Network data intelligent management system and method based on artificial intelligence

Abstract

The invention discloses a network data intelligent management system and method based on artificial intelligence, and relates to the technical field of network data management, original sensor data is projected to a sparse domain through an auto-encoder, whether the sparsity of the original sensor data meets a compressed sensing condition is verified, if yes, a Gaussian random matrix is constructed, and if not, a Gaussian random matrix is constructed; the Gaussian random matrix maps sparse signals to a low-dimensional space, a server side uses a basis tracking reconstruction algorithm to recover original sensor data according to the Gaussian random matrix, and after performance data in actual deployment are collected, the performance score of the deployment is calculated and obtained according to the evaluation model. And a corresponding optimization strategy is generated according to a performance score calculation result. According to the invention, in an application scene with limited bandwidth and limited equipment energy consumption, the data quality is ensured, the cost of data acquisition and transmission is significantly reduced, and the overall efficiency and intelligent level of network management are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of network data management, and in particular to an artificial intelligence-based network data intelligent management system and method. Background Art

[0002] With the widespread application of emerging technologies such as cloud computing, 5G technology, and the Internet of Things (IoT), the amount of data generated in the network has increased dramatically. How to efficiently manage this data has become a major challenge. The traditional manual management mode can no longer meet the needs of real-time and efficient data processing. The main purpose of the management system is to use artificial intelligence technology to intelligently manage, analyze and optimize network data, improve network performance, reduce management costs, and ensure efficient, secure and stable operation of the network.

[0003] The prior art has the following defects: When a sensor network is deployed in multiple areas of a city (for real-time monitoring of environmental parameters such as air quality, temperature, and humidity), the sensor nodes will periodically collect data and transmit the data to a central server for analysis and processing. However, due to limited network bandwidth and sensor power consumption, the existing management system's scheduled sampling and transmission method not only increases the bandwidth burden, but also leads to a large amount of unnecessary data redundancy.

[0004] Based on this, the present invention proposes an artificial intelligence-based network data intelligent management system and method, which can ensure data quality in application scenarios with limited bandwidth and limited equipment energy consumption, while significantly reducing the cost of data collection and transmission, thereby improving the overall efficiency and intelligence level of network management. Summary of the invention

[0005] The purpose of the present invention is to provide an artificial intelligence-based network data intelligent management system and method to solve the shortcomings of the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: an artificial intelligence-based network data intelligent management method, the management method comprising the following steps: The collection end uses AI algorithms to adjust the sampling frequency in real time according to network conditions and sensor sampling data. After collecting raw sensor data using a random sampling method, the raw sensor data is projected into a sparse domain through an autoencoder. Verify whether the sparsity of the original sensor data meets the compressed sensing conditions. If so, construct a Gaussian random matrix. The server uses the basis pursuit reconstruction algorithm to restore the original sensor data based on the Gaussian random matrix. After collecting the performance data from the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result.

[0007] In a preferred embodiment, after collecting the performance data in the actual deployment, the performance score of the deployment is calculated according to the evaluation model, and the corresponding optimization strategy is generated according to the performance score calculation result, including the following steps: Collect performance data from actual deployment, including compression ratio, reconstruction error, reconstruction time, and transmission efficiency. Substitute the compression ratio, reconstruction error, reconstruction time, and transmission efficiency into the evaluation model to calculate the performance score of this deployment. The obtained performance score is compared with the preset score threshold. The score threshold is used to evaluate whether the performance of this deployment meets the standard. If the performance score is greater than or equal to the score threshold, the performance of this deployment is evaluated to meet the standard. If the performance score is less than the score threshold, the performance of this deployment is evaluated to fail to meet the standard.

[0008] In a preferred embodiment, the compression rate, reconstruction error, reconstruction time and transmission efficiency are substituted into the evaluation model to calculate the performance score of this deployment, and the expression is: , where To rate performance, is the compression ratio, For transmission efficiency, is the reconstruction error, For the reconstruction time, are regression coefficients, and all regression coefficients are greater than 0.

[0009] In a preferred embodiment, verifying whether the sparsity of the original sensor data satisfies the compressed sensing condition comprises the following steps: After obtaining the non-zero element ratio value and norm ratio value of the original sensor data, the sparse coefficient of the element sensor data is calculated and the expression is: , where is the sparse coefficient, is the non-zero element ratio value, is the norm ratio, , is the adjustment coefficient, and , All are greater than 0; The obtained sparse coefficient is compared with the preset sparse threshold. If the sparse coefficient is greater than the sparse threshold, it is judged that the sparsity of the original sensor data does not meet the compressed sensing conditions. If the sparse coefficient is less than or equal to the sparse threshold, it is judged that the sparsity of the original sensor data meets the compressed sensing conditions.

[0010] In a preferred embodiment, the logic for obtaining the non-zero element ratio value is: obtaining the number of non-zero coefficients after the original sensor data is transformed, obtaining the total dimension of the original sensor data, and dividing the number of non-zero coefficients after the transformation by the total dimension to obtain the non-zero element ratio value; The logic for obtaining the norm ratio is: calculate the L1 norm, and the expression is: , calculate the L2 norm, the expression is: , the norm ratio is obtained by dividing the L1 norm by the L2 norm.

[0011] In a preferred embodiment, after collecting the raw sensor data by random sampling method, projecting the raw sensor data into a sparse domain by an autoencoder includes the following steps: The Bernoulli sampling method is used to sample the original sensor data. The signal obtained after sampling is a sparse representation of the original data. Suppose the original sensor data is: , then the sensor data after sampling is: ; The autoencoder includes an encoder and a decoder. The encoder maps the sampled sensor data y to a sparse domain representation z through a neural network, and the decoder reconstructs the original sensor data from the sparse domain representation z. , let the encoder of the autoencoder be: , the decoder is: ,but: ; ; During the autoencoder optimization process, the goal is to minimize the trade-off between reconstruction error and sparsity constraints. The objective function is: , where is the reconstruction error, is the sparsity constraint, is a regularization parameter that controls sparsity.

[0012] In a preferred embodiment, the collection end adjusts the sampling frequency in real time according to the network status and sensor sampling data through an AI algorithm, including the following steps: The acquisition end obtains network status data, including bandwidth occupancy rate, and obtains sensor sampling data, including signal change rate. The adjustment factor is calculated based on the bandwidth occupancy rate and signal change rate, and the sampling frequency is corrected by the obtained adjustment factor. The expression is: , where is the corrected sampling frequency, is the sampling frequency before correction, is the adjustment factor.

[0013] In a preferred embodiment, the calculation logic of the adjustment factor is: normalize the bandwidth occupancy rate and the signal change rate so that the value range of the bandwidth occupancy rate and the signal change rate is mapped to between [0,1], obtain the normalized value of the bandwidth occupancy rate and the normalized value of the signal change rate, and subtract the normalized value of the signal change rate from the normalized value of the bandwidth occupancy rate to obtain the adjustment factor.

[0014] In a preferred embodiment, the calculation logic of the bandwidth occupancy rate is: obtaining the currently used bandwidth of the network and the total network bandwidth, and dividing the total network bandwidth by the currently used bandwidth of the network to obtain the bandwidth occupancy rate; The calculation expression of the signal change rate is: , where is the signal change rate, is the number of sampling points, For sensor The sampling value of times.

[0015] An artificial intelligence-based network data intelligent management system, comprising an adjustment module, a matrix construction module, a reconstruction module, and a performance evaluation module; Adjustment module: The sampling frequency is adjusted in real time according to the network status and sensor sampling data through the AI ​​algorithm, and the raw sensor data is collected using the random sampling method. The raw sensor data is sent to the matrix construction module; Matrix construction module: Project the raw sensor data to the sparse domain through the autoencoder and verify whether the sparsity of the raw sensor data meets the compressed sensing conditions. If so, a Gaussian random matrix is ​​constructed and sent to the reconstruction module. Reconstruction module: Use the basis pursuit reconstruction algorithm to restore the original sensor data according to the Gaussian random matrix, and send the reconstruction results to the performance evaluation module; Performance evaluation module: After collecting performance data from actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention projects the original sensor data to the sparse domain through the autoencoder, and verifies whether the sparsity of the original sensor data meets the compressed sensing conditions. If so, a Gaussian random matrix is ​​constructed, which maps the sparse signal to a low-dimensional space. The server uses a basis pursuit reconstruction algorithm to restore the original sensor data based on the Gaussian random matrix. After collecting the performance data in the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result. In application scenarios with limited bandwidth and limited equipment energy consumption, it can significantly reduce the cost of data collection and transmission while ensuring data quality, and improve the overall efficiency and intelligence level of network management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Example 1: Please refer to Figure 1 As shown, the present embodiment provides an artificial intelligence-based network data intelligent management method, the management method comprising the following steps: The collection end uses AI algorithms to adjust the sampling frequency in real time according to network conditions and sensor sampling data. After collecting the original sensor data using a random sampling method, the original sensor data is projected into a sparse domain through an autoencoder, and it is verified whether the sparsity of the original sensor data meets the compressed sensing conditions. If so, a Gaussian random matrix is ​​constructed, which maps the sparse signal to a low-dimensional space. In this process, the sensor only needs to collect a small part of the data, rather than the original complete signal. The server uses a basis pursuit reconstruction algorithm to restore the original sensor data based on the Gaussian random matrix. After collecting the performance data in the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result.

[0021] This application projects the original sensor data into a sparse domain through an autoencoder, and verifies whether the sparsity of the original sensor data meets the compressed sensing conditions. If so, a Gaussian random matrix is ​​constructed, which maps the sparse signal into a low-dimensional space. The server uses a basis pursuit reconstruction algorithm to restore the original sensor data based on the Gaussian random matrix. After collecting the performance data in the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result. In application scenarios with limited bandwidth and limited device energy consumption, it can significantly reduce the cost of data collection and transmission while ensuring data quality, and improve the overall efficiency and intelligence level of network management.

[0022] Embodiment 2: The collection end adjusts the sampling frequency in real time according to the network status and sensor sampling data through an AI algorithm, including the following steps: The collection end obtains network status data, including bandwidth occupancy rate, obtains sensor sampling data, including signal change rate, normalizes bandwidth occupancy rate and signal change rate, maps their value ranges to [0,1], obtains bandwidth occupancy rate normalized value and signal change rate normalized value, subtracts signal change rate normalized value from bandwidth occupancy rate normalized value to obtain adjustment factor, the larger the adjustment factor, the more the sampling frequency needs to be reduced, and the sampling frequency is corrected by the obtained adjustment factor, the expression is: , where is the corrected sampling frequency, is the sampling frequency before correction, is the adjustment factor.

[0023] The calculation logic of bandwidth utilization is as follows: obtain the current network used bandwidth and the total network bandwidth, and divide the total network bandwidth by the current network used bandwidth to obtain the bandwidth utilization. The larger the bandwidth utilization, the more limited the bandwidth is, and the sampling frequency needs to be reduced to avoid network overload. The calculation expression of the signal change rate is: , where is the signal change rate, is the number of sampling points, For sensor The smaller the signal change rate is, the smaller the sensor data change is, which means the data change is relatively stable and the sampling frequency needs to be lowered.

[0024] After collecting the raw sensor data using the random sampling method, the raw sensor data is projected into the sparse domain through the autoencoder, which includes the following steps: The Bernoulli sampling method is used to sample the original sensor data. The signal obtained after sampling is a sparse representation of the original data. Suppose the original sensor data is: , then the sensor data after sampling is: ; An autoencoder is an unsupervised neural network that learns low-dimensional representations of data. It consists of two main parts: an encoder that maps input data to a latent low-dimensional space, and a decoder that reconstructs the original data from the latent space.

[0025] In data compression sensing applications, the encoder is used to map sparse sampled data to a sparse domain, and data compression is achieved by learning the representation of sparse features. The encoder: maps the sampled sensor data y to the sparse domain representation z through a neural network, and the decoder: tracks and reconstructs the original signal from the sparse domain representation z. ; Let the encoder of the autoencoder be: , the decoder is: ,but: (Encoding process, projection to sparse domain); (Decoding process, reconstructing the original signal); In an autoencoder, sparsity is achieved through the architecture and training of the network. The goal of the encoder is to find a mapping that makes the representation in the sparse domain z as sparse as possible. Through training, the autoencoder automatically learns how to represent the signal in the sparse domain while minimizing the reconstruction error.

[0026] During the training process of the autoencoder, sparsity constraints can be added to encourage the network to learn sparse feature representations. Common sparsity constraints include: L1 regularization: encouraging weight sparsity through L1 norm constraints, sparse coding: forcing some elements in the encoder's output z to be close to zero, making it sparse.

[0027] During the optimization process, the goal is to minimize the trade-off between the reconstruction error and the sparsity constraint. The objective function is usually: , where is the reconstruction error (L2 norm), is the sparsity constraint (L1 norm), is a regularization parameter that controls sparsity.

[0028] Verify whether the sparsity of the original sensor data meets the compressed sensing conditions. If so, construct a Gaussian random matrix that maps the sparse signal to a low-dimensional space. In this process, the sensor only needs to collect a small portion of data, rather than the original complete signal. The following steps are included: After obtaining the non-zero element ratio value and norm ratio value of the original sensor data, the sparse coefficient of the element sensor data is calculated and the expression is: , where is the sparse coefficient, is the non-zero element ratio value, is the norm ratio, , is the adjustment coefficient, and , All are greater than 0; The logic for obtaining the non-zero element ratio value is as follows: obtain the number of non-zero coefficients after the original sensor data is transformed, obtain the total dimension of the original sensor data, and divide the number of non-zero coefficients after the transformation by the total dimension to obtain the non-zero element ratio value. The smaller the non-zero element ratio value is, the original sensor data can be considered to be sparse and suitable for compressed sensing.

[0029] In compressed sensing, the L1 norm of sparse signals is usually small, while the L2 norm is relatively large. Therefore, the ratio between the L1 norm and the L2 norm of the signal can be used to evaluate sparsity. The logic for obtaining the norm ratio is: calculate the L1 norm, and the expression is: , calculate the L2 norm, the expression is: , the norm ratio is obtained by dividing the L1 norm by the L2 norm. The larger the norm ratio, the lower the sparsity of the original sensor data, and the smaller the norm ratio, the sparse the signal.

[0030] The smaller the sparse coefficient, the more the sparsity of the original sensor data meets the conditions. The obtained sparse coefficient is compared with the preset sparse threshold. If the sparse coefficient is greater than the sparse threshold, it is judged that the sparsity of the original sensor data does not meet the compressed sensing conditions. The sparsity of the data can be improved through other transform domains. If the sparse coefficient is less than or equal to the sparse threshold, it is judged that the sparsity of the original sensor data meets the compressed sensing conditions. If the sparsity of the original sensor data does not meet the compressed sensing conditions, the sparsity of the data can be improved through other transform domains. Commonly used transform domains include: Wavelet transform: Suitable for multi-resolution analysis, especially when the signal has local characteristics (such as images or sensor data).

[0031] Fourier transform: For periodic signals or signals with obvious frequency characteristics, Fourier transform can usually significantly improve sparsity.

[0032] Discrete Cosine Transform (DCT): Especially suitable for images, video signals, etc., and often used for data compression.

[0033] Principal Component Analysis (PCA): For high-dimensional data sets, PCA can help extract the most important components, making the data sparse; By projecting the signal into different transform domains, it may be possible to find a more suitable domain that improves the sparsity of the signal.

[0034] When the sparsity of the original sensor data meets the compressed sensing conditions, a Gaussian random matrix is ​​constructed for compressed sensing sampling. The characteristic of the Gaussian random matrix is ​​that each element is independent and conforms to the standard normal distribution, which ensures the good properties of the sampling matrix and can effectively retain the key information of the signal during the compressed sensing process.

[0035] Generate a Gaussian random matrix: , where m is the dimension of the sampled sensor data, , n is the dimension of the original sensor data, and each element of the Gaussian random matrix conforms to the standard normal distribution : , where i is the index of the sampling dimension and j is the index of the original dimension of the signal; The original signal is sampled using a Gaussian random matrix. The sampling process obtains low-dimensional data through matrix multiplication: .

[0036] The server uses the basis pursuit reconstruction algorithm to restore the original sensor data based on the Gaussian random matrix, including the following steps: The original sensor data is sparsely represented using the discrete cosine transform dictionary D, and the original sensor data x is reconstructed based on the sampled sensor data y. The optimization problem is expressed as: , where is a sparse coefficient vector, yes norm, which is used to measure the error between the reconstructed signal and the compressed data, yes norm, which controls sparsity and encourages most elements in sparse coefficient vectors to be zero, is a regularization parameter used to balance reconstruction error and sparsity.

[0037] Initialize the sparse coefficients to a zero vector and calculate the residual of the reconstructed signal : , in each iteration, update To minimize the objective function. Common update methods include: Matching-Pursuit: In each iteration, the basis vector in the dictionary that best matches the current residual is selected and the sparse coefficients are updated.

[0038] Orthogonal-Matching-Pursuit (OMP): gradually select the dictionary basis most relevant to the residual and solve the sparse coefficients by the least squares method.

[0039] Iterative-Thresholding: Apply a threshold to each sparse coefficient, retain the most important coefficients, and gradually approach the optimal solution.

[0040] According to the selected algorithm, the residual is updated and the iteration is repeated until a predetermined stopping criterion is met (such as the residual reaches a certain threshold or the maximum number of iterations is reached).

[0041] The sparse coefficients are obtained through the optimization process , and then through the discrete cosine transform dictionary D and sparse coefficients Reconstruct the signal: , is the original signal reconstructed by basis pursuit.

[0042] After collecting the performance data from the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result, including the following steps: Collect performance data from actual deployment, including compression rate, reconstruction error, reconstruction time, and transmission efficiency. Substitute the compression rate, reconstruction error, reconstruction time, and transmission efficiency into the evaluation model to calculate the performance score of this deployment. The expression is: , where To rate performance, is the compression ratio, For transmission efficiency, is the reconstruction error, For the reconstruction time, is the regression coefficient, and the regression coefficients are all greater than 0; The larger the performance score, the better the overall performance of this deployment. The obtained performance score is compared with the preset score threshold. The score threshold is used to evaluate whether the performance of this deployment meets the standard. If the performance score is greater than or equal to the score threshold, the performance of this deployment meets the standard. If the performance score is less than the score threshold, the performance of this deployment does not meet the standard. When the deployment performance is evaluated as not up to standard, the generated optimization strategy is: Redesign the network structure to reduce data transmission paths and communication delays. Increase network bandwidth, avoid network bottlenecks, and ensure smooth data transmission. Use efficient transmission protocols, such as QUIC or UDP instead of TCP, to reduce handshake and retransmission delays. Add backup links to improve network reliability and reduce performance degradation caused by single link failure. Implement traffic distribution and load balancing to avoid performance degradation caused by overload of a certain node. Use packet retransmission mechanism or forward error correction technology (FEC) to improve the reliability of packet transmission. If the data loss rate or error is high, consider using higher precision or more reliable sensor equipment. Avoid single point failures and ensure system stability by deploying redundant equipment. Increase processors, memory, storage capacity, etc. to improve the computing and storage capabilities of the device and support more data processing tasks.

[0043] Dynamically adjust the data collection frequency according to network conditions and device load to avoid overload caused by too high sampling frequency. Use efficient compression algorithms to reduce the amount of data transmitted and reduce network bandwidth usage. Improve data processing algorithms, such as using more efficient transmission encoding or data analysis methods. Distribute data processing tasks to multiple nodes to avoid performance bottlenecks of a single node. Use load balancing strategies to ensure that each node in the system is not overloaded or lacks resources. Add more computing nodes to share processing tasks and improve the overall processing capacity of the system. Use intelligent scheduling algorithms based on load, priority, and delay to ensure that critical tasks are executed first and optimize resource allocation.

[0044] Avoid storing redundant data and use compressed storage formats to reduce storage space usage. Use more efficient storage media, such as SSDs, to increase data read and write speeds. Accelerate data access and transmission through multi-level cache mechanisms to reduce data latency. Distribute data to multiple nodes to improve storage scalability and reliability.

[0045] Dynamically adjust the power consumption mode of the device according to the load, reduce the power consumption of the device when the load is low, and extend the service life of the device. Optimize energy consumption, reduce the overall energy consumption of the system, and improve energy efficiency. Optimize the processing flow of the system, reduce delays, and improve the user interaction experience. Simplify the user interface and interaction process to improve the convenience and efficiency of user operations. Implement real-time monitoring, track the performance indicators of the system, and dynamically adjust the system configuration based on real-time data. Establish a fault warning system to detect potential problems in a timely manner and handle them.

[0046] Collect feedback from users to understand the shortcomings of the system in actual use and provide a basis for further optimization. Based on performance data and user feedback, continuously adjust and optimize strategies to improve system performance. According to the complexity and bottlenecks of the system, gradually implement the optimization strategy in stages. You can start with the most important bottleneck problem and gradually optimize other aspects. For the optimization plan, you can conduct A / B testing to verify the effects of different plans to ensure that the optimization strategy is effective.

[0047] Embodiment 3: The network data intelligent management system based on artificial intelligence described in this embodiment includes an adjustment module, a matrix construction module, a reconstruction module, and a performance evaluation module; Adjustment module: The sampling frequency is adjusted in real time according to the network status and sensor sampling data through the AI ​​algorithm, and the raw sensor data is collected using the random sampling method. The raw sensor data is sent to the matrix construction module; Matrix construction module: Project the raw sensor data to the sparse domain through the autoencoder and verify whether the sparsity of the raw sensor data meets the compressed sensing conditions. If so, a Gaussian random matrix is ​​constructed and sent to the reconstruction module. Reconstruction module: Use the basis pursuit reconstruction algorithm to restore the original sensor data according to the Gaussian random matrix, and send the reconstruction results to the performance evaluation module; Performance evaluation module: After collecting performance data from actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result.

[0048] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0049] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0050] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A network data intelligent management method based on artificial intelligence, characterized in that: The management method comprises the following steps: The collection end uses AI algorithms to adjust the sampling frequency in real time according to network conditions and sensor sampling data. After collecting raw sensor data using a random sampling method, the raw sensor data is projected into a sparse domain through an autoencoder. Verify whether the sparsity of the original sensor data meets the compressed sensing conditions. If so, construct a Gaussian random matrix. The server uses the basis pursuit reconstruction algorithm to restore the original sensor data based on the Gaussian random matrix. After collecting the performance data from the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result.

2. The method for intelligent management of network data based on artificial intelligence according to claim 1, characterized in that: After collecting the performance data from the actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result, including the following steps: Collect performance data from actual deployment, including compression ratio, reconstruction error, reconstruction time, and transmission efficiency. Substitute the compression ratio, reconstruction error, reconstruction time, and transmission efficiency into the evaluation model to calculate the performance score of this deployment. The obtained performance score is compared with the preset score threshold. The score threshold is used to evaluate whether the performance of this deployment meets the standard. If the performance score is greater than or equal to the score threshold, the performance of this deployment is evaluated to meet the standard. If the performance score is less than the score threshold, the performance of this deployment is evaluated to fail to meet the standard.

3. The method for intelligent management of network data based on artificial intelligence according to claim 2, characterized in that: Substitute the compression rate, reconstruction error, reconstruction time, and transmission efficiency into the evaluation model to calculate the performance score of this deployment. The expression is: , where To rate performance, is the compression ratio, For transmission efficiency, is the reconstruction error, For the reconstruction time, are regression coefficients, and all regression coefficients are greater than 0.

4. The method for intelligent management of network data based on artificial intelligence according to claim 3, characterized in that: Verifying whether the sparsity of the original sensor data meets the compressed sensing conditions includes the following steps: After obtaining the non-zero element ratio value and norm ratio value of the original sensor data, the sparse coefficient of the element sensor data is calculated and the expression is: , where is the sparse coefficient, is the non-zero element ratio value, is the norm ratio, , is the adjustment coefficient, and , All are greater than 0; The obtained sparse coefficient is compared with the preset sparse threshold. If the sparse coefficient is greater than the sparse threshold, it is judged that the sparsity of the original sensor data does not meet the compressed sensing conditions. If the sparse coefficient is less than or equal to the sparse threshold, it is judged that the sparsity of the original sensor data meets the compressed sensing conditions.

5. The method for intelligent management of network data based on artificial intelligence according to claim 4, characterized in that: The logic for obtaining the non-zero element ratio value is as follows: obtaining the number of non-zero coefficients after the original sensor data is transformed, obtaining the total dimension of the original sensor data, and dividing the number of non-zero coefficients after the transformation by the total dimension to obtain the non-zero element ratio value; The logic for obtaining the norm ratio is: calculate the L1 norm, and the expression is: , calculate the L2 norm, the expression is: , the norm ratio is obtained by dividing the L1 norm by the L2 norm.

6. The method for intelligent management of network data based on artificial intelligence according to claim 1, characterized in that: After collecting the raw sensor data using the random sampling method, the raw sensor data is projected into the sparse domain through the autoencoder, which includes the following steps: The Bernoulli sampling method is used to sample the original sensor data. The signal obtained after sampling is a sparse representation of the original data. Suppose the original sensor data is: , then the sensor data after sampling is: ; The autoencoder includes an encoder and a decoder. The encoder maps the sampled sensor data y to a sparse domain representation z through a neural network, and the decoder reconstructs the original sensor data from the sparse domain representation z. , let the encoder of the autoencoder be: , the decoder is: ,but: ; ; During the autoencoder optimization process, the goal is to minimize the trade-off between reconstruction error and sparsity constraints. The objective function is: , where is the reconstruction error, is the sparsity constraint, is the regularization parameter that controls sparsity.

7. The method for intelligent management of network data based on artificial intelligence according to claim 6, characterized in that: The collection end uses AI algorithms to adjust the sampling frequency in real time according to network conditions and sensor sampling data, including the following steps: The acquisition end obtains network status data, including bandwidth occupancy rate, and obtains sensor sampling data, including signal change rate. The adjustment factor is calculated based on the bandwidth occupancy rate and signal change rate, and the sampling frequency is corrected by the obtained adjustment factor. The expression is: , where is the corrected sampling frequency, is the sampling frequency before correction, is the adjustment factor.

8. The method for intelligent management of network data based on artificial intelligence according to claim 7, characterized in that: The calculation logic of the adjustment factor is as follows: normalize the bandwidth occupancy rate and the signal change rate so that their value ranges are mapped to [0, 1], obtain the normalized value of the bandwidth occupancy rate and the normalized value of the signal change rate, and subtract the normalized value of the signal change rate from the normalized value of the bandwidth occupancy rate to obtain the adjustment factor.

9. The method for intelligent management of network data based on artificial intelligence according to claim 8, characterized in that: The calculation logic of the bandwidth occupancy rate is as follows: obtaining the currently used bandwidth of the network and the total network bandwidth, and dividing the total network bandwidth by the currently used bandwidth of the network to obtain the bandwidth occupancy rate; The calculation expression of the signal change rate is: , where is the signal change rate, is the number of sampling points, For sensor The sampling value of times.

10. An artificial intelligence-based network data intelligent management system, used to implement the management method according to any one of claims 1 to 9, characterized in that: It includes adjustment module, matrix construction module, reconstruction module and performance evaluation module; Adjustment module: The sampling frequency is adjusted in real time according to the network status and sensor sampling data through the AI ​​algorithm, and the raw sensor data is collected using the random sampling method. The raw sensor data is sent to the matrix construction module; Matrix construction module: Project the raw sensor data to the sparse domain through the autoencoder and verify whether the sparsity of the raw sensor data meets the compressed sensing conditions. If so, a Gaussian random matrix is ​​constructed and sent to the reconstruction module. Reconstruction module: Use the basis pursuit reconstruction algorithm to restore the original sensor data according to the Gaussian random matrix, and send the reconstruction results to the performance evaluation module; Performance evaluation module: After collecting performance data from actual deployment, the performance score of this deployment is calculated based on the evaluation model, and the corresponding optimization strategy is generated based on the performance score calculation result.

Images