Enterprise platform operation and maintenance method and system based on machine learning

Abstract

The application discloses an enterprise platform operation and maintenance method and system based on machine learning, belongs to the technical field of enterprise platform operation and maintenance and data processing, and collects operation and maintenance data of an enterprise platform first, learns the collected operation and maintenance data of the enterprise platform through a machine learning technology, so that a rule for analyzing operation and maintenance data of the enterprise platform can be formed, and then the rule is relied on to analyze subsequent operation and maintenance data of the enterprise platform, so that real-time monitoring of the operation state of the enterprise platform, fault prediction, performance optimization and intelligent decision-making can be realized, the operation and maintenance efficiency of the enterprise platform is effectively improved, operation and maintenance costs are reduced, and stable operation of the platform is ensured.

Description

Technical Field

[0001] This invention belongs to the field of enterprise platform operation and maintenance technology and data processing technology, specifically relating to an enterprise platform operation and maintenance method and system based on machine learning. Background Technology

[0002] Enterprise IT operations and maintenance (O&M), also known as enterprise IT operations, refers to a series of management tasks and operational activities carried out by an enterprise's internal IT department to ensure the normal operation of IT infrastructure and services. The goal of enterprise O&M is to guarantee the stability, reliability, and security of IT systems, while optimizing resource utilization and improving service quality and response speed. With the rapid development of enterprise informatization, enterprise platform O&M faces many challenges, such as increased system complexity, frequent failures, and insufficient O&M personnel. Traditional O&M methods mainly rely on manual experience, which is difficult to cope with large-scale, high-concurrency platform environments. Therefore, researching an efficient and intelligent O&M method is of great significance. Summary of the Invention

[0003] This invention provides an enterprise platform operation and maintenance method and system based on machine learning to solve the problems existing in the prior art.

[0004] On the one hand, this invention provides a machine learning-based enterprise platform operation and maintenance method, including:

[0005] The system receives historical enterprise platform operation and maintenance data and corresponding historical fault types from human-computer interaction input to obtain historical sample data. After preprocessing the historical sample data, it obtains preprocessed historical sample data.

[0006] Select at least one machine learning model from the machine learning model library, and train the selected machine learning model using preprocessed historical sample data to obtain at least one trained machine learning model.

[0007] Based on at least one trained machine learning model, determine the final machine learning model and deploy the final machine learning model;

[0008] Real-time enterprise platform operation and maintenance data is collected from the enterprise platform, and the deployed final machine learning model is used to identify the real-time enterprise platform operation and maintenance data to determine the real-time fault type; wherein, the real-time fault type includes no fault or specific fault type;

[0009] When the real-time fault type is a specific fault type, a warning message of enterprise platform operation and maintenance abnormality is generated and transmitted to the device designated by the staff to complete the enterprise platform operation and maintenance based on machine learning.

[0010] Furthermore, pre-set enterprise platform operation and maintenance parameter adjustment strategies for each specific fault type;

[0011] When the real-time fault type is a specific fault type, the target enterprise platform operation and maintenance parameter adjustment strategy corresponding to the specific fault type is determined by search and matching, and the target enterprise platform operation and maintenance parameter adjustment strategy is executed to realize the automatic operation and adjustment of the enterprise platform.

[0012] Further, after preprocessing the historical sample data, preprocessed historical sample data is obtained, including:

[0013] Remove historical sample data containing missing or outlier values ​​to obtain historical sample data after initial processing;

[0014] The non-numerical data in the historical sample data after initial processing is converted into characters to convert all data into numerical data, resulting in the historical sample data after character conversion.

[0015] The historical sample data after character conversion is normalized to obtain normalized historical sample data, which is then used as the preprocessed historical sample data.

[0016] Furthermore, the machine learning model library is a pre-set database and includes at least LSTM machine learning models, CNN machine learning models, BP machine learning models, and SVM machine learning models.

[0017] Furthermore, at least one selected machine learning model is trained using preprocessed historical sample data to obtain at least one trained machine learning model, including:

[0018] For any machine learning model, initialize the model parameters of the machine learning model, and combine all the model parameters into a particle. Repeat this process to obtain particles multiple times to complete the initialization of the particle swarm.

[0019] The fitness of each particle is obtained using preprocessed historical sample data, and the optimal particle in the particle swarm is obtained based on the fitness of each particle.

[0020] An information interaction algorithm carrying an oscillation term is used to quickly update each particle in the particle swarm, resulting in a rapidly updated particle.

[0021] If the fitness of a particle after a rapid update is greater than the fitness of the best particle, then the particle with the greater fitness is replaced with the best particle.

[0022] Determine whether the total number of times the optimal particle has not been replaced has reached the preset training number threshold K. If so, perform a global optimal search on the optimal particle to obtain the updated optimal particle; otherwise, perform a global jump search on the optimal particle to obtain the updated optimal particle.

[0023] Determine if the current training iterations have reached the maximum number of training iterations. If so, based on the particles after the rapid update and the optimal particle after the update, redetermine the optimal particle and use the parameters in the redetermined optimal particle as the final parameters of the machine learning model to obtain the trained machine learning model. Otherwise, return to the rapid update step.

[0024] Iterate through all selected machine learning models to obtain at least one trained machine learning model.

[0025] Furthermore, the fitness of each particle is obtained using preprocessed historical sample data, and the optimal particle in the particle swarm is obtained based on the fitness of each particle, including:

[0026] For any given particle, the model parameters contained in the particle are applied to the machine learning model. The historical enterprise platform operation and maintenance data in the preprocessed historical sample data are used as the input of the machine learning model, and the historical fault types in the preprocessed historical sample data are used as the expected label values ​​to obtain the error function value of the machine learning model.

[0027] The fitness of a particle is obtained by adding the error function value of the machine learning model to a preset constant term and taking the reciprocal.

[0028] Based on the fitness of each particle, the particle with the highest fitness is determined as the optimal particle.

[0029] Furthermore, an information interaction algorithm carrying an oscillation term is used to quickly update each particle in the particle swarm, resulting in the following updated particles:

[0030] For any given particle, randomly match the particle with the closest Euclidean distance and the higher fitness to obtain the target following particle corresponding to the particle.

[0031] Obtain the Euclidean distance between the particle and its corresponding target following particle, and determine the oscillation term that adapts to the number of training iterations based on the current number of training iterations and the Euclidean distance.

[0032] After determining the difference term between the particle and the optimal particle, and using the first random coefficient and the interaction coefficient for limitation, the first update term corresponding to the particle is obtained;

[0033] After determining the difference term between the particle and the target following particle, and using the second random coefficient and the gravity coefficient for limitation, the second update term corresponding to the particle is obtained;

[0034] Based on the first update item and the second update item, the particles are quickly updated to obtain the particles after the quick update.

[0035] Furthermore, a global optimal search is performed on the optimal particle to obtain the updated optimal particle, including:

[0036] The third update term corresponding to the optimal particle is determined by using a random distribution vector and an exponential function that is affected by the current training iterations.

[0037] A historical memory strategy is used to remember the best particle and determine the fourth update term corresponding to the best particle.

[0038] Based on the third and fourth update terms, a global optimal search is performed on the optimal particle to obtain the optimal particle after the fast update.

[0039] Furthermore, a global jump search is performed on the optimal particle to obtain the updated optimal particle, including:

[0040] Among the particles other than the optimal particle, a first random particle is randomly selected from the 20% of particles with the highest fitness, and a second random particle is randomly selected from the other 80% of particles.

[0041] Based on the first random particle and the second random particle, a global jump search is performed on the optimal particle to obtain the updated optimal particle.

[0042] On the other hand, the present invention provides an enterprise platform operation and maintenance system based on machine learning, including: a data acquisition module, a model selection and training module, a model deployment module, a model application module, and a warning module;

[0043] The data acquisition module is used to receive historical enterprise platform operation and maintenance data and historical fault types corresponding to the historical enterprise platform operation and maintenance data input by human-computer interaction, obtain historical sample data, and obtain preprocessed historical sample data after preprocessing the historical sample data.

[0044] The model selection and training module is used to select at least one machine learning model from the machine learning model library and train the selected at least one machine learning model using preprocessed historical sample data to obtain at least one trained machine learning model.

[0045] The model deployment module is used to determine the final machine learning model based on at least one trained machine learning model, and to deploy the final machine learning model.

[0046] The model application module is used to collect real-time enterprise platform operation and maintenance data from the enterprise platform, and use the deployed final machine learning model to identify the real-time enterprise platform operation and maintenance data to determine the real-time fault type; wherein, the real-time fault type includes no fault or specific fault type;

[0047] The warning module is used to generate warning information of enterprise platform operation and maintenance abnormality when the real-time fault type is a specific fault type, and transmit the warning information to the device designated by the staff to complete the enterprise platform operation and maintenance based on machine learning.

[0048] This invention provides a machine learning-based enterprise platform operation and maintenance method and system. By first collecting enterprise platform operation and maintenance data, and then using machine learning technology to learn from the collected data, a rule for analyzing the enterprise platform operation and maintenance data can be formed. This rule is then used to analyze subsequent enterprise platform operation and maintenance data, thereby enabling real-time monitoring of the enterprise platform's operating status, fault prediction, performance optimization, and intelligent decision-making. This effectively improves the efficiency of enterprise platform operation and maintenance, reduces operation and maintenance costs, and ensures stable platform operation. Attached Figure Description

[0049] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0050] Figure 1 A flowchart illustrating an enterprise platform operation and maintenance method based on machine learning, provided as an embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram of an enterprise platform operation and maintenance system based on machine learning, provided as an embodiment of the present invention.

[0052] Among them, 201-Data Acquisition Module, 202-Model Selection and Training Module, 203-Model Deployment Module, 204-Model Application Module, and 205-Warning Module.

[0053] The accompanying drawings have illustrated specific embodiments of the invention, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0054] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0055] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0056] like Figure 1 As shown, this embodiment of the invention provides a machine learning-based enterprise platform operation and maintenance method, including:

[0057] S101. Receive historical enterprise platform operation and maintenance data and historical fault types corresponding to the historical enterprise platform operation and maintenance data from human-computer interaction input, obtain historical sample data, and preprocess the historical sample data to obtain preprocessed historical sample data.

[0058] Operational data can include data from specified locations in system logs, enterprise platform performance metrics, and user operation records, etc. It's important to note that the data length should be consistent across all data collection sessions to ensure the machine learning model can correctly identify the data. If data may have varying lengths, shorter data can be padded with zeros to ensure consistent length and that data at the same location belongs to the same data type. Other methods can also be used to ensure the accuracy of data length and location.

[0059] S102. Select at least one machine learning model from the machine learning model library, and train the selected machine learning model using preprocessed historical sample data to obtain at least one trained machine learning model.

[0060] One or more machine learning models can be set in the machine learning model library. Users can also select one or more machine learning models during the training process to determine the optimal machine learning model.

[0061] S103. Based on at least one trained machine learning model, determine the final machine learning model and deploy the final machine learning model;

[0062] Determining the final machine learning model based on at least one trained machine learning model may include: extracting some historical sample data and using the extracted historical sample data to test the performance metrics of at least one trained machine learning model, and determining the machine learning model with the best performance metrics as the final machine learning model, thereby ensuring the performance of the machine learning model.

[0063] S104. Collect real-time enterprise platform operation and maintenance data from the enterprise platform, and use the deployed final machine learning model to identify the real-time enterprise platform operation and maintenance data to determine the real-time fault type; wherein, the real-time fault type includes no fault or specific fault type;

[0064] The specific fault type can be any common type, and this embodiment of the invention does not impose any limitations. For example, faults such as enterprise platform attacks, security vulnerabilities, data leaks, backup failures, etc.

[0065] Optionally, real-time fault types can also include no fault or potential fault. When a potential fault is detected, a warning message is generated to enable staff to handle the anomaly, thereby ensuring the security of the enterprise platform.

[0066] S105. When the real-time fault type is a specific fault type, a warning message of enterprise platform operation and maintenance abnormality is generated, and the warning message is transmitted to the device designated by the staff to complete the enterprise platform operation and maintenance based on machine learning.

[0067] This invention provides a machine learning-based enterprise platform operation and maintenance method. By first collecting enterprise platform operation and maintenance data, and then using machine learning technology to learn from the collected data, a rule for analyzing the enterprise platform operation and maintenance data can be formed. This rule is then used to analyze subsequent enterprise platform operation and maintenance data, thereby enabling real-time monitoring of the enterprise platform's operating status, fault prediction, performance optimization, and intelligent decision-making. This effectively improves the efficiency of enterprise platform operation and maintenance, reduces operation and maintenance costs, and ensures stable platform operation.

[0068] In this embodiment of the invention, a pre-set enterprise platform operation and maintenance parameter adjustment strategy is provided for each specific fault type.

[0069] When the real-time fault type is a specific fault type, the target enterprise platform operation and maintenance parameter adjustment strategy corresponding to the specific fault type is determined by search and matching, and the target enterprise platform operation and maintenance parameter adjustment strategy is executed to realize the automatic operation and adjustment of the enterprise platform.

[0070] In this embodiment of the invention, after preprocessing the historical sample data, preprocessed historical sample data is obtained, including:

[0071] Remove historical sample data containing missing or outlier values ​​to obtain historical sample data after initial processing;

[0072] The non-numerical data in the historical sample data after initial processing is converted into characters to convert all data into numerical data, resulting in the historical sample data after character conversion.

[0073] The historical sample data after character conversion is normalized to obtain normalized historical sample data, which is then used as the preprocessed historical sample data.

[0074] In this embodiment of the invention, the machine learning model library is a pre-set database, and includes at least LSTM (Long Short-Term Memory) machine learning models, CNN (Convolutional Neural Network) machine learning models, BP (Back Propagation) machine learning models, and SVM (Support Vector Machine) machine learning models.

[0075] It is worth noting that you can also set up only one machine learning model, for example, use only a CNN machine learning model for operation and maintenance management, and train the weights and bias parameters of the CNN machine learning model to obtain the trained machine learning model.

[0076] In this embodiment of the invention, at least one selected machine learning model is trained using preprocessed historical sample data to obtain at least one trained machine learning model, including:

[0077] For any machine learning model, initialize the model parameters of the machine learning model, and combine all the model parameters into a particle. Repeat this process to obtain particles multiple times to complete the initialization of the particle swarm.

[0078] The fitness of each particle is obtained by using preprocessed historical sample data, and the best particle in the swarm is obtained based on the fitness of each particle. For example, the error function value (root mean square error or cross-entropy error) of the particle can be obtained first, and then the error function value can be added to a very small constant term (such as 0.001) and the reciprocal can be taken to obtain the fitness of the particle.

[0079] An information interaction algorithm carrying an oscillation term is used to quickly update each particle in the particle swarm, resulting in a rapidly updated particle.

[0080] If the fitness of a particle after a rapid update is greater than the fitness of the best particle, then the particle with the greater fitness is replaced with the best particle.

[0081] Determine whether the total number of times the optimal particle has not been replaced has reached the preset training number threshold K. If so, perform a global optimal search on the optimal particle to obtain the updated optimal particle; otherwise, perform a global jump search on the optimal particle to obtain the updated optimal particle.

[0082] Determine if the current training iterations have reached the maximum number of training iterations. If so, based on the particles after the rapid update and the optimal particle after the update, redetermine the optimal particle and use the parameters in the redetermined optimal particle as the final parameters of the machine learning model to obtain the trained machine learning model. Otherwise, return to the rapid update step.

[0083] Iterate through all selected machine learning models to obtain at least one trained machine learning model.

[0084] In existing technologies, particle swarm optimization (PSO) algorithms are often used to optimize the parameters of machine learning models. Although PSO algorithms have the advantage of fast training speed, they are prone to getting trapped in local optima. Therefore, this invention provides an improved PSO algorithm to solve the technical problems existing in existing PSO algorithms.

[0085] In this embodiment of the invention, the fitness of each particle is obtained using preprocessed historical sample data, and the optimal particle in the particle swarm is obtained based on the fitness of each particle, including:

[0086] For any given particle, the model parameters contained in the particle are applied to the machine learning model. The historical enterprise platform operation and maintenance data in the preprocessed historical sample data are used as the input of the machine learning model, and the historical fault types in the preprocessed historical sample data are used as the expected label values ​​to obtain the error function value of the machine learning model.

[0087] The fitness of a particle is obtained by adding the error function value of the machine learning model to a preset constant term and taking the reciprocal.

[0088] Based on the fitness of each particle, the particle with the highest fitness is determined as the optimal particle.

[0089] In this embodiment of the invention, an information interaction algorithm carrying an oscillation term is used to quickly update each particle in the particle swarm, resulting in the following particles after rapid update:

[0090] For any given particle, randomly match the particle with the closest Euclidean distance and the higher fitness to obtain the target following particle corresponding to the particle.

[0091] Obtain the Euclidean distance between a particle and its corresponding target following particle. Based on the current number of training iterations and the Euclidean distance, an oscillation term that adaptively changes with the number of training iterations is determined. in, Let i represent the i-th particle in the t-th training process, where i = 1, 2, ..., L, and L represents the total number of particles in the swarm. Represents particles The corresponding target follows the particle, and T represents the preset maximum number of training iterations. Let represent the first random vector, and The total dimension of the parameters is the same as the total dimension of the particle parameters. Each dimension of the number is a uniformly distributed random number between [-1, 1].

[0092] After determining the difference term between the particle and the optimal particle, and using the first random coefficient r1 and the interaction coefficient β as constraints, the first update term corresponding to the particle is obtained as follows: Where r1 represents the first random number between (0,1), and β represents the interaction coefficient between (0,2). Represents the optimal particle;

[0093] After determining the difference term between the particle and the target-following particle, and constraining it using the second random coefficient r2 and the gravitational coefficient α, the second update term corresponding to the particle is obtained as follows: Where r2 represents the second random number between (0,1), and α represents the gravitational coefficient between (0,2);

[0094] Based on the first update item and the second update item, the particles are quickly updated to obtain the particles after the quick update: in, Indicates the particles after rapid update

[0095] The information interaction algorithm with oscillation term provided in this embodiment of the invention can use information interaction to enable particles to quickly search with global and local optimal information. At the same time, by setting an adaptive oscillation term, it is not easy to get trapped in local optima in the early stage of the algorithm, and it has high search accuracy in the later stage of the algorithm. This solves the technical problem of being easily trapped in local optima in the prior art, and also ensures the optimization capability and optimization accuracy of the algorithm.

[0096] In this embodiment of the invention, a global optimal search is performed on the optimal particle to obtain the updated optimal particle, including:

[0097] Using a random distribution vector and an exponential function influenced by the current training iterations, the third update term corresponding to the optimal particle is determined as follows: in, Let e ​​represent the second random vector, and let e represent the natural constant. This represents the third update term corresponding to the optimal particle;

[0098] A historical memory strategy is used to remember the optimal particle, and the fourth update term corresponding to the optimal particle is determined as follows: Where r3 represents a third random number between (0,1);

[0099] Based on the third and fourth update terms, a global optimal search is performed on the optimal particle, resulting in the optimal particle after the fast update: in, Let represent the optimal particle during the (t-1)th training process.

[0100] The global optimal search for the optimal particle provided by the embodiments of the present invention allows the optimal particle to search in the known optimal region in the solution space, thereby ensuring the local optimization capability of the algorithm.

[0101] In this embodiment of the invention, a global jump search is performed on the optimal particle to obtain the updated optimal particle, including:

[0102] Among all particles except the optimal particle, a first random particle is randomly selected from the 20% of particles with the highest fitness. And randomly select a second random particle from the other 80% of the particles.

[0103] Based on the first and second random particles, a global jump search is performed on the optimal particle to obtain the updated optimal particle as follows: Where r4 represents the fourth random number between (0,1);

[0104] The global jump search for the optimal particle provided in this embodiment of the invention is only performed if the optimal particle has not been replaced multiple times. This proves that the population may be trapped in a local optimum. Therefore, by performing crossover operations on random particles, a global search can be achieved.

[0105] Optionally, during the global jump search for the optimal particle, random flying particles can be used to further enhance the randomness, such as... in This refers to particles generated by Levi's flight.

[0106] It's worth noting that while a greedy strategy can be used to control each search step to further accelerate convergence, excessive greed may lead the algorithm into local optima. Therefore, it's advisable to control only the global jump search. Furthermore, after each update, boundary-crossing checks can be performed on particles to ensure the algorithm's effectiveness.

[0107] like Figure 2 As shown, this embodiment of the invention provides an enterprise platform operation and maintenance system based on machine learning, including: a data acquisition module 201, a model selection and training module 202, a model deployment module 203, a model application module 204, and a warning module 205;

[0108] The data acquisition module 201 is used to receive historical enterprise platform operation and maintenance data and historical fault types corresponding to the historical enterprise platform operation and maintenance data input by human-computer interaction, obtain historical sample data, and obtain preprocessed historical sample data after preprocessing the historical sample data.

[0109] The model selection and training module 202 is used to select at least one machine learning model from the machine learning model library and train the selected at least one machine learning model using preprocessed historical sample data to obtain at least one trained machine learning model.

[0110] The model deployment module 203 is used to determine the final machine learning model based on at least one trained machine learning model, and to deploy the final machine learning model.

[0111] The model application module 204 is used to collect real-time enterprise platform operation and maintenance data from the enterprise platform, and use the deployed final machine learning model to identify the real-time enterprise platform operation and maintenance data to determine the real-time fault type; wherein, the real-time fault type includes no fault or specific fault type;

[0112] The warning module 205 is used to generate warning information of enterprise platform operation and maintenance abnormality when the real-time fault type is a specific fault type, and transmit the warning information to the device designated by the staff to complete the enterprise platform operation and maintenance based on machine learning.

[0113] The enterprise platform operation and maintenance system based on machine learning provided in this embodiment of the invention can execute the above-mentioned method and technical solution. Its principle and beneficial effects are similar, and will not be described again here.

[0114] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0115] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0118] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0119] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A machine learning-based enterprise platform operation and maintenance method, characterized in that, include: The system receives historical enterprise platform operation and maintenance data and corresponding historical fault types from human-computer interaction input to obtain historical sample data. After preprocessing the historical sample data, it obtains preprocessed historical sample data. Select at least one machine learning model from the machine learning model library, and train the selected machine learning model using preprocessed historical sample data to obtain at least one trained machine learning model. Based on at least one trained machine learning model, determine the final machine learning model and deploy the final machine learning model; Real-time enterprise platform operation and maintenance data is collected from the enterprise platform, and the deployed final machine learning model is used to identify the real-time enterprise platform operation and maintenance data to determine the real-time fault type; wherein, the real-time fault type includes no fault or specific fault type; When the real-time fault type is a specific fault type, a warning message of abnormal enterprise platform operation and maintenance is generated and transmitted to the device designated by the staff to complete the enterprise platform operation and maintenance based on machine learning. At least one selected machine learning model is trained using preprocessed historical sample data to obtain at least one trained machine learning model, including: For any machine learning model, initialize the model parameters of the machine learning model, and combine all the model parameters into a particle. Repeat this process to obtain particles multiple times to complete the initialization of the particle swarm. The fitness of each particle is obtained using preprocessed historical sample data, and the optimal particle in the particle swarm is obtained based on the fitness of each particle. An information interaction algorithm carrying an oscillation term is used to quickly update each particle in the particle swarm, resulting in a rapidly updated particle. If the fitness of a particle after a rapid update is greater than the fitness of the best particle, then the particle with the greater fitness is replaced with the best particle. Determine whether the total number of times the optimal particle has not been replaced has reached the preset training number threshold K. If so, perform a global optimal search on the optimal particle to obtain the updated optimal particle; otherwise, perform a global jump search on the optimal particle to obtain the updated optimal particle. Determine if the current training iterations have reached the maximum number of training iterations. If so, based on the particles after the rapid update and the optimal particle after the update, redetermine the optimal particle and use the parameters in the redetermined optimal particle as the final parameters of the machine learning model to obtain the trained machine learning model. Otherwise, return to the rapid update step. Iterate through all selected machine learning models to obtain at least one trained machine learning model.

2. The enterprise platform operation and maintenance method based on machine learning according to claim 1, characterized in that, Pre-configure enterprise platform operation and maintenance parameter adjustment strategies for each specific fault type; When the real-time fault type is a specific fault type, the target enterprise platform operation and maintenance parameter adjustment strategy corresponding to the specific fault type is determined by search and matching, and the target enterprise platform operation and maintenance parameter adjustment strategy is executed to realize the automatic operation and adjustment of the enterprise platform.

3. The enterprise platform operation and maintenance method based on machine learning according to claim 1, characterized in that, After preprocessing the historical sample data, preprocessed historical sample data is obtained, including: Remove historical sample data containing missing or outlier values ​​to obtain historical sample data after initial processing; The non-numerical data in the historical sample data after initial processing is converted into characters to convert all data into numerical data, resulting in the historical sample data after character conversion. The historical sample data after character conversion is normalized to obtain normalized historical sample data, which is then used as the preprocessed historical sample data.

4. The enterprise platform operation and maintenance method based on machine learning according to claim 1, characterized in that, The machine learning model library is a pre-set database and includes at least LSTM machine learning models, CNN machine learning models, BP machine learning models, and SVM machine learning models.

5. The enterprise platform operation and maintenance method based on machine learning according to claim 1, characterized in that, The fitness of each particle is obtained using preprocessed historical sample data, and the optimal particle in the swarm is obtained based on the fitness of each particle, including: For any given particle, the model parameters contained in the particle are applied to the machine learning model. The historical enterprise platform operation and maintenance data in the preprocessed historical sample data are used as the input of the machine learning model, and the historical fault types in the preprocessed historical sample data are used as the expected label values ​​to obtain the error function value of the machine learning model. The fitness of a particle is obtained by adding the error function value of the machine learning model to a preset constant term and taking the reciprocal. Based on the fitness of each particle, the particle with the highest fitness is determined as the optimal particle.

6. The enterprise platform operation and maintenance method based on machine learning according to claim 1, characterized in that, An information interaction algorithm carrying an oscillation term is used to quickly update each particle in the particle swarm, resulting in the following updated particles: For any given particle, randomly match the particle with the closest Euclidean distance and the higher fitness to obtain the target following particle corresponding to the particle. Obtain the Euclidean distance between the particle and its corresponding target following particle, and determine the oscillation term that adapts to the number of training iterations based on the current number of training iterations and the Euclidean distance. After determining the difference term between the particle and the optimal particle, and using the first random coefficient and the interaction coefficient for limitation, the first update term corresponding to the particle is obtained; After determining the difference term between the particle and the target following particle, and using the second random coefficient and the gravity coefficient for limitation, the second update term corresponding to the particle is obtained; Based on the first update item and the second update item, the particles are quickly updated to obtain the particles after the quick update.

7. The enterprise platform operation and maintenance method based on machine learning according to claim 6, characterized in that, A global optimal search is performed on the optimal particle to obtain the updated optimal particle, including: The third update term corresponding to the optimal particle is determined by using a random distribution vector and an exponential function that is affected by the current training iterations. A historical memory strategy is used to remember the best particle and determine the fourth update term corresponding to the best particle. Based on the third and fourth update terms, a global optimal search is performed on the optimal particle to obtain the optimal particle after the fast update.

8. The enterprise platform operation and maintenance method based on machine learning according to claim 7, characterized in that, A global jump search is performed on the optimal particle to obtain the updated optimal particle, including: Among the particles other than the optimal particle, a first random particle is randomly selected from the 20% of particles with the highest fitness, and a second random particle is randomly selected from the other 80% of particles. Based on the first random particle and the second random particle, a global jump search is performed on the optimal particle to obtain the updated optimal particle.

9. A machine learning-based enterprise platform operation and maintenance system, wherein the machine learning-based enterprise platform operation and maintenance system is used to execute the machine learning-based enterprise platform operation and maintenance method according to any one of claims 1 to 8, characterized in that, include: The module includes a data acquisition module, a model selection and training module, a model deployment module, a model application module, and a warning module. The data acquisition module is used to receive historical enterprise platform operation and maintenance data and historical fault types corresponding to the historical enterprise platform operation and maintenance data input by human-computer interaction, obtain historical sample data, and obtain preprocessed historical sample data after preprocessing the historical sample data. The model selection and training module is used to select at least one machine learning model from the machine learning model library and train the selected at least one machine learning model using preprocessed historical sample data to obtain at least one trained machine learning model. The model deployment module is used to determine the final machine learning model based on at least one trained machine learning model, and to deploy the final machine learning model. The model application module is used to collect real-time enterprise platform operation and maintenance data from the enterprise platform, and use the deployed final machine learning model to identify the real-time enterprise platform operation and maintenance data to determine the real-time fault type; wherein, the real-time fault type includes no fault or specific fault type; The warning module is used to generate warning information of enterprise platform operation and maintenance abnormality when the real-time fault type is a specific fault type, and transmit the warning information to the device designated by the staff to complete the enterprise platform operation and maintenance based on machine learning.

Images