executed by a processor.

[0045] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0046] In some or related embodiments, a data set corresponding to a target node set is obtained according to business demand information, wherein the data set includes a user behavior data subset and a system performance data subset; the data set is identified and processed using a first prediction model to obtain a state identification result corresponding to a grayscale switch; and the on / off state of the grayscale switch is controlled according to the state identification result. Therefore, by obtaining a data set through business demand information, the matching of the data set and the business demand information can be improved, and the on / off state of the grayscale switch can be controlled according to the state identification result, thereby improving the flexibility of grayscale control, supporting fine-grained functional control, and without modifying business codes or configuration files, the intrusiveness of business codes can be reduced, maintenance costs can be reduced, and the security of grayscale switch control can be improved. Moreover, by using a prediction model for identification, the grayscale switch state can be automatically adjusted according to real-time data to ensure system performance and user experience, thereby reducing the intrusiveness of grayscale switch control and improving the flexibility and intelligence of grayscale switch control.

[0047] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0049] Figure 1 is a flow chart of a grayscale switch control method according to an exemplary embodiment;

[0050] Figure 2 is a flow chart of a grayscale switch control method according to an exemplary embodiment;

[0051] Figure 3 is a schematic diagram showing an example of a grayscale switch control method according to an exemplary embodiment;

[0052] Figure 4 is a block diagram of a grayscale switch control device according to an exemplary embodiment;

[0053] Figure 5The invention is a block diagram of a network device according to an exemplary embodiment. DETAILED DESCRIPTION

[0054] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0055] The embodiments of the present disclosure provide grayscale switch control methods, devices, network devices, and storage media. In some embodiments, grayscale switch control methods, information processing methods, communication methods, and other terms can be interchangeable, grayscale switch control devices, information processing devices, communication devices, and other terms can be interchangeable, and information processing systems, communication systems, and other terms can be interchangeable.

[0056] The embodiments of the present disclosure are not exhaustive, but are only illustrative of some embodiments, and are not intended to be a specific limitation on the scope of protection of the present disclosure. In the absence of contradiction, each step in a certain embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a certain embodiment can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment can be arbitrarily exchanged. In addition, the optional implementation methods in a certain embodiment can be arbitrarily combined; in addition, the embodiments can be arbitrarily combined, for example, some or all of the steps of different embodiments can be arbitrarily combined, and a certain embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0057] In each embodiment of the present disclosure, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between the embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form a new embodiment based on their internal logical relationships.

[0058] The terms used in the embodiments of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0059] In the embodiments of the present disclosure, unless otherwise specified, elements expressed in the singular form, such as "a", "an", "the", "above", "said", "aforementioned", "this", etc., may mean "one and only one", or "one or more", "at least one", etc. For example, when using articles such as "a", "an", "the" in English in translation, the noun after the article may be understood as a singular expression or a plural expression.

[0060] In the embodiments of the present disclosure, “plurality” refers to two or more.

[0061] In some embodiments, the terms “at least one,” “one or more,” “a plurality of,” “multiple,” etc. may be used interchangeably.

[0062] In some embodiments, "at least one of A and B", "A and / or B", "A in one case, B in another case", "in response to one case A, in response to another case B", etc., may include the following technical solutions according to the situation: in some embodiments, A (A is executed independently of B); in some embodiments, B (B is executed independently of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). When there are more branches such as A, B, C, etc., the above is also similar.

[0063] In some embodiments, devices and equipment may be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. In some cases, they may also be understood as "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "subject", etc.

[0064] In some embodiments, "terminal" or "terminal device" can be referred to as "user equipment (UE)", "user terminal" "mobile station (MS)", "mobile terminal (MT)", subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, etc.

[0065] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0066] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0067] According to some embodiments, the control method for the grayscale switch may include multiple methods, among which, for example, the grayscale switch control can be performed by hardware coding, or by configuration files, or by dynamic service discovery and routing, or by grayscale switch control based on a rule engine. Among them, when controlling by hard-coded switches, for example, configuration switches (also called feature switches or function flags) can be directly embedded in the application code, and the enabling or disabling of new functions can be controlled by modifying the status of these switches. This method requires developers to modify the code directly and redeploy the entire application each time it is released. However, there will be a need to frequently modify and deploy the code, which increases the maintenance cost, and the switch state change usually requires restarting the application or redeploying, and the response speed is slow. Among them, when controlling the switch by a configuration file, the configuration of the grayscale switch can be stored in an external configuration file (such as a JSON, YAML, or XML file), and the application reads these configuration files at startup or runtime to determine the status of the grayscale switch. This method reduces the need to modify the code directly, but still requires restarting the application or performing a hot update of the configuration file. However, the loss or error of the configuration file may cause malfunction. The update of the configuration file may require additional deployment steps or tool support, rely on external configuration and make the update complicated.

[0068] Among them, when using dynamic service discovery and routing, service registration and discovery mechanisms (such as Eureka, Consul or ZooKeeper) and intelligent routing services (such as Nginx Plus, Istio or Envoy) can be used to dynamically route traffic to different service versions based on user request attributes (such as geographic location, user ID, request header, etc.). This method can achieve more fine-grained traffic control and grayscale testing, but usually involves complex network configuration and service governance. However, additional service discovery and routing components need to be deployed and maintained, and operation and maintenance personnel need to master additional technology stacks. The system is highly complex and has a steep learning curve. Among them, when grayscale switch control is performed through grayscale policies based on rule engines, rule engines (such as Drools, EasyRules, etc.) can be used to define grayscale release rules, and different release strategies can be dynamically applied according to preset conditions (such as user attributes, request parameters, etc.). This method provides high flexibility and configurability, but the definition and maintenance of rules may be more complicated. However, as the business complexity increases, the number and complexity of rules will also increase, and the processing of the rule engine may introduce additional performance overhead, so rule management is complex and performance considerations are required. In summary, these methods are highly invasive and inflexible, and cannot automatically adjust the switch status according to real-time data, cannot effectively respond to system performance fluctuations or changes in user behavior, and lack intelligence.

[0069] Figure 1 is a flow chart of a grayscale switch control method according to an exemplary embodiment. Figure 1 As shown, the following steps are included:

[0070] In step S11, according to the business demand information, a data set corresponding to the target node set is obtained, wherein the data set includes a user behavior data subset and a system performance data subset;

[0071] In some embodiments, the execution subject of the embodiments of the present disclosure may be, for example, a network device, and the name of the network device is not limited. For example, the network device may also be called a server, a base station, or a server cluster, etc. The embodiments of the present disclosure are not limited to this.

[0072] In some embodiments, the technical solution of the embodiments of the present disclosure can be used, for example, for the grayscale release and management of new functions in large-scale distributed systems. Specifically, for example, it can be applied to scenarios where artificial intelligence technology is used for non-invasive method-level intelligent control of grayscale switches.

[0073] According to some embodiments, the business requirement information may be used to indicate information required for the current business execution. The business requirement information does not specifically refer to a fixed information. For example, when the application scenario corresponding to the business requirement information changes, the business requirement information may also change accordingly.

[0074] In some embodiments, the target node set may be, for example, a group formed by at least one node. The target node may, for example, refer to a node involved in the current business operation. The target node set is not just a fixed set. For example, when the current running business changes, the target node set may also change accordingly. For example, when a node in the target node set changes, the target node set may also change accordingly.

[0075] The node may be, for example, a critical network node, wherein the node may be, for example, a location in the network where data traffic is greater than a traffic threshold, the number of interactions is greater than a number threshold, or the requirements for business continuity are met.

[0076] According to some embodiments, for example, a data collection tool may be deployed on the server side, wherein the data collection tool may be used to collect various data generated when the server processes a business request, including but not limited to user behavior logs, system performance indicators (such as CPU usage, memory usage, response time, etc.), business transaction data, etc. Among them, these data are useful for evaluating system functional performance, user satisfaction, and system stability. The server may be a Web server, an application server, a database server, etc., and the specific type of the server depends on the system architecture and business requirements.

[0077] According to some embodiments, a data set may be, for example, a collection of at least one data. The data set does not specifically refer to a fixed data. For example, when the embodiment of the present disclosure is applied to online insurance of an insurance company, the data set includes but is not limited to a user behavior data subset and a system performance data subset, that is, the data set includes but is not limited to the user's insurance behavior data and the performance indicator data of the server. The behavior data may include, for example, browsing insurance type data, filled-in insurance data, and insurance application related data. The performance indicator data of the server includes but is not limited to response time and error rate. The data set does not specifically refer to a fixed set. For example, when the number of data included in the data set changes, the data set may also change accordingly. For example, when a certain data in the data set changes, the data set may also change accordingly.

[0078] In some embodiments, the user behavior data subset may be, for example, a collection of at least one user behavior data. The user behavior data subset does not specifically refer to a fixed set. For example, when the number of user behavior data included in the user behavior data subset changes, the user behavior data subset may also change accordingly. For example, when a certain user behavior data in the user behavior data subset changes, the user behavior data subset may also change accordingly.

[0079] According to some embodiments, the system performance data subset may be, for example, a collection of at least one system performance data. The system performance data subset does not specifically refer to a fixed set. For example, when the number of system performance data included in the system performance data subset changes, the system performance data subset may also change accordingly. For example, when a certain system performance data in the system performance data subset changes, the system performance data subset may also change accordingly.

[0080] According to some embodiments, a data set corresponding to the target node set may be acquired based on business demand information, wherein the data set includes a user behavior data subset and a system performance data subset.

[0081] In step S12, the data set is identified using the first prediction model to obtain a state identification result corresponding to the grayscale switch;

[0082] In some embodiments, the first prediction model may be, for example, a model that has been trained and can be used for state recognition, and the first prediction model does not specifically refer to a fixed model. For example, when the training conditions of the first prediction model change, the first prediction model may also change accordingly. For example, when the training data set changes, the first prediction model may also change accordingly.

[0083] According to some embodiments, the identification process may be, for example, a process of identifying the grayscale of a grayscale switch.

[0084] In some embodiments, grayscale switching can be a term commonly used in the software development and release process, which usually refers to a mechanism for controlling the gradual release of new features or changes to the user group. Through grayscale switching, developers can open new features or changes to a small number of users first, collect feedback and test the effects, and then decide whether to promote them to a wider range of users based on the collected data. This approach helps reduce the risks that may be brought about by the launch of new features and ensure the stable operation of the system.

[0085] In some embodiments, the state recognition result can be used to indicate the state of the grayscale switch, for example. The state recognition result does not specifically refer to a fixed result. For example, when the time point of state recognition changes, the state recognition result can also change accordingly. For example, when the first prediction model changes, the state recognition result can also change accordingly. For example, when the grayscale switch changes, the state recognition result can also change accordingly.

[0086] In some embodiments, when a data set is obtained, a first prediction model may be used to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch.

[0087] In step S13, the on / off state of the grayscale switch is controlled according to the state recognition result.

[0088] In some embodiments, the on / off state may include, for example, an on state or a off state. According to the state recognition result, the on / off state of the grayscale switch may be controlled. For example, it may include turning on the grayscale switch according to the state recognition result. For example, it may include turning off the grayscale switch according to the state recognition result.

[0089] In some or related embodiments, a data set corresponding to a target node set is obtained according to business demand information, wherein the data set includes a user behavior data subset and a system performance data subset; a first prediction model is used to identify and process the data set to obtain a state identification result corresponding to a grayscale switch; and the on / off state of the grayscale switch is controlled according to the state identification result. Therefore, the data set can be obtained through business demand information, which can improve the matching of the data set with the business demand information, and the machine learning model can be used to predict the optimal state of the grayscale switch, and automatically adjusted according to real-time data, effectively responding to system performance fluctuations and user behavior changes, reducing the risks brought by the launch of new functions, and the on / off state of the grayscale switch can be adjusted according to the state identification result, improving the flexibility of grayscale control, supporting fine-grained function control, and without modifying business codes or configuration files, which can reduce the intrusiveness of business codes, reduce maintenance costs, and improve the security of grayscale switch control. The prediction model is used for identification, and the grayscale switch state can be automatically adjusted according to real-time data to ensure system performance and user experience, thereby reducing the intrusiveness of grayscale switch control and improving the high flexibility and intelligence of grayscale switch control.

[0090] Figure 2 is a flow chart of a grayscale switch control method according to an exemplary embodiment. Figure 2 As shown, the following steps are included:

[0091] In step S21, according to the data type and data attribute, a first training data set is collected through a node set, wherein the node set includes a network node subset and a server subset;

[0092] In some embodiments, the execution subject of the embodiments of the present disclosure may be, for example, a network device, and the name of the network device is not limited. For example, the network device may also be called a server, a base station, or a server cluster, etc. The embodiments of the present disclosure are not limited to this. The application scenarios of the embodiments of the present disclosure may be used, for example, for grayscale switch control, specifically, for example, for grayscale testing of new functions, testing of new functions of e-commerce platforms, personalized recommendations for financial services, content updates of online education platforms, and performance optimization of enterprise-level applications.

[0093] The specific process is as described above and will not be repeated here.

[0094] In some embodiments, the data type can be used to indicate the type of message data that needs to be collected. The data type includes but is not limited to the JSON data type and the XML data type, etc. These two data types have obvious structural characteristics and the convenience of related conversion is high. The data type does not specifically refer to a fixed data type. For example, when a modification instruction for the data type is received, the data type can also change accordingly. Among them, the data type of the embodiment of the present disclosure can be, for example, the JSON data type.

[0095] Among them, when a data set is obtained, the data type of the data set can be identified. When the data type of the data set is not a preset data type, the type of the data set can be converted to obtain a data set in a preset format, which can improve the uniformity of training data acquisition and improve the accuracy of model training.

[0096] In some embodiments, the preset data type may be, for example, an XML data type. When a data set of a JSON data type is obtained, the data set may be formatted and XML tags and text content may be extracted from the converted data.

[0097] According to some embodiments, data attributes may be used to indicate the attributes of the acquired data, including but not limited to timestamp, sender information, receiver information, message content, etc. The data attribute does not specifically refer to a fixed attribute. For example, when the number of attributes corresponding to the data attribute changes, the data attribute may also change accordingly. For example, when a modification instruction for the data attribute is received, the data attribute may also change accordingly.

[0098] According to some embodiments, the node set may be, for example, a collection of at least one node. Each node in the node set is provided with a data collection tool. The node set does not specifically refer to a fixed set. For example, when the current scene changes, the determined node set may also change accordingly. For example, when the number of nodes corresponding to the node set changes, the node set may also change accordingly.

[0099] In some embodiments, the network node subset may be, for example, a group formed by at least one network node. The network node subset does not specifically refer to a fixed set. For example, when the number of network nodes corresponding to the network node subset changes, the network node subset may also change accordingly. For example, when a network node in the network node subset changes, the network node subset may also change accordingly.

[0100] According to some embodiments, the server subset may be, for example, a group formed by at least one server. The server subset does not specifically refer to a fixed set. For example, when the number of servers corresponding to the server subset changes, the server subset may also change accordingly. For example, when a server in the server subset changes, the server subset may also change accordingly.

[0101] In some embodiments, the first training data set may be, for example, a data set used for model training. When the first one of the first training data set changes with the remaining training data sets, the first training data set may also change accordingly. The first training data set does not specifically refer to a fixed set. For example, when the amount of data included in the first training data set changes, the first training data set may also change accordingly. For example, when the method for obtaining the first training data set changes, the first training data set changes accordingly.

[0102] In some embodiments, a first training data set is collected through a node set according to the data type and data attribute. The first training data set may include, for example, user behavior data and performance indicator data within a historical period. The name of the performance indicator data is not limited, and may also be called system performance data, for example.

[0103] The training data may be obtained by, for example, capturing and analyzing data packets in the network, and a log collection system may be used to aggregate log data of servers, applications, and systems.

[0104] According to some embodiments, when a data set for building a model is obtained, the data can be divided according to a ratio to obtain a first training data set and a validation data set. When the data set is obtained, the data set can be preprocessed, and the preprocessing includes but is not limited to feature scaling, data validation, and encoding processing.

[0105] The data verification method mainly includes at least one of the following:

[0106] Completeness verification: Checks whether the collected data is complete and no key fields or records are missing.

[0107] Consistency verification: Check whether the logical relationship between data is consistent, for example, whether there is a reasonable correlation between user behavior data and system performance indicators.

[0108] Accuracy verification: Checking the accuracy of data, for example by verifying the authenticity of data by comparing it to multiple data sources.

[0109] Deduplication verification: Remove duplicate data records to ensure the purity of the data set.

[0110] According to some embodiments, the data may be processed and converted, wherein the conversion process is as described above and will not be described in detail herein. The format conversion may be, for example, converted into a format suitable for processing by a machine learning model, such as CSV or Pandas DataFrame. Data processing includes data cleaning, and unnecessary labels or attributes may be removed.

[0111] In step S22, iteratively training the second prediction model corresponding to the current application scenario on the first training data set to obtain a third prediction model;

[0112] In some embodiments, the current application scenario may refer to, for example, a scenario in which grayscale switch control is currently being performed. The application scenario includes, but is not limited to, an insurance scenario, a clothing shopping scenario, etc. The current application scenario of the embodiment of the present disclosure may be, for example, an insurance scenario.

[0113] According to some embodiments, the second prediction model may be, for example, a model that corresponds to the current application scenario and has not been trained, and the second in the second prediction model does not specifically refer to a fixed model. For example, when the model type corresponding to the second prediction model changes, the second prediction model may also change accordingly. For example, when the current application scenario changes, the second prediction model may also change accordingly.

[0114] For example, a loss function and an optimizer corresponding to the second prediction model can be selected, wherein the loss function can be used to measure the gap between the model prediction and the actual value, and the optimizer can include a gradient descent method or its variants (such as Adam, RMSprop, etc.) for minimizing the loss function.

[0115] The choice of loss function depends on the specific machine learning task and goal.

[0116] Classification problem: loss functions include cross-entropy loss, logarithmic loss, hinge loss, etc.

[0117] Regression problem: loss functions include mean squared error (MSE), mean absolute error (MAE), etc.

[0118] According to some embodiments, the third prediction model may be, for example, a third training model obtained after iterative training using training data. The third prediction model may be the first prediction model, or may not be the first prediction model, and the third prediction model may be determined after verification of the third prediction model based on the verification data set.

[0119] According to some embodiments, the second prediction model corresponding to the current application scenario is iteratively trained on the first training data set to obtain a third prediction model. In each iterative training, for example, the parameters of the second prediction model can be adjusted according to the value of the loss function, and the model parameters of the prediction model can be gradually optimized through multiple epochs (the number of times the entire data set is traversed).

[0120] According to some embodiments, the training process may also include, for example, hyperparameter adjustment, wherein the hyperparameters may include, for example, learning rate, batch size, number of epochs, etc. By adjusting the hyperparameters, for example, the training effect of the model may be improved. For example, grid search, random search, or Bayesian optimization may be used to find the best hyperparameter combination. The embodiments of the present disclosure are not limited to this.

[0121] In step S23, the third prediction model is verified using the verification data set to obtain a verification result corresponding to the third prediction model;

[0122] The specific process is as described above and will not be repeated here.

[0123] In some embodiments, the verification data set may be, for example, a collection of at least one training data. The verification data set may be, for example, a verification data set obtained by proportional distribution in the data set. The verification data set does not specifically refer to a fixed set. For example, when the amount of data included in the verification data set changes, the verification data set may also change accordingly.

[0124] According to some embodiments, the third prediction model is verified using a verification data set to obtain a verification result corresponding to the third prediction model.

[0125] According to some embodiments, the using the verification data set to verify the third prediction model and obtaining the verification result corresponding to the third prediction model includes:

[0126] Splitting the verification data set to obtain a preset number of verification data subsets;

[0127] A preset number of validation data subsets are used to cross-validate the third prediction model to obtain a cross-validation result corresponding to the third prediction model, wherein the cross-validation result includes at least one of classification accuracy, classification precision, classification recall and F1 score.

[0128] According to some embodiments, the method further comprises:

[0129] Acquire the second prediction model according to the data volume corresponding to the first training data set;

[0130] or,

[0131] Acquire the second prediction model according to the data characteristics corresponding to each data in the first training data set;

[0132] or,

[0133] According to the model interpretability requirement, obtaining the second prediction model;

[0134] or,

[0135] The second prediction model is obtained according to the category distribution information corresponding to the first training data set. Therefore, the matching of the prediction model with the current application scenario can be improved, the accuracy of the prediction model acquisition can be improved, and the accuracy of the prediction can be improved.

[0136] According to some embodiments, when performing model selection, the selected prediction model may include, for example, logistic regression, support vector machine, random forest, etc.

[0137] According to some embodiments, for example, corresponding labels (such as insurance type, branch, etc.) can be generated for each XML message based on existing business rules or existing data, and these labels can be used as target variables in supervised learning.

[0138] According to some embodiments, for example, feature engineering may be performed on the data, specifically including at least one of the following:

[0139] Data cleaning: Remove outliers, noise, and irrelevant data to ensure the quality of the data set.

[0140] Feature extraction: Extract features useful for model training from raw data. For example, extract features such as user activity and preference categories from user behavior data; extract features such as response time and error rate from system performance data.

[0141] Feature encoding: For categorical features, encoding processing (such as one-hot encoding, label encoding, etc.) is required so that the model can process it.

[0142] Feature scaling: For numerical features, standardization or normalization may be required to eliminate the dimensional differences between different features.

[0143] Feature selection: Select the features that are most helpful for model training based on business needs and data characteristics.

[0144] In step S24, when the verification result indicates that the third prediction model does not meet the model requirements, the third prediction model is adjusted until the first prediction model that meets the model requirements is obtained;

[0145] The specific process is as described above and will not be repeated here.

[0146] According to some embodiments, when the verification result indicates that the third prediction model does not meet the model requirements, the third prediction model is adjusted until the first prediction model that meets the model requirements is obtained.

[0147] In some embodiments, for example, the model can be prevented from overfitting by using regularization techniques, where the regularization techniques may include L1 and L2 regularization, or the risk of underfitting may be reduced by increasing the amount of data and using ensemble learning methods.

[0148] In some embodiments, when the first prediction model is obtained, the first prediction model can be stored and deployed to the corresponding application scenario to improve the convenience of the prediction or classification task. When the model is deployed, for example, it can be deployed to the same environment as the training environment to reduce the situation where the prediction model is inaccurate due to inconsistent deployment of the prediction model.

[0149] In some embodiments, for example, the state of the grayscale switch can be predicted and analyzed through historical data. The specific process may include, for example, obtaining a historical data set, wherein the type of prediction model used may be determined based on the amount of data corresponding to the historical data set. When the amount of data is greater than the preset amount of data, the prediction model may include, for example, a random forest classifier, a support vector machine (SVM), or a neural network. When the amount of data is greater than the preset amount of data, the prediction model may include, for example, a linear regression or a decision tree classifier, which may reduce the occurrence of overfitting.

[0150] In some embodiments, for example, a prediction model may be selected according to the characteristics of the data. For example, when the feature in the data set is of a classification type, the prediction model may be, for example, a decision tree or a naive Bayes classifier. For example, when the feature in the data set is of a numerical type, the prediction model may be, for example, a vector machine or a neural network.

[0151] In some embodiments, for example, the prediction model can also be selected based on the interpretability and accuracy of the model. For example, when selecting the reason for making a predetermined prediction, a decision tree or logistic regression can be selected. For example, when selecting accuracy as the main factor, a random forest or deep learning model can be selected.

[0152] In some embodiments, for example, a prediction model may be selected based on the type of processing imbalance. For example, when processing imbalance is selected, a random forest, a support vector machine, or a neural network model may be selected.

[0153] In some embodiments, for example, a random forest classifier may be selected as a machine learning model. Random forest is an ensemble learning method that improves the stability and accuracy of the model by constructing multiple decision trees and voting or averaging their prediction results.

[0154] Advantages of the random forest classifier include, for example, the following:

[0155] High accuracy: By integrating the predictions of multiple decision trees, random forests are often able to achieve higher prediction accuracy than a single decision tree.

[0156] Resistance to overfitting: Since each decision tree is trained on a randomly sampled subset of the data and each node split only considers a randomly selected subset of features, the random forest model is not prone to overfitting.

[0157] Ability to handle large number of features: Random forests can effectively handle datasets with a large number of features without the need for feature selection.

[0158] Easy to parallelize: Each tree in a random forest can be trained independently, which makes the algorithm easy to parallelize, thereby improving computational efficiency.

[0159] According to some embodiments, when a historical data set is obtained, the data set may be preprocessed, wherein the preprocessing may include, for example, cleaning the data, specifically, for example, removing or filling missing values. For categorical features, for example, encoding processing may also be included, wherein the encoding may include, for example, one-hot encoding, and the data processing may also include standardization or normalization of numerical features.

[0160] In some embodiments, when a historical data set is obtained, the historical data set can be divided into a training data set, a verification data set, and a test data set. The training data set is used to train the prediction model; the verification data set is used to verify the trained model, adjust the model parameters, and evaluate the model performance; the test data set is used to finally evaluate the generalization ability of the model.

[0161] In some embodiments, a random forest model is trained using a training data set. During the training process, model performance can be optimized by adjusting parameters (such as the number of trees, maximum depth, minimum sample split, etc.).

[0162] The parameter setting for the prediction model may include, for example, multiple parameters of the random forest model, specifically, for example, the number of trees (n_estimators), the maximum depth (max_depth), the minimum sample split (min_samples_split), etc. For example, n_estimators=100 may be set, indicating that 100 decision trees are constructed in the random forest.

[0163] According to some embodiments, a random forest model is trained using a training data set. Each tree is trained on a random subset of the training set, and each node splits considering only a random subset of features. During the training process, the model learns how to predict the target variable based on the input features.

[0164] In some embodiments, a validation data set is used to evaluate the final performance of the model. The classification performance of the model can be evaluated by indicators such as accuracy, precision, recall, and F1 score.

[0165] In order to more accurately evaluate the performance of the prediction model, cross-validation techniques can be used. For example, 5-fold cross-validation divides the training set into 5 parts, uses 4 of them for training each time, and the remaining 1 part for validation, repeating 5 times, using a different part as the validation set each time. Through cross-validation, the performance of the model on different data sets can be obtained, thereby more comprehensively evaluating the stability and generalization ability of the model.

[0166] In some embodiments, a validation data set or a test data set may be used to evaluate the performance of the prediction model. Common classification evaluation indicators include accuracy, precision, recall, and F1 score. For example, if the accuracy of the model on the validation set is 90%, it means that the prediction model can correctly classify 90% of the samples, and it can be determined that the prediction model meets the model requirements and obtains the prediction model.

[0167] In some embodiments, the parameters of the prediction model may be adjusted, for example, including adjusting the model parameters according to the performance on the validation data set. For example, if the prediction model is found to be overfitting, the number of trees may be increased or the maximum depth may be reduced to reduce the complexity of the model. Through multiple attempts and adjustments, an optimal set of parameter configurations may be found.

[0168] In step S25, according to the business demand information, a data set corresponding to the target node set is obtained, wherein the data set includes a user behavior data subset and a system performance data subset;

[0169] The specific process is as described above and will not be repeated here.

[0170] According to some embodiments, the data set may include, for example, a user behavior data subset and a system data subset. The user behavior data subset may include, for example, user operation data in the insurance business system, such as browsing data, purchase data, consultation data, etc. The system performance data subset may include, for example, data such as the response time, CPU usage, and memory usage of the collection server.

[0171] According to some embodiments, for example, data processing may be performed on the acquired data set, which may specifically include cleaning and preprocessing data, removing outliers and noise, etc.

[0172] Wherein, obtaining the data set may, for example, be periodically obtaining the data set or obtaining the data set by triggering a specific event. The embodiments of the present disclosure are not limited to this. Wherein, obtaining the data set may, for example, be obtaining data collected from multiple sources without affecting the normal operation of the system, and the collected data may be preliminarily verified to improve the integrity and accuracy of the data. The collected data may be stored in a secure and scalable storage system for subsequent processing and analysis.

[0173] According to some embodiments, before the first prediction model is used to perform recognition processing on the data set to obtain the state recognition result corresponding to the grayscale switch, the method further includes:

[0174] According to the business requirements, feature engineering is performed on the data set to obtain a feature set corresponding to the data set;

[0175] According to the label generation rule, label identification is performed on each data in the data set to obtain label information corresponding to each data. Therefore, the label information and the first prediction model can be used to perform identification processing on the data set to obtain the state identification result corresponding to the grayscale switch, which can improve the accuracy of result acquisition and improve the accuracy of grayscale switch control.

[0176] In step S26, the data set is identified using the first prediction model to obtain a state identification result corresponding to the grayscale switch;

[0177] The specific process is as described above and will not be repeated here.

[0178] According to some embodiments, the using the first prediction model to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch includes:

[0179] According to the label information and the attribute information, the data set is preprocessed to obtain a preprocessed data set, wherein the data preprocessing includes at least one of cleaning data processing, standardizing data processing, and normalizing data processing;

[0180] When the format information corresponding to the preprocessed data set does not meet the format requirement, performing format conversion processing on the cleaned data set to obtain a format-converted data set;

[0181] The first prediction model is used to perform recognition processing on the data set after the format conversion processing to obtain a state recognition result corresponding to the grayscale switch.

[0182] According to some embodiments, the first prediction model is updated every preset time period;

[0183] or,

[0184] When the second training data set is obtained, the first prediction model is updated. Therefore, the accuracy of model training can be improved. For example, data can be re-collected and the model can be re-trained to improve the prediction performance of the prediction model and improve the model's adaptability to changes in data distribution.

[0185] For example, the prediction model may be updated once every preset time period, wherein the preset time period may be determined, for example, based on data distribution or data change trend, and the preset time period may be, for example, every month or every quarter.

[0186] In step S27, the on / off state of the grayscale switch is controlled according to the state recognition result.

[0187] The specific process is as described above and will not be repeated here.

[0188] According to some embodiments, for example, grayscale switch control can be performed by users, for example, the insurance type code (ProductCode) and branch code (indirectly represented by AgentCode) in the message can be used to implement grayscale switch control based on insurance type and branch.

[0189] For example, an insurance company is testing a new feature for life insurance products (such as a smarter way to calculate premiums) and wants to conduct a grayscale test in Branch A first.

[0190] Message example:

[0191] xml

[0192] <policy>

[0193] <!--...other policy related fields...-->

[0194] <ProductCode tc="73263"> 73263 <! -- Assume 73263 represents life insurance -->

[0195] <!--...Other product information...-->

[0196] <AgentCode tc="SH001"> SH001<! --Assume that SH001 represents the agent code of Branch A -->

[0197] <!--...Other Policy fields...-->

[0198] < / policy>

[0199] Grayscale switch logic:

[0200] When the system receives a message containing ProductCode = 73263 (life insurance) and AgentCode starting with "SH", it determines whether the current user request should be included in the grayscale test scope according to the grayscale rules. If it meets the grayscale test conditions, the new function for life insurance products is enabled; otherwise, the request is processed using the existing function.

[0201] According to some embodiments, an insurance company plans to test a new health insurance claims process optimization function at Branch B.

[0202] Message example:

[0203] xml

[0204] <policy>

[0205] <!--...other policy related fields...-->

[0206] <ProductCode tc="73264"> 73264<!--Assume 73264 represents health insurance-->

[0207] <!--...Other product information...-->

[0208] <AgentCode tc="BJ002"> BJ002<! --Assume that BJ002 represents the agent code of Branch B -->

[0209] <!--...Other Policy fields...-->

[0210] < / policy>

[0211] For example, the grayscale switch logic can be that when the system detects that ProductCode = 73264 (health insurance) and AgentCode starts with "BJ", it determines whether the new claim process optimization function needs to be enabled according to the preset grayscale rules. If the user request meets the grayscale test conditions, it automatically switches to the new process; otherwise, the old process continues to be used to process the claim request.

[0212] According to some embodiments, in order to more comprehensively evaluate the impact of new functions, the insurance company decides to conduct grayscale testing on multiple insurance types and multiple branches at the same time, but the testing strategies are different.

[0213] Message example:

[0214] For the testing of life insurance products at Branch A (as shown in the example above).

[0215] For the testing of health insurance products at Branch B (as shown in the example above).

[0216] It may also include combined testing of other types of insurance (such as accident insurance) and other branches (such as C Branch).

[0217] The grayscale switch logic may be, for example, that after the system receives the message, it first parses out the ProductCode and AgentCode.

[0218] According to the preset grayscale rule library, find the specific testing strategy for the combination of this insurance type and branch.

[0219] Determine whether to enable new features based on the policy, and adjust the grayscale switch status accordingly.

[0220] Record key indicators during the testing process (such as user feedback, system performance, etc.) for subsequent evaluation and optimization.

[0221] According to some embodiments, a life insurance company is developing a new online insurance system, which includes multiple new functions, such as intelligent recommendation of insurance products, automatic underwriting process, etc. In order to ensure the stability and user experience of the new function after it is launched, the company decides to use a non-intrusive method-level grayscale switch to test the new function.

[0222] Figure 3 FIG. 1 is a block diagram of a grayscale switch control device according to an exemplary embodiment. Figure 3 As shown, it may include: user request reception -> data collection -> data processing -> model prediction -> grayscale switch state adjustment -> return processing results. When receiving relevant messages, automatically capture and store relevant information, especially ProductCode (insurance product code) and user identification (such as can be indirectly obtained through the information in the Party tag). Clean and preprocess the collected data to ensure data quality. Store the processed data in the database for subsequent model training and prediction. Use historical insurance data (including user behavior, insurance product performance, etc.) to train the machine learning model. The goal of the model is to predict whether enabling a new function will bring positive effects (such as increasing the insurance success rate, reducing the complaint rate, etc.) for a specific insurance product (such as ProductCode = 73263) and a specific user group. When a new insurance request message is received, extract relevant features (such as ProductCode, user information, etc.). Input the features into the trained model to predict whether the new function should be enabled for the request. According to the model prediction results, automatically adjust the grayscale switch state to decide whether to enable the new function. Continue to collect data after the new function is launched to evaluate its impact on system performance and user experience. Based on the evaluation results, the machine learning model is continuously optimized to improve prediction accuracy. For example, the grayscale strategy can be adjusted, such as expanding or reducing the grayscale test range.

[0223] In some or related embodiments, according to the data type and data attributes, a first training data set is collected through a node set, wherein the node set includes a network node subset and a server subset; a second prediction model corresponding to the current application scenario is iteratively trained using the first training data set to obtain a third prediction model; the third prediction model is verified using a verification data set to obtain a verification result corresponding to the third prediction model; when the verification result indicates that the third prediction model does not meet the model requirements, the third prediction model is adjusted until the first prediction model that meets the model requirements is obtained. Therefore, according to the data type and data attribute, a first training data set can be collected through a node set, and the prediction model can be trained using the training data set, which can improve the accuracy of the prediction model training and the accuracy of the prediction model acquisition, and the model parameters can be continuously adjusted using historical data to improve the prediction accuracy.

[0224] Figure 4 is a block diagram of a grayscale switch control device according to an exemplary embodiment. Figure 4 , the grayscale switch control device 400 comprises:

[0225] A set acquisition unit 401 is used to acquire a data set corresponding to a target node set according to business demand information, wherein the data set includes a user behavior data subset and a system performance data subset;

[0226] A result acquisition unit 402 is used to perform recognition processing on the data set using a first prediction model to obtain a state recognition result corresponding to the grayscale switch;

[0227] The switch control unit 403 is used to control the on / off state of the grayscale switch according to the state recognition result.

[0228] According to some embodiments, the result acquisition unit 402 is further configured to:

[0229] According to the data type and the data attribute, a first training data set is collected through a node set, wherein the node set includes a network node subset and a server subset;

[0230] Iteratively training a second prediction model corresponding to the current application scenario on the first training data set to obtain a third prediction model;

[0231] Using a verification data set to verify the third prediction model, and obtaining a verification result corresponding to the third prediction model;

[0232] When the verification result indicates that the third prediction model does not meet the model requirements, the third prediction model is adjusted until the first prediction model that meets the model requirements is obtained.

[0233] According to some embodiments, the result acquisition unit 402 is used to verify the third prediction model using the verification data set, and when obtaining the verification result corresponding to the third prediction model, specifically:

[0234] Splitting the verification data set to obtain a preset number of verification data subsets;

[0235] A preset number of validation data subsets are used to cross-validate the third prediction model to obtain a cross-validation result corresponding to the third prediction model, wherein the cross-validation result includes at least one of classification accuracy, classification precision, classification recall and F1 score.

[0236] According to some embodiments, the result acquisition unit 402 is further configured to:

[0237] Acquire the second prediction model according to the data volume corresponding to the first training data set;

[0238] or,

[0239] Acquire the second prediction model according to the data characteristics corresponding to each data in the first training data set;

[0240] or,

[0241] According to the model interpretability requirement, obtaining the second prediction model;

[0242] or,

[0243] The second prediction model is obtained according to the category distribution information corresponding to the first training data set.

[0244] According to some embodiments, the result acquisition unit 402 is further configured to use the first prediction model to perform recognition processing on the data set, and before obtaining the state recognition result corresponding to the grayscale switch:

[0245] According to the business requirements, feature engineering is performed on the data set to obtain a feature set corresponding to the data set;

[0246] According to the label generation rule, label identification is performed on each data in the data set to obtain label information corresponding to each data.

[0247] According to some embodiments, the result acquisition unit 402 is used to perform recognition processing on the data set using the first prediction model to obtain the state recognition result corresponding to the grayscale switch, specifically for:

[0248] According to the label information and the attribute information, the data set is preprocessed to obtain a preprocessed data set, wherein the data preprocessing includes at least one of cleaning data processing, standardizing data processing, and normalizing data processing;

[0249] When the format information corresponding to the preprocessed data set does not meet the format requirement, performing format conversion processing on the cleaned data set to obtain a format-converted data set;

[0250] The first prediction model is used to perform recognition processing on the data set after the format conversion processing to obtain a state recognition result corresponding to the grayscale switch.

[0251] According to some embodiments, the result acquisition unit 402 is further configured to:

[0252] Updating the first prediction model once every preset time period;

[0253] or,

[0254] When the second training data set is obtained, the first prediction model is updated.

[0255] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0256] In some or related embodiments, a set acquisition unit is used to acquire a data set corresponding to a target node set according to business demand information, wherein the data set includes a user behavior data subset and a system performance data subset; a result acquisition unit is used to identify and process the data set using a first prediction model to acquire a state recognition result corresponding to a grayscale switch; and a switch control unit is used to control the on / off state of the grayscale switch according to the state recognition result. Therefore, by acquiring a data set through business demand information, the matching of the data set and the business demand information can be improved, and the on / off state of the grayscale switch can be controlled according to the state recognition result, thereby improving the flexibility of grayscale control, supporting fine-grained functional control, and without modifying business codes or configuration files, the intrusiveness of business codes can be reduced, maintenance costs can be reduced, and the security of grayscale switch control can be improved. Moreover, by using a prediction model for identification, the grayscale switch state can be automatically adjusted according to real-time data to ensure system performance and user experience, thereby reducing the intrusiveness of grayscale switch control and improving the flexibility and intelligence of grayscale switch control.

[0257] Figure 5is a block diagram of a network device 500 provided in an embodiment of the present disclosure. For example, the network device 500 may be provided as a network side device. Figure 5 The network device 500 includes a processing component 522, which further includes at least one processor, and a memory resource represented by a memory 532 for storing instructions executable by the processing component 522, such as an application. The application stored in the memory 532 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 522 is configured to execute instructions to perform any method of the aforementioned method applied to the network side device.

[0258] The network device 500 may also include a power supply component 527 configured to perform power management of the network device 500, a wired or wireless network interface 550 configured to connect the network device 500 to a network, and an input / output (I / O) interface 558. The network device 500 may operate based on an operating system stored in the memory 532, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM or the like.

[0259] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0260] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable grayscale switch control device, so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0261] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0262] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0263] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0264] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0265] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0266] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A grayscale switch control method, characterized in that: include: According to the business demand information, a data set corresponding to the target node set is obtained, wherein the data set includes a user behavior data subset and a system performance data subset; Using the first prediction model to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch; According to the state recognition result, the on / off state of the grayscale switch is controlled.

2. The method according to claim 1, characterized in that The method further comprises: According to the data type and the data attribute, a first training data set is collected through a node set, wherein the node set includes a network node subset and a server subset; Iteratively training a second prediction model corresponding to the current application scenario on the first training data set to obtain a third prediction model; Using a verification data set to verify the third prediction model, and obtaining a verification result corresponding to the third prediction model; When the verification result indicates that the third prediction model does not meet the model requirements, the third prediction model is adjusted until the first prediction model that meets the model requirements is obtained.

3. The method according to claim 2, characterized in that The using the verification data set to verify the third prediction model and obtaining a verification result corresponding to the third prediction model includes: Splitting the verification data set to obtain a preset number of verification data subsets; A preset number of validation data subsets are used to cross-validate the third prediction model to obtain a cross-validation result corresponding to the third prediction model, wherein the cross-validation result includes at least one of classification accuracy, classification precision, classification recall and F1 score.

4. The method according to claim 2 or 3, characterized in that: The method further comprises: Acquire the second prediction model according to the data volume corresponding to the first training data set; or, Acquire the second prediction model according to the data characteristics corresponding to each data in the first training data set; or, According to the model interpretability requirement, obtaining the second prediction model; or, The second prediction model is obtained according to the category distribution information corresponding to the first training data set.

5. The method according to claim 1, characterized in that: Before the first prediction model is used to perform recognition processing on the data set to obtain the state recognition result corresponding to the grayscale switch, the method further includes: According to the business requirements, feature engineering is performed on the data set to obtain a feature set corresponding to the data set; According to the label generation rule, label identification is performed on each data in the data set to obtain label information corresponding to each data.

6. The method according to claim 1, characterized in that The adopting the first prediction model to perform recognition processing on the data set to obtain a state recognition result corresponding to the grayscale switch includes: According to the label information and the attribute information, the data set is preprocessed to obtain a preprocessed data set, wherein the data preprocessing includes at least one of cleaning data processing, standardizing data processing, and normalizing data processing; When the format information corresponding to the preprocessed data set does not meet the format requirement, performing format conversion processing on the cleaned data set to obtain a format-converted data set; The first prediction model is used to perform recognition processing on the data set after the format conversion processing to obtain a state recognition result corresponding to the grayscale switch.

7. The method according to claim 1, characterized in that The method further comprises: Updating the first prediction model once every preset time period; or, When the second training data set is obtained, the first prediction model is updated.

8. A grayscale switch control device, characterized in that: include: A set acquisition unit, used to acquire a data set corresponding to a target node set according to business demand information, wherein the data set includes a user behavior data subset and a system performance data subset; A result acquisition unit, used to perform recognition processing on the data set using a first prediction model to obtain a state recognition result corresponding to the grayscale switch; A switch control unit is used to control the on / off state of the grayscale switch according to the state recognition result.

9. A network device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the grayscale switch control method as described in any one of claims 1 to 7.

10. A storage medium, when instructions in the storage medium are executed by a processor of a network device, the network device is enabled to execute the grayscale switch control method according to any one of claims 1 to 7.

Images