Data management method based on artificial intelligence

Abstract

The present invention discloses a data management method based on artificial intelligence, which belongs to the field of data management technology. An artificial intelligence model is used to construct a data classification model, and sample data corresponding to each data type is used to train the data classification model. The trained data classification model is used to identify the target data to obtain the data type corresponding to the target data. Finally, a data operation method set by a staff is executed according to the data type corresponding to the target data, thereby realizing intelligent data management, which can not only effectively improve the data management efficiency, but also improve the data management accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of data management, and in particular relates to a data management method based on artificial intelligence. Background Art

[0002] Data management refers to the effective organization, storage, maintenance, retrieval, and use of data resources to ensure data quality, availability, security, and compliance. With the advent of the big data era, the amount of data faced by enterprises is exploding. Existing technologies typically store collected data directly or analyze it before storing it. This often involves unmanaged data or manual data management. This disorganized data storage and management process leads to inefficiencies and poor accuracy when processing massive amounts of data. Furthermore, existing data management methods lack intelligence and are unable to provide personalized data services to users. Summary of the Invention

[0003] The present invention provides an artificial intelligence-based data management method to solve the problems of low data management efficiency, low accuracy and inability to provide personalized services in the existing technology.

[0004] A data management method based on artificial intelligence, comprising:

[0005] Obtain the data operation method corresponding to each data type input by the staff, and obtain the sample data corresponding to each data type;

[0006] An artificial intelligence model is used to build a data classification model, and the data classification model is trained using sample data corresponding to each data type to obtain a trained data classification model;

[0007] Collect target data generated in real time, and use the trained data classification model to identify the target data to obtain the data type corresponding to the target data;

[0008] Based on the data operation method corresponding to the data type corresponding to the target data, the target data is managed to complete the data management of artificial intelligence.

[0009] Furthermore, the data operation method corresponding to each data type input by the staff is obtained, including: discarding data, storing data in a designated area, or issuing an early warning for the data.

[0010] Furthermore, an artificial intelligence model is used to construct a data classification model, including: using a convolutional neural network model to construct a data classification model.

[0011] Furthermore, the data classification model is trained using sample data corresponding to each data type to obtain a trained data classification model, including:

[0012] For each data type, assign different outputs of the data classification model to the data type, determine the output category of each data type, and set the expected label of the sample data corresponding to each data type to the corresponding output category;

[0013] Based on the parameters of the data classification model, multiple different parameter vectors are generated to complete the training initialization;

[0014] Take the sample data as the actual input and the output category corresponding to the sample data as the expected output, and obtain the loss function value corresponding to each parameter vector;

[0015] According to the loss function value corresponding to each parameter vector, the optimal vector and the historical optimal value corresponding to each parameter vector are determined;

[0016] Perform a sinusoidal local search on the parameter vector to obtain a parameter vector after the sinusoidal local search;

[0017] Perform boundary information fusion search on the parameter vector after the sinusoidal local search to obtain the parameter vector after the boundary information fusion search;

[0018] According to the optimal vector and the historical optimal value corresponding to each parameter vector, the parameter vector after the boundary information fusion search is subjected to an adaptive speed change information fusion search to obtain the parameter vector after the adaptive speed change information fusion search;

[0019] Performing a global random flight search on the parameter vector after the adaptive speed change information fusion search to obtain the parameter vector after the global random flight search;

[0020] Sine local search, boundary information fusion search, adaptive variable speed information fusion search and global random flight search are repeatedly performed until the maximum number of training times is reached, the optimal vector is obtained again, and the optimal vector is decoded into the final parameters of the data classification model to obtain the trained data classification model.

[0021] Furthermore, a sinusoidal local search is performed on the parameter vector to obtain a parameter vector after the sinusoidal local search, including:

[0022] Generate a first random number between (0, 1), and obtain a search factor using a sine function based on the first random number;

[0023] For any parameter vector, randomly match the parameter vector with another parameter vector to obtain the first target parameter vector corresponding to each parameter vector;

[0024] Obtaining a first difference vector between the parameter vector and its corresponding first target parameter vector, and processing the first difference vector using a search factor to obtain a processed first difference vector;

[0025] The parameter vector is updated according to the processed first difference vector to obtain the parameter vector after the sinusoidal local search.

[0026] Furthermore, a boundary information fusion search is performed on the parameter vector after the sinusoidal local search to obtain a parameter vector after the boundary information fusion search, including:

[0027] Obtain a random information vector according to the upper bound vector and the lower bound vector corresponding to the parameter vector;

[0028] For any parameter vector after the sinusoidal local search, based on the random information vector, and using the parameter vector and reference vector in the previous training process, obtain a reference vector in the current training process;

[0029] For any parameter vector after the sinusoidal local search, a second difference vector between the parameter vector and the optimal vector is obtained, and the second difference vector is processed using a second random number between (0, 1) to obtain a processed second difference vector;

[0030] Obtaining a third difference vector between the parameter vector and the reference vector, and processing the third difference vector using a third random number between (0, 1) to obtain a processed third difference vector;

[0031] The parameter vector after the sinusoidal local search is updated according to the processed second difference vector and the processed third difference vector to obtain the parameter vector after the boundary information fusion search.

[0032] Furthermore, based on the optimal vector and the historical optimal value corresponding to each parameter vector, an adaptive speed change information fusion search is performed on the parameter vector after the boundary information fusion search to obtain the parameter vector after the adaptive speed change information fusion search, including:

[0033] Based on the current number of training times, the adaptive speed change weight is obtained;

[0034] For the parameter vector after the boundary information fusion search, the adaptive speed change weight is used to weight the historical search speed of the adaptive speed change information fusion search in the previous training process to obtain the weighted historical search speed;

[0035] Obtaining a social learning information vector between the parameter vector and the optimal vector, and processing the social learning information vector using a social learning factor to obtain a processed social learning information vector;

[0036] Obtain the individual learning information between the parameter vector and its historical optimal value, and process the individual learning information using the individual learning factor to obtain the processed individual learning information;

[0037] Obtain the real-time search speed during the current training process based on the weighted historical search speed, the processed social learning information vector, and the processed individual learning information volume;

[0038] According to the real-time search speed in the current training process, the parameter vector after the boundary information fusion search is updated to obtain the parameter vector after the adaptive speed change information fusion search.

[0039] Furthermore, a global random flight search is performed on the parameter vector after the adaptive speed change information fusion search to obtain the parameter vector after the global random flight search, including:

[0040] Determine a global search vector for the parameter vector after the adaptive speed change information fusion search;

[0041] Use Lévy flight to generate random Lévy flight step length;

[0042] For the vector to be globally searched, randomly match another parameter vector to obtain a second target parameter vector corresponding to each vector to be globally searched;

[0043] Obtaining a fourth difference vector between the global search vector and its corresponding second target parameter vector, and processing the fourth difference vector using a Levy flight step size and a preset search control factor to obtain a processed fourth difference vector;

[0044] For the vector to be globally searched, a fourth random number between (0, 1) is used to process the vector to be globally searched to obtain the processed vector to be globally searched;

[0045] The optimal vector is processed using a fifth random number between (0, 1) to obtain the optimal vector after processing;

[0046] The global search vector is updated according to the processed fourth difference vector, the processed vector to be globally searched, and the processed optimal vector to be globally searched, to obtain a parameter vector after the global random flight search.

[0047] Furthermore, the target data generated in real time is collected and the trained data classification model is used to identify the target data to obtain the data type corresponding to the target data, including:

[0048] Collecting target data generated in real time, and converting the target data into data in the same format as the sample data to obtain converted target data;

[0049] The converted target data is constructed as the input data of the trained data classification model, and is input into the trained data classification model to obtain the data type corresponding to the target data.

[0050] Furthermore, the target data is managed based on the data operation method corresponding to the data type corresponding to the target data, including:

[0051] According to the data type corresponding to the target data, determine the data operation method corresponding to the target data from the data operation methods corresponding to each data type input by the staff;

[0052] Execute the data operation method corresponding to the target data to manage the target data.

[0053] The present invention provides an artificial intelligence-based data management method, which constructs a data classification model by adopting an artificial intelligence model, and trains the data classification model with sample data corresponding to each data type, and uses the trained data classification model to identify the target data to obtain the data type corresponding to the target data. Finally, the data operation method set by the staff is executed according to the data type corresponding to the target data, thereby realizing intelligent data management, which can not only effectively improve the data management efficiency, but also improve the data management accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0055] Figure 1 A flowchart of an artificial intelligence-based data management method provided in an embodiment of the present invention.

[0056] The above drawings illustrate specific embodiments of the present invention, which will be described in more detail below. These drawings and the accompanying description are not intended to limit the scope of the present invention in any way, but rather to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0057] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0058] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0059] like Figure 1 As shown, an embodiment of the present invention provides a data management method based on artificial intelligence, including:

[0060] S101. Obtain the data operation method corresponding to each data type input by the staff, and simultaneously obtain the sample data corresponding to each data type;

[0061] Data manipulation methods can include data modification, data deletion, data storage, data warning, and other data-related operations, which are set by staff. For example, when processing motor operation data, when the motor is operating normally, its data is useless except for monitoring, so the motor data can be deleted or stored. However, when the motor is operating abnormally, the data can be warned and stored.

[0062] In the embodiment of the present invention, the staff can customize the data operation method, which can effectively realize the customized management of data and realize the intelligentization of data management.

[0063] S102. Using an artificial intelligence model to construct a data classification model, and using sample data corresponding to each data type to train the data classification model to obtain a trained data classification model;

[0064] Artificial intelligence models such as convolutional neural networks can be used to build data classification models, enabling rapid data identification. By training the data classification model with sample data corresponding to each data type, the data classification model can identify data relationships between data, thereby acquiring data classification capabilities and ultimately completing intelligent data management.

[0065] S103, collecting target data generated in real time, and using the trained data classification model to identify the target data to obtain the data type corresponding to the target data;

[0066] It is worth noting that in order to ensure that the data classification model after training can accurately identify the data, the data type and data structure between the target data and the sample data should be the same, and the length of data of the same type should be the same. If they are different, zeros should be added to the front of the data to ensure normal recognition of the data.

[0067] S104. Manage the target data based on the data operation method corresponding to the data type corresponding to the target data, and complete the data management of artificial intelligence.

[0068] The embodiment of the present invention not only realizes intelligent data management but also improves the standardization and accuracy of data management by executing the data operation method set by the staff.

[0069] In this embodiment of the present invention, the data operation method corresponding to each data type input by the staff member includes: discarding the data, storing the data in a designated area, or issuing an early warning for the data. It is worth noting that the above data operation method is only a preferred method of this embodiment of the present invention, and other data operation methods can also be used to achieve more comprehensive data management.

[0070] In an embodiment of the present invention, an artificial intelligence model is used to construct a data classification model, including: constructing a data classification model using a convolutional neural network model. It is worth noting that the convolutional neural network model is only a preferred method in an embodiment of the present invention. An artificial intelligence model can also be used to construct a data classification model to achieve the same data classification effect.

[0071] In an embodiment of the present invention, the data classification model is trained using sample data corresponding to each data type to obtain a trained data classification model, including:

[0072] For each data type, assign different outputs of the data classification model to the data type, determine the output category of each data type, and set the expected label of the sample data corresponding to each data type to the corresponding output category;

[0073] Based on the parameters of the data classification model, multiple different parameter vectors are generated to complete the training initialization; for example, initialization can be performed by random initialization between an upper limit and a lower limit.

[0074] For example, based on processing the operating parameters of the motor, assuming that the operating parameters of the motor are current, voltage, speed, vibration frequency, etc., then the data type can be normal type and abnormal type. At this time, the data classification model has two outputs. The first output category can be determined as the output category of the normal type, and the other output category can be determined as the output category of the abnormal type. Then, for any sample data, the expected label can be constructed as [1,0] or [0,1]. [1,0] indicates that the sample data is normal type data, and [0,1] indicates that the sample data is abnormal type data. It is worth noting that the above label setting process is only an example of an embodiment of the present invention. The number of abnormal types can be set according to actual needs, so as to achieve more refined data classification and data operation, and improve data management capabilities.

[0075] Take the sample data as the actual input and the output category corresponding to the sample data as the expected output, and obtain the loss function value corresponding to each parameter vector;

[0076] According to the loss function value corresponding to each parameter vector, the optimal vector and the historical optimal value corresponding to each parameter vector are determined;

[0077] Perform a sinusoidal local search on the parameter vector to obtain a parameter vector after the sinusoidal local search;

[0078] Perform boundary information fusion search on the parameter vector after the sinusoidal local search to obtain the parameter vector after the boundary information fusion search;

[0079] According to the optimal vector and the historical optimal value corresponding to each parameter vector, the parameter vector after the boundary information fusion search is subjected to an adaptive speed change information fusion search to obtain the parameter vector after the adaptive speed change information fusion search;

[0080] Performing a global random flight search on the parameter vector after the adaptive speed change information fusion search to obtain the parameter vector after the global random flight search;

[0081] Sine local search, boundary information fusion search, adaptive variable speed information fusion search and global random flight search are repeatedly performed until the maximum number of training times is reached, the optimal vector is obtained again, and the optimal vector is decoded into the final parameters of the data classification model to obtain the trained data classification model.

[0082] In existing technologies, gradient descent optimization algorithms are often used to train hyperparameters (such as weights) of data classification models. However, this algorithm can easily fall into local optima, leading to premature convergence and poor data classification results. Therefore, embodiments of the present invention provide a new intelligent optimization algorithm to enhance global search capabilities and ultimately improve data management accuracy.

[0083] Optionally, after each search, the parameter vector may be subjected to out-of-bounds processing to ensure that the parameter is always within a valid range.

[0084] In an embodiment of the present invention, performing a sinusoidal local search on a parameter vector to obtain a parameter vector after the sinusoidal local search includes:

[0085] Generate a first random number r1 between (0, 1), and based on the first random number r1, use the sine function to obtain a search factor: sin(2πr1)*(r6+1); where sin represents the sine function, π represents pi, and r6 represents the sixth random number between (0, 1);

[0086] For any parameter vector, randomly match the parameter vector with another parameter vector to obtain the first target parameter vector corresponding to each parameter vector;

[0087] Obtain a first difference vector between the parameter vector and its corresponding first target parameter vector, and process the first difference vector using a search factor to obtain the processed first difference vector: Among them, Δ1 represents the first difference vector, represents the i-th parameter vector in the t-th training process, i=1,2,…,I, I represents the total number of parameter vectors, Represents the parameter vector The corresponding first target parameter vector, r6 represents the sixth random number between (0,1);

[0088] According to the first difference vector after processing, the parameter vector is updated, and the parameter vector after the sine local search is obtained as follows: in, represents the parameter vector after the sinusoidal local search

[0089] The sinusoidal local search provided by the embodiment of the present invention can perform local search on the parameter vector using a sinusoidal waveform, thereby improving the efficiency of local development. In the process of local development, it is also possible to search for the global optimal area.

[0090] In an embodiment of the present invention, a boundary information fusion search is performed on the parameter vector after the sinusoidal local search to obtain the parameter vector after the boundary information fusion search, including:

[0091] According to the upper bound vector and lower bound vector corresponding to the parameter vector, the random information vector is obtained as follows: α*cos(θ)*((ub-lb)*r7+lb; where ub represents the upper bound vector, i.e., the vector composed of the upper limit of the parameters of each dimension; lb represents the lower bound vector, i.e., the vector composed of the lower limit of the parameters of each dimension; cos represents the cosine function, θ represents the random degree between (0, 2π), r7 represents the seventh random number between (0, 1); α represents the adaptive adjustment parameter, and α=(1-t / T) 2t / T , T represents the preset maximum number of training times;

[0092] For any parameter vector after the sinusoidal local search, based on the random information vector and using the parameter vector and reference vector in the previous training process, the reference vector in the current training process is obtained as follows: in, Represents the reference vector corresponding to the mth parameter vector in the last training process, Represents the reference vector corresponding to the mth parameter vector in the current training process, m = 1, 2, ..., I;

[0093] For any parameter vector after the sinusoidal local search, obtain the second difference vector between the parameter vector and the optimal vector, and use the second random number r2 between (0,1) to process the second difference vector. The second difference vector after processing is: in, represents the optimal vector, Δ2 represents the second difference vector;

[0094] Obtain a third difference vector between the parameter vector and the reference vector, and process the third difference vector using a third random number r3 between (0, 1). The processed third difference vector is: Wherein, Δ3 represents the third difference vector;

[0095] According to the processed second difference vector and the processed third difference vector, the parameter vector after the sine local search is updated, and the parameter vector after the boundary information fusion search is obtained as follows: in, Represents the parameter vector after boundary information fusion search represents the parameter vector after the mth sinusoidal local search during the tth training process.

[0096] The boundary information fusion search provided by the embodiment of the present invention can perform random search with boundary information, and combine with reference vectors to realize memory search with random search, and further integrate the search of optimal vectors, which not only increases the diversity of the algorithm and improves the global search capability, but also effectively guarantees the search speed of the algorithm.

[0097] In an embodiment of the present invention, based on the optimal vector and the historical optimal value corresponding to each parameter vector, an adaptive speed change information fusion search is performed on the parameter vector after the boundary information fusion search to obtain the parameter vector after the adaptive speed change information fusion search, including:

[0098] Based on the current number of training times, the adaptive speed change weight is obtained as: Among them, ω represents the adaptive speed change weight, ω max Represents the maximum value of the adaptive speed change weight, ω min represents the minimum value of the adaptive speed change weight, r8 represents the eighth random number between (0,1), b represents the intermediate parameter, and b=ln(ω max / ω min );

[0099] In the early stages of the search, the probability of the adaptive speed weight taking a larger value increases, allowing the parameter vector to search the entire space at a higher speed and determine the range of the global optimal solution, thus ensuring the population's global search capability. In the later stages of the search, the probability of the adaptive speed weight taking a smaller value increases, allowing the parameter vector to achieve a lower speed, thereby enabling a more precise search of the optimal solution's location neighborhood and ensuring the population's local development capabilities. This balances the algorithm's search performance. Furthermore, during the search process, the adaptive speed weight value dynamically adjusts with the number of iterations, continuously correcting the speed and position of the parameter vector and strengthening its interaction. This helps the parameter vector quickly find the optimal solution and prevents the algorithm from falling into a local optimum. Furthermore, the dynamic changes in the adaptive speed weight satisfy the nonlinear and complex nature of the actual search process.

[0100] For the parameter vector after the boundary information fusion search, the adaptive speed change weight is used to weight the historical search speed of the adaptive speed change information fusion search in the previous training process. The weighted historical search speed is: in, Indicates the speed corresponding to the parameter vector after the nth boundary information fusion search during the tth training process;

[0101] Obtain the social learning information vector between the parameter vector and the optimal vector, and use the social learning factor to process the social learning information vector. The social learning information vector after processing is: Among them, c1 represents the social learning factor, r9 represents the ninth random number between (0,1), represents the optimal parameter vector, represents the parameter vector after the nth boundary information fusion search in the tth training process, n = 1, 2, ..., I;

[0102] The amount of individual learning information between the parameter vector and its historical optimal value is obtained, and the amount of individual learning information is processed using the individual learning factor. The individual learning information after processing is obtained as follows: Among them, c2 represents the individual learning factor, r 10 Represents the tenth random number between (0,1), Represents the historical optimal value of the parameter vector corresponding to the nth boundary information fusion search in the tth training process;

[0103] According to the weighted historical search speed, the processed social learning information vector, and the processed individual learning information volume, the real-time search speed in the current training process is obtained as follows: in, Indicates the real-time search speed of the parameter vector after the nth boundary information fusion search in the current training process;

[0104] According to the real-time search speed in the current training process, the parameter vector after the boundary information fusion search is updated, and the parameter vector after the adaptive speed change information fusion search is obtained as follows: in, Represents the parameter vector after adaptive speed change information fusion search

[0105] The adaptive speed-varying information fusion search provided by the embodiment of the present invention can effectively improve the search speed and search quality of the algorithm and provide higher precision in the later stage of the algorithm.

[0106] In an embodiment of the present invention, a global random flight search is performed on the parameter vector after the adaptive speed change information fusion search to obtain the parameter vector after the global random flight search, including:

[0107] For the parameter vector after the adaptive speed change information fusion search, determine the vector to be globally searched; for example, determine the K parameter vectors with the largest loss function value as the vector to be globally searched;

[0108] Use Levy flight to generate random Levy flight step length levy;

[0109] For the vector to be globally searched, randomly match another parameter vector to obtain a second target parameter vector corresponding to each vector to be globally searched;

[0110] The fourth difference vector between the global search vector and its corresponding second target parameter vector is obtained as: Wherein, Δ4 represents the fourth difference vector, Indicates the kth vector to be globally searched during the tth training process, k=1,2,…,K, Represents the vector to be globally searched The corresponding second target parameter vector;

[0111] The fourth difference vector is processed using the Levy flight step length and the preset search control factor, and the processed fourth difference vector is obtained as follows: Δ4*levy*γ; wherein γ represents the preset search control factor;

[0112] For the vector to be globally searched, a fourth random number r4 between (0, 1) is used to process the vector to be globally searched, and the processed vector to be globally searched is:

[0113] The optimal vector is processed using the fifth random number r5 between (0,1), and the optimal vector after processing is obtained as follows:

[0114] According to the processed fourth difference vector, the processed global search vector and the processed optimal vector, the global search vector is updated to obtain the parameter vector after the global random flight search: in, represents the parameter vector after global random flight search

[0115] The global random flight search provided by the embodiment of the present invention can not only effectively help the algorithm escape from the local optimum, but also does not affect the algorithm search speed.

[0116] Optionally, an annealing simulation algorithm may be used to control the global random flight search, or a greedy algorithm may be used to control the global random flight search, thereby further improving the algorithm search speed.

[0117] In an embodiment of the present invention, target data generated in real time is collected and the trained data classification model is used to identify the target data to obtain the data type corresponding to the target data, including:

[0118] Collecting target data generated in real time, and converting the target data into data in the same format as the sample data to obtain converted target data;

[0119] The converted target data is constructed as the input data of the trained data classification model, and is input into the trained data classification model to obtain the data type corresponding to the target data.

[0120] In an embodiment of the present invention, the target data is managed based on the data operation method corresponding to the data type corresponding to the target data, including:

[0121] According to the data type corresponding to the target data, determine the data operation method corresponding to the target data from the data operation methods corresponding to each data type input by the staff;

[0122] Execute the data operation method corresponding to the target data to manage the target data.

[0123] The present invention provides an artificial intelligence-based data management method, which constructs a data classification model by adopting an artificial intelligence model, and trains the data classification model with sample data corresponding to each data type, and uses the trained data classification model to identify the target data to obtain the data type corresponding to the target data. Finally, the data operation method set by the staff is executed according to the data type corresponding to the target data, thereby realizing intelligent data management, which can not only effectively improve the data management efficiency, but also improve the data management accuracy.

[0124] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and variations can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A data management method based on artificial intelligence, characterized in that: include: Obtain the data operation method corresponding to each data type input by the staff, and obtain the sample data corresponding to each data type; An artificial intelligence model is used to build a data classification model, and the data classification model is trained using sample data corresponding to each data type. The trained data classification model includes: Collect target data generated in real time, and use the trained data classification model to identify the target data to obtain the data type corresponding to the target data; Based on the data operation method corresponding to the data type corresponding to the target data, the target data is managed to complete the data management of artificial intelligence; According to the loss function value corresponding to each parameter vector, the optimal vector and the historical optimal value corresponding to each parameter vector are determined; According to the optimal vector and the historical optimal value corresponding to each parameter vector, the parameter vector after the boundary information fusion search is subjected to an adaptive speed change information fusion search to obtain the parameter vector after the adaptive speed change information fusion search, including: Based on the current number of training times, the adaptive speed change weight ω is obtained as: Among them, ω max Represents the maximum value of the adaptive speed change weight, ω min represents the minimum value of the adaptive speed change weight, r8 represents the eighth random number between (0,1), t represents the current training number, T represents the maximum training number, b represents the intermediate parameter, and b=ln(ω max / ω min ); Parameter vector after fusion search for the nth boundary information , the adaptive speed change weight ω is used to adjust the historical search speed of the adaptive speed change information fusion search in the previous training process Perform weighting to obtain the weighted historical search speed ; Get parameter vector With the optimal vector The social learning information vector between ;in, represents the social learning factor, Represents the ninth random number between (0,1); Get parameter vector Its historical optimal value The amount of individual learning information between ;in, represents the individual learning factor, Represents the tenth random number between (0,1); According to the weighted historical search speed, the processed social learning information vector and the processed individual learning information, the real-time search speed in the current training process is obtained. for: According to the real-time search speed in the current training process, the parameter vector after the boundary information fusion search is updated to obtain the parameter vector after the adaptive speed change information fusion search; Determine a global search vector for the parameter vector after the adaptive speed change information fusion search; For the vector to be globally searched, randomly match another parameter vector to obtain a second target parameter vector corresponding to each vector to be globally searched; Obtaining a fourth difference vector between the global search vector and its corresponding second target parameter vector, and processing the fourth difference vector using a Levy flight step length and a preset search control factor; For the global search vector to be processed, a fourth random number r4 between (0,1) is used to process the global search vector; The optimal vector is processed using a fifth random number r5 between (0,1). The global search vector is updated according to the processed fourth difference vector, the processed global search vector and the processed optimal vector to obtain a parameter vector after global random flight search.

2. The data management method based on artificial intelligence according to claim 1, characterized in that: Obtain the data operation method corresponding to each data type entered by the staff, including: discarding data, storing data in a designated area, or issuing an early warning for the data.

3. The data management method based on artificial intelligence according to claim 1, characterized in that: Adopting artificial intelligence models to build data classification models, including: adopting convolutional neural network models to build data classification models.

4. The data management method based on artificial intelligence according to claim 1, characterized in that: Collect target data generated in real time and use the trained data classification model to identify the target data to obtain the data type corresponding to the target data, including: Collecting target data generated in real time, and converting the target data into data in the same format as the sample data to obtain converted target data; The converted target data is constructed as the input data of the trained data classification model, and is input into the trained data classification model to obtain the data type corresponding to the target data.

5. The artificial intelligence-based data management method according to claim 4, characterized in that: The target data is managed based on the data operation method corresponding to the data type corresponding to the target data, including: According to the data type corresponding to the target data, determine the data operation method corresponding to the target data from the data operation methods corresponding to each data type input by the staff; Execute the data operation method corresponding to the target data to manage the target data.

Images