Wireless network resource joint optimization method and system based on federal feature sparsification

Abstract

The application discloses a wireless network resource joint optimization method and system based on federal feature sparsification, relates to the technical field of wireless communication network resource management, and discloses the wireless network resource joint optimization method and system based on federal feature sparsification, which reduces the transmission data amount through feature sparsification processing, protects privacy in combination with noise disturbance, and dynamically generates a resource allocation strategy based on global feature analysis, can improve resource allocation efficiency and reduce communication overhead, and simultaneously improves the privacy protection capability.

Description

Technical Field

[0001] This application relates to the field of wireless communication network resource management technology, and in particular to a method and system for joint optimization of wireless network resources based on federated feature sparsity. Background Technology

[0002] In the current field of wireless network communication, with the rapid increase in the number of terminal devices and the diversification of network applications, the management and optimization of wireless network resources face unprecedented challenges. Traditional wireless network resource allocation methods often employ centralized processing, requiring the status information of all terminal devices to be uploaded to a central node for unified processing. This not only leads to huge communication overhead but also easily causes privacy leaks. Furthermore, due to the spatiotemporal differences in network status among different terminal devices, centralized processing struggles to accurately reflect the actual needs of each terminal in real time, resulting in uneven resource allocation and low efficiency. In addition, existing technologies lack effective feature filtering mechanisms when processing massive amounts of terminal device data, leading to excessive transmission of redundant data and further exacerbating network congestion. Regarding privacy protection, traditional methods typically employ simple data anonymization or encrypted transmission, failing to achieve effective privacy protection while ensuring data availability. For the dynamic allocation of network resources, most existing solutions rely on static models for decision-making, which cannot adapt to the rapid changes in the wireless network environment, resulting in low resource utilization.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this application is to provide a method and system for joint optimization of wireless network resources based on federated feature sparsity, which aims to improve resource allocation efficiency and reduce communication overhead.

[0005] To achieve the above objectives, this application proposes a joint optimization method for wireless network resources based on federated feature sparsity, the method comprising:

[0006] The terminal device collects the raw dataset of local network status and performs feature extraction and standardization on the raw dataset to generate a standardized feature vector;

[0007] The terminal device performs a feature importance analysis operation based on the standardized feature vector to generate a feature importance vector;

[0008] The terminal device receives channel capacity indication information sent by the base station, and performs a sparse mask generation operation in combination with the feature importance vector to generate a binary feature mask.

[0009] The terminal device extracts a subset of key features from the standardized feature vector based on the binary feature mask, adds noise perturbation to the subset of key features to generate privacy-preserving feature data, and uploads the privacy-preserving feature data to the edge aggregation node.

[0010] The edge aggregation node receives the privacy protection feature data uploaded by multiple terminals, performs cross-device feature alignment operation on the privacy protection feature data, and generates an aligned feature dataset;

[0011] A weighted fusion process is performed on the aligned feature dataset to generate a global feature space matrix;

[0012] Based on the global feature space matrix analysis of the device spatiotemporal correlation, a network resource strategy including channel allocation parameters and terminal scheduling sequence is generated.

[0013] The base station controller executes the network resource policy and broadcasts the updated channel state parameters to the terminal devices.

[0014] In one embodiment, the terminal device collects a raw dataset of local network status and performs feature extraction and standardization processing on the raw dataset to generate a standardized feature vector, including the following steps:

[0015] Terminal devices collect raw data on local network signal strength, transmission delay, and device power consumption.

[0016] The original data is segmented using a preset time window, and time-domain statistical features are calculated within the time window;

[0017] Perform Fourier transform on the data within the time window to extract the energy features of the principal components in the frequency domain;

[0018] The time-domain statistical features and the frequency-domain principal component energy features are respectively subjected to minimum-maximum normalization, and the normalized time-domain statistical features and frequency-domain principal component energy features are concatenated to generate the standardized feature vector.

[0019] In one embodiment, the step of the terminal device performing feature importance analysis based on the standardized feature vector to generate a feature importance vector includes:

[0020] The standardized feature vector is copied to generate a test vector copy, and the specified feature dimension value in the test vector copy is set to zero to generate a masked feature vector;

[0021] Based on a preset network performance prediction function, calculate the predicted network throughput value corresponding to the shielding feature vector;

[0022] Calculate the deviation between the baseline network throughput prediction value of the standardized feature vector and the network throughput prediction value of the masked feature vector;

[0023] An initial sequence of feature importance is generated based on the deviation, and a sliding window smoothing process is performed on the initial sequence of feature importance to generate the feature importance vector.

[0024] In one embodiment, the terminal device receives channel capacity indication information from the base station and performs a sparse mask generation operation in conjunction with the feature importance vector to generate a binary feature mask. The steps include:

[0025] The available transmission bits are parsed from the channel capacity indication information, and the average number of coded bits per feature dimension is calculated based on the locally stored historical transmission records.

[0026] The maximum number of transmittable dimensions is calculated based on the available number of transmittable bits and the average number of encoded bits per single feature dimension.

[0027] The feature dimensions are sorted in descending order according to the feature importance vector, and the top K dimensions are selected as the set of retained feature dimensions, where K is equal to the maximum number of transmittable dimensions.

[0028] Construct a binary mask matrix by setting the corresponding positions of the reserved feature dimension set to 1 and the rest to 0 to generate the binary feature mask.

[0029] In one embodiment, the step of generating privacy-preserving feature data by adding noise perturbation to the subset of key features includes:

[0030] Based on the feature type of the key feature subset, query the preset noise intensity coefficient mapping table to obtain the noise intensity coefficient of the current feature, and generate a Laplace distribution random number based on the noise intensity coefficient.

[0031] The random number is superimposed onto the numerical portion of the key feature subset;

[0032] Boundary truncation is performed on the superimposed feature values ​​to generate the privacy-preserving feature data.

[0033] In one embodiment, the edge aggregation node receives privacy-preserving feature data uploaded by multiple terminals, and performs cross-device feature alignment on the privacy-preserving feature data to generate an aligned feature dataset. The steps include:

[0034] Parse the dimensional description information of the privacy protection feature data uploaded by each terminal, and extract the correspondence between feature names and dimensional indices; perform cross-terminal dimensional matching based on feature name similarity to generate a terminal feature dimension mapping table;

[0035] Adjust the dimensional order of privacy protection feature data for each terminal according to the terminal feature dimension mapping relationship table;

[0036] The missing dimension locations in the adjusted data are detected, and the missing dimension locations are filled with the historical average of global features from the most recent preset first number of historical data collection periods to generate the aligned feature dataset.

[0037] In one embodiment, the step of performing a weighted fusion process on the aligned feature dataset to generate a global feature space matrix includes:

[0038] The number of local data samples for each terminal is counted, and the aggregation weight coefficient for each terminal is determined based on the number of local data samples and the total number of samples.

[0039] Perform a weighted calculation on the feature values ​​of each feature dimension in the aligned feature dataset;

[0040] The weighted calculation results are normalized by standard deviation, and the normalized values ​​of all dimensions are integrated to generate the global feature space matrix.

[0041] In one embodiment, the step of generating a network resource strategy including channel allocation parameters and terminal scheduling sequences based on the spatiotemporal correlation of the devices analyzed by the global feature space matrix includes:

[0042] Obtain the geographic location coordinate data reported by the terminal device, and calculate the Euclidean distance matrix between the terminal devices based on the geographic location coordinate data;

[0043] Construct a spatiotemporal correlation matrix based on the Euclidean distance matrix and the device service type;

[0044] Cluster analysis of the spatiotemporal correlation matrix is ​​performed to identify communication device groups, and a channel bandwidth ratio is allocated to each group based on the communication device groups;

[0045] Generate a terminal scheduling priority sequence based on the urgency of the service;

[0046] The network resource strategy is generated by integrating the channel bandwidth allocation ratio and the terminal scheduling priority sequence.

[0047] In one embodiment, the method further includes:

[0048] Calculate the cosine similarity between the network resource strategies for a second consecutive predetermined number of decision periods;

[0049] When the cosine similarity is less than a preset threshold, a sparsity adjustment instruction is generated based on the cosine similarity and sent to the terminal device;

[0050] The terminal device updates the sliding window size in the feature importance analysis according to the sparsity adjustment instruction.

[0051] Furthermore, to achieve the above objectives, this application also proposes a joint optimization system for wireless network resources based on federated feature sparsity. The system includes: a memory, a processor, and a joint optimization program for wireless network resources based on federated feature sparsity stored in the memory and executable on the processor. The joint optimization program for wireless network resources based on federated feature sparsity is configured to implement the steps of the joint optimization method for wireless network resources based on federated feature sparsity.

[0052] The proposed method and system for joint optimization of wireless network resources based on federated feature sparsity reduces the amount of transmitted data through feature sparsity processing, protects privacy by combining noise perturbation, and dynamically generates resource allocation strategies based on global feature analysis. This can improve resource allocation efficiency and reduce communication overhead, while enhancing privacy protection capabilities. Attached Figure Description

[0053] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart illustrating an embodiment of the joint optimization method for wireless network resources based on federated feature sparsity in this application.

[0056] Figure 2 For this application Figure 1 A detailed flowchart of step S100;

[0057] Figure 3 For this application Figure 1 A detailed flowchart of step S200;

[0058] Figure 4 For this application Figure 1 Detailed flowchart of step S300;

[0059] Figure 5 For this application Figure 1 Detailed flowchart of step S400;

[0060] Figure 6 For this applicationFigure 1 A detailed flowchart of step S500;

[0061] Figure 7 For this application Figure 1 A detailed flowchart of step S600;

[0062] Figure 8 For this application Figure 1 A detailed flowchart of step S700;

[0063] Figure 9 This is a flowchart illustrating another embodiment of the joint optimization method for wireless network resources based on federated feature sparsity in this application.

[0064] Figure 10 This is a schematic diagram of a structural embodiment of the wireless network resource joint optimization system based on federated feature sparsity provided in this application.

[0065] Explanation of icon numbers:

[0066] 10. Memory; 20. Processor.

[0067] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0068] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0069] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0070] In existing technologies, the field of wireless network communication faces challenges brought about by the surge in the number of terminal devices and the diversification of network applications. Traditional methods rely on centralized processing, requiring all terminal devices to upload complete status information to a central node, resulting in excessive communication overhead and privacy risks. The network status of different terminal devices varies in time and space, and centralized processing cannot reflect the actual needs of each terminal in real time, leading to uneven resource allocation and inefficiency. For example, in dense urban environments, when a large number of mobile terminals simultaneously request network resources, the central node cannot quickly process the dynamically changing device status, resulting in increased communication latency for some terminals.

[0071] To address the aforementioned issues, a distributed resource optimization method is needed that can reduce communication overhead while protecting privacy. Considering the local data processing capabilities of terminal devices, feature filtering can reduce the amount of transmitted data, but it's crucial to ensure that the filtered features accurately reflect the network status. If a federated learning framework is adopted, where each terminal processes data independently and uploads only key information, global resource optimization must also be achieved while protecting privacy. Therefore, by introducing a dynamic feature sparsity mechanism, the feature dimension is adaptively adjusted according to channel capacity, balancing transmission efficiency and information integrity. Furthermore, cross-device feature alignment and fusion at edge nodes can effectively capture spatiotemporal correlations, providing a basis for resource allocation decisions.

[0072] Based on this, embodiments of this application provide a joint optimization method for wireless network resources based on federated feature sparsity, referring to... Figure 1 The wireless network resource joint optimization method based on federated feature sparsity includes steps S100 to S800, wherein:

[0073] Step S100: The terminal device collects the original dataset of local network status and performs feature extraction and standardization processing on the original dataset to generate a standardized feature vector.

[0074] Step S200: The terminal device performs a feature importance analysis operation based on the standardized feature vector to generate a feature importance vector;

[0075] In step S300, the terminal device receives the channel capacity indication information sent by the base station, and performs a sparse mask generation operation in combination with the feature importance vector to generate a binary feature mask.

[0076] Step S400: The terminal device extracts a subset of key features from the standardized feature vector based on the binary feature mask, adds noise perturbation to the subset of key features to generate privacy-preserving feature data, and uploads the privacy-preserving feature data to the edge aggregation node.

[0077] Step S500: The edge aggregation node receives the privacy protection feature data uploaded by multiple terminals, performs a cross-device feature alignment operation on the privacy protection feature data, and generates an aligned feature dataset.

[0078] Step S600: Perform weighted fusion processing on the aligned feature dataset to generate a global feature space matrix;

[0079] Step S700: Based on the global feature space matrix analysis of the device spatiotemporal correlation, generate a network resource strategy that includes channel allocation parameters and terminal scheduling sequences;

[0080] In step S800, the base station controller executes the network resource policy and broadcasts the updated channel state parameters to the terminal device.

[0081] In this embodiment, the standardized feature vector refers to uniformly scaled data formed by extracting time-domain statistical features and transforming frequency-domain energy features from the original network state data. Specifically, this can be achieved by methods such as calculating the mean and variance through time window segmentation and extracting principal component energy through Fourier transform, used to eliminate differences in the dimensions of different features. Feature importance analysis refers to evaluating the impact of perturbing specific feature dimensions on network performance prediction. Specifically, this can be achieved by masking some features and comparing the deviation of the prediction results, used to identify feature dimensions that have a significant impact on resource allocation decisions. Binary feature masking refers to a selector that dynamically determines the transmittable feature dimensions based on channel capacity constraints. Specifically, it can be generated by truncating low-value dimensions after sorting by importance, used to prioritize the transmission of high-value features under limited bandwidth. Privacy-preserving feature data refers to data with noise superimposed on the selected key features. Specifically, it can use a Laplace mechanism to generate perturbation values ​​that meet differential privacy requirements, used to prevent the original feature information from being reverse-engineered during transmission. Cross-device feature alignment refers to the operation of unifying the feature dimensions and order of data uploaded from different terminals. This can be achieved by using feature name matching and filling missing dimensions with historical means, thus eliminating the impact of device heterogeneity on global analysis. The global feature space matrix is ​​a unified representation formed by aggregating feature data from multiple terminals. This can be generated using sample size weighting fusion and standard deviation normalization, comprehensively reflecting the spatiotemporal distribution characteristics of network status. Device spatiotemporal correlation analysis refers to the operation of mining communication relationships between devices by combining geographical location and service type. This can be achieved by using Euclidean distance matrix construction and cluster analysis, used to identify strongly correlated device groups to optimize channel allocation.

[0082] In this embodiment, the terminal device first collects raw data such as local signal strength and transmission delay, extracts time-domain statistical features through time window segmentation, and extracts frequency-domain energy features using Fourier transform. After normalization, these are concatenated to form a standardized feature vector, eliminating differences in the dimensions of different features. Subsequently, by masking some feature dimensions and evaluating their impact on network throughput prediction, a vector reflecting the importance of each dimension is generated. Based on the current channel capacity information issued by the base station, the maximum number of transmittable feature dimensions is calculated, and key dimensions are selected according to importance to generate a binary mask. After extracting a feature subset using this mask, noise that meets privacy protection requirements is superimposed to generate irreversible perturbation data for uploading. After receiving data from multiple terminals, the edge aggregation node adjusts the dimension order by matching feature names and fills missing dimensions with historical averages to form an aligned dataset with a unified structure. The aligned data is subjected to sample-weighted fusion and normalization to generate a global feature matrix. A spatial distance matrix is ​​calculated by combining the device's geographical location, and a spatiotemporal correlation matrix is ​​constructed by superimposing service type similarity. Cluster analysis is used to divide devices into groups and allocate channel resources. A scheduling priority sequence is generated based on the urgency of the services, forming a network strategy that includes channel allocation ratios and scheduling order. After the base station executes the policy, it broadcasts the updated channel state parameters to the terminal equipment, forming a closed-loop optimization.

[0083] In this embodiment, the proposed solution significantly reduces the amount of transmitted data through dynamic feature filtering. Feature perturbation is performed at the terminal side to ensure that the original data does not leave the local device. Global feature fusion and clustering analysis accurately identify device grouping characteristics, improving channel utilization. The feature selection threshold is adjusted based on real-time channel capacity to achieve a balance between transmission efficiency and information integrity. Thus, this application effectively reduces the communication burden between the terminal and the base station, avoiding the transmission of redundant feature data. While protecting user privacy, cross-device feature alignment and fusion accurately capture the spatiotemporal distribution patterns of network status. The channel allocation strategy based on device grouping characteristics improves spectrum resource utilization in dense scenarios. The dynamically adjusted feature sparsity mechanism enables the system to adapt to changes in channel capacity, maintaining the real-time performance and accuracy of resource allocation. The synergistic effect of noise perturbation and feature filtering reduces the risk of privacy leakage while ensuring the effective transmission of critical information.

[0084] In one feasible implementation, refer to Figure 2 Step S100 includes steps S110 to S140, wherein:

[0085] Step S110: The terminal device collects raw data of signal strength, transmission delay, and device power consumption of the local network.

[0086] Step S120: The original data is segmented by a preset time window, and the time domain statistical features are calculated within the time window;

[0087] Step S130: Perform Fourier transform on the data within the time window to extract the energy features of the principal components in the frequency domain;

[0088] Step S140: Perform maximum and minimum normalization on the time-domain statistical features and the frequency-domain principal component energy features respectively, and concatenate the normalized time-domain statistical features and frequency-domain principal component energy features to generate the standardized feature vector.

[0089] In this embodiment, the preset time window refers to the interval into which continuously collected raw data is divided according to a fixed duration. Specifically, a sliding window mechanism can be used, for example, dividing the data into 5-second or 10-second segments to capture dynamic changes in network status. Time-domain statistical features refer to statistical indicators such as the mean, variance, and extreme values ​​calculated for the data within the time window. Specifically, this can be achieved through a moving average algorithm or difference calculation, used to characterize the short-term fluctuations in signal strength, transmission delay, and device power consumption. Fourier transform is a mathematical method for converting a time-domain signal into a frequency-domain signal. Specifically, a fast Fourier transform algorithm can be used to extract periodic components from the signal. Frequency-domain principal component energy features refer to the feature vectors of frequency-domain energy concentration regions selected through principal component analysis. Specifically, a covariance matrix decomposition method can be used to characterize the energy distribution characteristics of the signal in the frequency domain. Max-min normalization is a standardization method that linearly maps feature values ​​to the [0, 1] interval. Specifically, this can be achieved by calculating the ratio between the maximum and minimum values ​​of the feature dimension, used to eliminate differences in the dimensions of different features.

[0090] In this embodiment, signal strength, transmission delay, and device power consumption data are segmented into fixed-length time windows. Statistical features are calculated in the time domain to capture short-term fluctuation patterns, while frequency domain energy distribution features are extracted using Fourier transform to reflect periodic variation patterns. The time-domain and frequency-domain features are then normalized and concatenated to form a multi-dimensional standardized feature vector. For example, the mean and variance of signal strength are calculated within the time window, and the energy proportion of its principal components in the frequency domain is extracted. Both are then normalized and merged into a unified feature representation. This processing method can simultaneously preserve the time-varying characteristics and periodic patterns of the network state, providing a multi-dimensional data foundation for subsequent feature selection.

[0091] In this embodiment, the proposed solution utilizes time-frequency dual-domain feature fusion to more comprehensively describe the spatiotemporal variation patterns of network states. Simultaneously, normalization eliminates dimensional differences among multi-source data, improving feature comparability and the stability of subsequent processing. This addresses the problems of traditional network state feature extraction, such as its single dimension and insufficient representational ability. The collaborative extraction of time-domain statistical features and frequency-domain energy features enhances the representational ability of network state data, providing a more comprehensive data foundation for subsequent feature sparsity processing. Furthermore, normalization reduces the difficulty of handling the compatibility of multi-source heterogeneous data, avoids interference from features with different dimensions on model training, and improves the accuracy of network resource optimization decisions.

[0092] In one feasible implementation, refer to Figure 3 Step S200 includes steps S210 to S240, wherein:

[0093] Step S210: Copy the standardized feature vector to generate a test vector copy, and set the specified feature dimension value in the test vector copy to zero to generate a masked feature vector;

[0094] Step S220: Calculate the predicted network throughput value corresponding to the masking feature vector based on the preset network performance prediction function;

[0095] Step S230: Calculate the deviation between the baseline network throughput prediction value of the standardized feature vector and the network throughput prediction value of the masked feature vector;

[0096] Step S240: Generate an initial sequence of feature importance based on the deviation amount, and perform sliding window smoothing on the initial sequence of feature importance to generate the feature importance vector.

[0097] In this embodiment, the test vector copy refers to a temporary data copy generated by copying the original feature vector, which can be implemented using memory copying. This is used to isolate the impact of zeroing the feature dimension on the original data. The masked feature vector refers to the feature data after zeroing a specified feature dimension. This can be achieved by indexing and locating the target dimension and assigning it a zero value. This is used to simulate the network performance change when the feature dimension is missing. The network performance prediction function is a mathematical model used to evaluate network throughput. This can be implemented using a gradient boosting tree-based regression model, which outputs the predicted throughput from the input feature vector. The bias is the difference between the baseline prediction value and the masked prediction value. This can be calculated using the absolute difference or a relative percentage. This is used to quantify the degree of influence of the feature dimension on network performance. The sliding window smoothing process refers to the operation of locally averaging the sequence data. This can be implemented using a mean filtering algorithm with a fixed window length. This is used to eliminate random fluctuations in the feature importance evaluation process.

[0098] In this embodiment, during the feature importance analysis process, a test copy is first generated by replicating the standardized feature vector to ensure the integrity of the original data. Multiple masked feature vectors are generated by sequentially setting different feature dimensions in the test copy to zero, and these are input into the network performance prediction function to calculate the corresponding network throughput prediction values. The deviation between the baseline prediction value of the original feature vector and each masked prediction value is calculated to form an initial sequence reflecting the importance of each feature dimension. Due to the dynamic nature of the wireless network environment, the initial sequence may experience short-term fluctuations. A sliding window smoothing process is used to average the evaluation results of multiple adjacent periods, ultimately generating a stable and reliable feature importance vector.

[0099] In this embodiment, the proposed solution utilizes a quantitative evaluation mechanism that masks feature dimensions to dynamically capture the actual impact of different features on network performance. Combined with sliding window processing, it effectively suppresses transient interference, making the feature importance evaluation results more closely reflect the real-time network environment. This allows for accurate identification of key feature dimensions affecting network resource allocation, avoiding the waste of communication resources caused by redundant feature data transmission. By dynamically adjusting the feature importance evaluation results, it can adapt to the differences in network states of different terminal devices, providing a reliable basis for subsequent sparsity processing and thus improving the effectiveness of the overall resource optimization strategy.

[0100] In one feasible implementation, refer to Figure 4 Step S300 includes steps S310 to S340, wherein:

[0101] Step S310: Parse the number of available transmission bits from the channel capacity indication information, and calculate the average number of coded bits per feature dimension based on the locally stored historical transmission records;

[0102] Step S320: Calculate the maximum number of transmittable dimensions based on the available number of transmittable bits and the average number of encoded bits per single feature dimension;

[0103] Step S330: Sort the feature dimensions in descending order according to the feature importance vector, and select the top K dimensions as the set of retained feature dimensions, where K is equal to the maximum number of transmittable dimensions.

[0104] Step S340: Construct a binary mask matrix by setting the corresponding positions of the retained feature dimension set to 1 and the rest to 0 to generate the binary feature mask.

[0105] In this embodiment, the channel capacity indication information refers to the numerical index of currently available wireless transmission resources quantified by the base station through the control channel. Specifically, it can be implemented by mapping the modulation and coding scheme index value carried in the physical layer control signaling to the number of bits. Its function is to provide data transmission capacity constraints for terminal devices under the current channel conditions. The average number of coded bits per feature dimension refers to the statistical average of the number of coded bits actually consumed by a single feature dimension during historical transmissions. Specifically, it can be implemented by using a sliding window to count the actual number of bits used by each feature dimension in the most recent N transmissions and calculating the average value. Its function is to provide a quantitative basis for dynamically adjusting the number of feature transmissions. The maximum number of transmittable dimensions refers to the upper limit of the number of feature dimensions allowed to be transmitted under the current channel capacity. Specifically, it can be implemented by dividing the available transmission bits by the average number of coded bits per feature dimension and then rounding down. Its function is to ensure that the amount of transmitted data does not exceed the channel carrying capacity. The descending order of feature importance vectors refers to arranging feature dimensions from high to low according to the degree of influence of features on network performance prediction. Specifically, it can be implemented by using a quicksort algorithm to sort the feature importance vectors. Its function is to prioritize retaining features that have a greater impact on resource optimization decisions.

[0106] In this embodiment, after the channel capacity indication information is parsed, the terminal device calculates the average number of encoded bits consumed by each feature dimension during past transmissions by querying locally stored historical transmission records. For example, if the historical record shows that a certain feature dimension requires an average of 3 bits for encoding, and the currently available transmission bits are 120 bits, then the maximum number of transmittable dimensions is 40. Subsequently, the feature importance vectors are sorted. For example, if a feature importance vector contains weight values ​​for 50 dimensions, the first 40 dimensions are selected as the retention set after being sorted in descending order. During the generation of the binary mask matrix, the matrix positions corresponding to the retained feature dimensions are set to 1, and the remaining positions are set to 0. For example, a binary sequence of length 50 is generated, with the first 40 bits being 1 and the last 10 bits being 0.

[0107] Compared to existing technologies, traditional methods often employ fixed feature selection strategies or ignore dynamic changes in channel capacity, leading to data transmission exceeding channel capacity or the omission of important features. For example, existing technologies may statically select only the top 30 dimensions based on feature importance, but this fails to fully utilize bandwidth when channel capacity is sufficient, or forces the discard of critical features when capacity is insufficient. This solution dynamically calculates the maximum number of transmittable dimensions, combining real-time channel capacity and feature importance ranking to achieve adaptive matching between the number of transmitted features and channel conditions. Through the above technical solution, this application solves the technical problem of balancing feature transmission integrity and communication resource constraints under limited channel capacity. By dynamically calculating the maximum number of transmittable dimensions, it ensures that the transmitted data volume strictly conforms to the current channel capacity constraints; through a dimension selection mechanism based on feature importance ranking, it prioritizes the transmission of features with greater impact on resource optimization decisions, avoiding invalid features occupying channel resources. For example, in scenarios with fluctuating channel capacity, the number of transmitted dimensions is automatically reduced when the number of available bits decreases, but features with high importance ranking are always retained, thereby reducing communication overhead while maintaining decision accuracy.

[0108] In one feasible implementation, refer to Figure 5 Step S400 includes steps S410 to S430, wherein:

[0109] Step S410: Query the preset noise intensity coefficient mapping table according to the feature type of the key feature subset, obtain the noise intensity coefficient of the current feature, and generate a Laplace distribution random number based on the noise intensity coefficient.

[0110] Step S420: The random number is superimposed on the numerical part of the key feature subset;

[0111] Step S430: Perform boundary truncation processing on the superimposed feature values ​​to generate the privacy-preserving feature data.

[0112] In this embodiment, the noise intensity coefficient mapping table refers to a pre-established lookup table containing the correspondence between feature types and noise intensities. It can be constructed using offline training or expert experience and is used to dynamically adjust the noise intensity based on the differences in privacy sensitivity of different features. Laplace distributed random numbers refer to random variables that conform to the Laplace probability density function. They can be generated using inverse transformation or lookup table methods, and their probability distribution characteristics meet the requirements of differential privacy protection. Boundary truncation refers to the operation of thresholding feature values ​​that exceed a preset numerical range. This can be implemented using piecewise linear functions or saturation functions to prevent feature values ​​from exceeding a reasonable range after noise superposition, thus preventing data distortion.

[0113] In this embodiment, before uploading feature data, the corresponding noise intensity coefficient is first matched in a preset mapping table according to the feature type. For example, a smaller coefficient can be configured for signal strength features to maintain data availability, while a larger coefficient can be configured for energy consumption features to enhance privacy protection. Laplace noise is generated based on the matched coefficients. For example, for a feature dimension with a coefficient of 0.3, its noise value can come from a Laplace distribution with a scale parameter of 0.3. After the noise value is superimposed on the original feature value, the values ​​exceeding the preset upper and lower limits are forcibly constrained to the boundary values ​​through boundary truncation. For example, feature values ​​exceeding the [0,1] interval are set to 0 or 1, ensuring that the processed data satisfies privacy requirements while retaining effective statistical characteristics.

[0114] In this embodiment, the proposed solution achieves a dynamic balance between privacy protection and data availability through a feature type-adaptive noise intensity adjustment mechanism. At the same time, it effectively suppresses abnormal feature value shifts caused by noise interference by combining boundary truncation operations. This enables fine-grained privacy protection within the federated learning framework. Differentiated noise perturbations are applied to different feature types, which avoids the risk of privacy leakage or data distortion caused by uniform noise intensity. Furthermore, boundary truncation maintains the effective distribution of feature data, providing a reliable data foundation for subsequent cross-device feature alignment and global resource optimization.

[0115] In one feasible implementation, refer to Figure 6 Step S500 includes steps S510 to S530, wherein:

[0116] Step S510: parse the dimensional description information of the privacy protection feature data uploaded by each terminal, extract the correspondence between feature names and dimensional indices; perform cross-terminal dimensional matching based on feature name similarity, and generate a terminal feature dimension mapping table.

[0117] Step S520: Adjust the dimensional order of privacy protection feature data of each terminal according to the terminal feature dimension mapping relationship table;

[0118] Step S530: Detect the location of missing dimensions in the adjusted data, and fill the missing dimension location with the global feature historical mean of the most recent preset first number of historical data collection periods to generate the aligned feature dataset.

[0119] In this embodiment, dimensional description information refers to metadata containing feature names, data types, and dimension position indices. This metadata can be implemented using JSON format or a protocol buffer and is used to identify the semantic information of feature vectors from different terminals. Cross-terminal dimensional matching refers to establishing feature correspondences between different devices by calculating the textual similarity of feature names. This can be implemented using a cosine similarity algorithm or an edit distance algorithm to solve the dimensional misalignment problem caused by differences in feature naming rules across different terminals. The global feature historical mean refers to the average value of all terminal feature dimensions stored in the aggregation node within the previous period. This value can be dynamically updated through a sliding window mechanism to provide a reasonable filler value when features are missing.

[0120] In this embodiment, during the cross-device feature alignment process, the feature metadata carried in the data uploaded by each terminal is first parsed to extract the mapping relationship between feature names and dimensional positions. By calculating the text similarity between feature names from different terminals—for example, by using a word vector model to convert feature names into vectors and then calculating cosine similarity—a cross-terminal feature correspondence table is established. Based on this table, the dimensional order of the feature vectors from each terminal is rearranged so that the same semantic features are in the same dimensional position in the data from different terminals. For feature loss due to device heterogeneity, such as a terminal not collecting specific sensor data, the global historical average of that feature dimension over the most recent several periods is used to fill the gaps, for example, the average of the most recent three historical data collection periods, thereby ensuring the integrity of the dataset.

[0121] In this embodiment, the solution effectively eliminates the data dimension inconsistency problem caused by device heterogeneity through a dynamic feature matching mechanism and a missing value imputation strategy based on historical averages. This improves the semantic consistency of the federated feature space, solves the dimension alignment difficulty caused by inconsistent feature naming rules of multiple terminals, and the feature missing problem caused by differences in device sensor configuration. It ensures that subsequent weighted fusion operations are performed in a unified dimension space, improves the integrity and availability of the global feature matrix, and provides an accurate data foundation for the generation of network resource strategies.

[0122] In one feasible implementation, refer to Figure 7 Step S600 includes steps S610 to S630, wherein:

[0123] Step S610: Count the number of local data samples of each terminal, and determine the aggregation weight coefficient of each terminal based on the number of local data samples and the total number of samples.

[0124] Step S620: Perform a weighted calculation on the feature values ​​of each feature dimension in the aligned feature dataset;

[0125] Step S630: Perform standard deviation normalization on the weighted calculation results and integrate the normalized values ​​of all dimensions to generate the global feature space matrix.

[0126] In this embodiment, the number of local data samples refers to the total number of data entries collected locally by the terminal device and used for feature generation. Specifically, this can be achieved by counting the number of samples in the data packets uploaded by each terminal using a counter, reflecting the scale of data contribution from different terminals. The aggregation weight coefficient is a parameter used to adjust the proportion of influence of each terminal's data in global fusion. Specifically, it can be calculated as the ratio of the number of local samples to the sum of the number of samples from all terminals, ensuring that terminals with larger data volumes have a higher weight in the global features. Weighted calculation refers to multiplying the data values ​​of each feature dimension from each terminal by their corresponding weight coefficient and then summing them. This can be achieved through matrix multiplication, ensuring that data sources of different sizes have a reasonable impact on the global features. Standard deviation normalization refers to scaling the weighted feature values ​​based on statistical distribution. Specifically, this can be achieved by calculating the global standard deviation and then performing a linear transformation by dividing each feature value by the standard deviation, eliminating dimensional differences between different feature dimensions.

[0127] In this embodiment, after the edge aggregation node completes cross-device feature alignment, it first iterates through the privacy-preserving feature data uploaded by all terminal devices and extracts the number of local data samples uploaded by each terminal during the current aggregation cycle. The total number of samples is obtained by summing the sample counts of each terminal, and the aggregation weight coefficient for that terminal is obtained by dividing the sample count of a single terminal by the total number of samples. For each feature dimension in the aligned feature dataset, the feature value of each terminal in that dimension is multiplied by the corresponding weight coefficient and then summed to obtain the weighted fusion value for that dimension. The standard deviation of the weighted fusion values ​​of all terminals in that dimension is calculated, and the normalized value is obtained by dividing the weighted fusion value of each dimension by the corresponding standard deviation. The normalized values ​​of all feature dimensions are arranged and combined in the original dimensional order to form a feature space matrix containing global statistical characteristics.

[0128] In this embodiment, by dynamically calculating weight coefficients based on data sample size, the data contribution of each terminal can be accurately reflected, avoiding excessive interference from terminals with small data volumes on global features. Simultaneously, standard deviation normalization is introduced to effectively eliminate fusion bias caused by different feature dimensions due to varying scales. Compared to the direct feature concatenation method in existing technologies, this significantly improves the data consistency of the global feature space. Thus, this application can dynamically adjust the fusion weights according to the actual data scale of the terminals, improving the accuracy of the global feature space in representing network resource distribution and solving the feature bias problem caused by traditional average fusion methods in scenarios with uneven terminal data volumes. Standard deviation normalization eliminates statistical distribution differences between multi-dimensional features, enhancing the stability of the global feature matrix in subsequent resource allocation decisions and reducing the risk of strategy oscillations caused by inconsistent feature scales.

[0129] In one feasible implementation, refer to Figure 8 Step S700 includes steps S710 to S750, wherein:

[0130] Step S710: Obtain the geographic location coordinate data reported by the terminal device, and calculate the Euclidean distance matrix between the terminal devices based on the geographic location coordinate data;

[0131] Step S720: Construct a spatiotemporal correlation matrix based on the Euclidean distance matrix and the device service type;

[0132] Step S730: Perform cluster analysis on the spatiotemporal correlation matrix to identify communication device groups, and allocate a channel bandwidth ratio to each group based on the communication device groups;

[0133] Step S740: Generate a terminal scheduling priority sequence based on the urgency of the service;

[0134] Step S750: Combine the channel bandwidth allocation ratio and the terminal scheduling priority sequence to generate the network resource strategy.

[0135] In this embodiment, geographic location coordinate data refers to the latitude and longitude coordinates collected by the terminal device through its built-in positioning module. This can be implemented using a GPS module or base station triangulation technology to determine the physical location relationship between devices, providing basic data for subsequent calculations of spatial correlation. The Euclidean distance matrix is ​​a symmetric matrix generated by calculating the square root of the sum of the squares of the differences in geographic location coordinates between any two devices. This can be implemented using a matrix operation library and is used to quantify the spatial distance between devices, reflecting the degree of potential signal interference. The spatiotemporal correlation matrix is ​​a multidimensional matrix formed by weighted concatenation of the Euclidean distance matrix and device service type features. This can be implemented using feature concatenation and normalization methods and is used to comprehensively characterize the correlation between devices in spatial distribution and service requirements. Communication device grouping refers to the set of devices formed by classifying the spatiotemporal correlation matrix using a clustering algorithm. This can be implemented using K-means or hierarchical clustering algorithms and is used to identify groups of devices with similar communication needs, providing a basis for channel allocation. The channel bandwidth ratio refers to the channel resource allocation weight dynamically calculated based on the number of devices and service types within a group. This can be implemented using a proportional allocation algorithm to ensure that the resource requirements of different groups are met in a differentiated manner. Terminal scheduling priority sequence refers to a list of device communication order generated according to the urgency of services. It can be implemented using a priority queue data structure to prioritize the transmission needs of high-urgency services when resources are limited.

[0136] In this embodiment, after the terminal device uploads the collected geographic coordinates to the edge aggregation node, it first calculates an Euclidean distance matrix using coordinate differences. This matrix reflects the spatial density of the devices. Further, it combines device service type data, such as service tags for real-time video transmission and IoT sensor data, to encode the service type into a feature vector and weightedly fuse it with the distance matrix to form a spatiotemporal correlation matrix. A clustering algorithm is used to analyze the spatiotemporal correlation matrix, grouping devices with similar spatial distributions and service requirements into the same communication group. Based on the number of devices and the data throughput requirements of the service type within each group, the channel bandwidth allocation ratio for each group is dynamically calculated. Simultaneously, according to the urgency parameters defined in the device service type (e.g., medical monitoring devices have the highest priority), a terminal scheduling sequence is generated. Finally, the channel allocation ratio and the scheduling sequence are merged to form a network resource strategy that balances spatial resource utilization efficiency and service timeliness.

[0137] Understandably, existing technologies typically allocate resources solely based on signal strength or service type, without fully considering the impact of device spatial distribution on channel interference. For example, existing technologies may only generate scheduling sequences based on the urgency of device services, but fail to address the issue of reduced transmission efficiency caused by channel contention among geographically proximate devices. This solution, however, constructs a spatiotemporal correlation matrix to jointly analyze device physical location and service characteristics. This allows channel allocation to avoid signal interference from spatially adjacent devices while matching the resource requirements of different service types, effectively reducing the probability of channel conflicts caused by spatial proximity between devices and improving the rationality of channel resource allocation. Simultaneously, by combining service urgency with the generation of scheduling sequences, it ensures timely transmission of high-priority services, optimizing overall network service quality in resource-constrained scenarios. Furthermore, the clustering-based channel allocation strategy can dynamically adapt to changes in device distribution, enhancing the robustness of the resource management solution.

[0138] In one feasible implementation, refer to Figure 9 The method further includes steps S910 to S930, wherein:

[0139] Step S910: Calculate the cosine similarity between the network resource strategies for a consecutive preset second number of decision periods;

[0140] Step S920: When the cosine similarity is less than a preset threshold, a sparsity adjustment instruction is generated based on the cosine similarity and sent to the terminal device.

[0141] In step S930, the terminal device updates the sliding window size in the feature importance analysis according to the sparsity adjustment instruction.

[0142] In this embodiment, cosine similarity refers to the degree of similarity between network resource strategies of different periods measured by a vector space model. Specifically, it can be calculated using the ratio of the vector inner product to the product of the magnitudes, and is used to detect the drasticness of strategy changes. The preset second quantity refers to the number of consecutive periods used for strategy similarity comparison. Specifically, the number of periods can be dynamically adjusted according to the network environment, for example, set to 3 to 5 periods. The sparsity adjustment instruction refers to the parameter update instruction triggered by changes in strategy similarity. Specifically, it can be generated as a binary trigger signal through a preset threshold comparison, used to control the sensitivity of feature importance analysis. The sliding window size refers to the length of the time window used for smoothing in feature importance analysis. Specifically, a circular buffer can be used to store historical data, and the stability of feature importance evaluation can be changed by adjusting the window length.

[0143] In this embodiment, the vector representation of network resource policies is extracted and similarity is calculated over multiple consecutive decision cycles. When the policy similarity falls below a preset threshold, it indicates a significant change in the network environment or device state, at which point an adjustment command is sent to the terminal device. Upon receiving the command, the terminal device dynamically increases or decreases the sliding window size in feature importance analysis, thereby altering the smoothness of feature evaluation. For example, the window size is increased to enhance stability when similarity is low, and decreased to improve response speed when similarity is high.

[0144] In some specific implementations, cosine similarity calculation can employ a sliding window mechanism, retaining the strategy vectors from the most recent N periods during each calculation. The generation of sparsity adjustment instructions can incorporate similarity change gradients, triggering adjustments when the gradient exceeds a set range. Updating the sliding window size can be done by setting an incremental step size, such as increasing or decreasing the window length by 10% with each adjustment, while simultaneously setting maximum and minimum value constraints.

[0145] Compared to existing technologies, traditional methods use feature analysis windows with fixed parameters, which cannot adapt to dynamic network environment changes. In existing technologies, the sliding window size is usually set to an empirical value, leading to inaccurate feature importance assessment when the network state changes abruptly. This solution, however, achieves automatic adaptation of feature analysis parameters through a dynamic feedback mechanism based on policy similarity, solving the problem of mismatch between static parameters and the dynamic environment. This allows the application to automatically optimize feature analysis parameters according to the changing trends of network resource policies, improving the response speed to network environment changes while ensuring the stability of feature importance assessment. This dynamic adjustment mechanism effectively avoids the problem of resource allocation policy lag caused by sudden environmental changes, improves the real-time performance and accuracy of wireless network resource optimization, and reduces the additional communication overhead caused by parameter mismatch.

[0146] In the embodiments of this application, the joint optimization method for wireless network resources based on federated feature sparsity reduces the amount of transmitted data through feature sparsity processing, protects privacy by combining noise perturbation, and dynamically generates resource allocation strategies based on global feature analysis. This can improve resource allocation efficiency and reduce communication overhead, while enhancing privacy protection capabilities.

[0147] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the joint optimization method for wireless network resources based on federated feature sparsity. Any simple modifications based on this technical concept are within the protection scope of this application.

[0148] This application also provides a joint optimization system for wireless network resources based on federated feature sparsity, referencing... Figure 10The system includes: a memory 10, a processor 20, and a joint optimization program for wireless network resources based on federated feature sparsity stored on the memory 10 and executable on the processor 20. The joint optimization program for wireless network resources based on federated feature sparsity is configured to implement the steps of the joint optimization method for wireless network resources based on federated feature sparsity.

[0149] The wireless network resource joint optimization system based on federated feature sparsity provided in this application employs the wireless network resource joint optimization method based on federated feature sparsity in the above embodiments, which can improve resource allocation efficiency and reduce communication overhead. Compared with the prior art, the beneficial effects of the wireless network resource joint optimization system based on federated feature sparsity provided in this application are the same as those of the wireless network resource joint optimization method based on federated feature sparsity provided in the above embodiments, and other technical features of the wireless network resource joint optimization system based on federated feature sparsity are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0150] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0151] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A joint optimization method for wireless network resources based on federated feature sparsity, characterized in that, The method includes: The terminal device collects the raw dataset of local network status and performs feature extraction and standardization on the raw dataset to generate a standardized feature vector; The terminal device performs a feature importance analysis operation based on the standardized feature vector to generate a feature importance vector; The terminal device receives channel capacity indication information sent by the base station, and performs a sparse mask generation operation in combination with the feature importance vector to generate a binary feature mask. The terminal device extracts a subset of key features from the standardized feature vector based on the binary feature mask, adds noise perturbation to the subset of key features to generate privacy-preserving feature data, and uploads the privacy-preserving feature data to the edge aggregation node. The edge aggregation node receives the privacy protection feature data uploaded by multiple terminals, performs cross-device feature alignment operation on the privacy protection feature data, and generates an aligned feature dataset; A weighted fusion process is performed on the aligned feature dataset to generate a global feature space matrix; Based on the global feature space matrix analysis of the device spatiotemporal correlation, a network resource strategy including channel allocation parameters and terminal scheduling sequence is generated. The base station controller executes the network resource policy and broadcasts the updated channel state parameters to the terminal device; The edge aggregation node receives the privacy-preserving feature data uploaded by multiple terminals, and performs cross-device feature alignment on the privacy-preserving feature data to generate an aligned feature dataset. The steps include: The dimensional description information of the privacy protection feature data uploaded by each terminal is parsed, and the correspondence between feature names and dimension indices is extracted; cross-terminal dimension matching is performed based on feature name similarity to generate a terminal feature dimension mapping table. Adjust the dimensional order of privacy protection feature data for each terminal according to the terminal feature dimension mapping relationship table; The missing dimension locations in the adjusted data are detected, and the missing dimension locations are filled with the historical average of global features from the most recent preset first number of historical data collection periods to generate the aligned feature dataset.

2. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 1, characterized in that, The steps of the terminal device collecting the original dataset of local network status and performing feature extraction and standardization on the original dataset to generate standardized feature vectors include: Terminal devices collect raw data on local network signal strength, transmission delay, and device power consumption. The original data is segmented using a preset time window, and time-domain statistical features are calculated within the time window; Perform Fourier transform on the data within the time window to extract the energy features of the principal components in the frequency domain; The time-domain statistical features and the frequency-domain principal component energy features are respectively subjected to minimum-maximum normalization, and the normalized time-domain statistical features and frequency-domain principal component energy features are concatenated to generate the standardized feature vector.

3. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 1, characterized in that, The steps by which the terminal device performs feature importance analysis based on the standardized feature vector to generate the feature importance vector include: The standardized feature vector is copied to generate a test vector copy, and the specified feature dimension value in the test vector copy is set to zero to generate a masked feature vector; Based on a preset network performance prediction function, calculate the predicted network throughput value corresponding to the shielding feature vector; Calculate the deviation between the baseline network throughput prediction value of the standardized feature vector and the network throughput prediction value of the masked feature vector; An initial sequence of feature importance is generated based on the deviation, and a sliding window smoothing process is performed on the initial sequence of feature importance to generate the feature importance vector.

4. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 1, characterized in that, The terminal device receives channel capacity indication information from the base station and performs a sparse mask generation operation in conjunction with the feature importance vector. The steps for generating a binary feature mask include: The available transmission bits are parsed from the channel capacity indication information, and the average number of coded bits per feature dimension is calculated based on the locally stored historical transmission records. The maximum number of transmittable dimensions is calculated based on the available number of transmittable bits and the average number of encoded bits per single feature dimension. The feature dimensions are sorted in descending order according to the feature importance vector, and the top K dimensions are selected as the set of retained feature dimensions, where K is equal to the maximum number of transmittable dimensions. Construct a binary mask matrix by setting the corresponding positions of the reserved feature dimension set to 1 and the rest to 0 to generate the binary feature mask.

5. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 1, characterized in that, The steps for generating privacy-preserving feature data by adding noise perturbation to the subset of key features include: Based on the feature type of the key feature subset, query the preset noise intensity coefficient mapping table to obtain the noise intensity coefficient of the current feature, and generate a Laplace distribution random number based on the noise intensity coefficient. The random number is superimposed onto the numerical portion of the key feature subset; Boundary truncation is performed on the superimposed feature values ​​to generate the privacy-preserving feature data.

6. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 1, characterized in that, The steps of performing weighted fusion processing on the aligned feature dataset to generate a global feature space matrix include: The number of local data samples for each terminal is counted, and the aggregation weight coefficient for each terminal is determined based on the number of local data samples and the total number of samples. Perform a weighted calculation on the feature values ​​of each feature dimension in the aligned feature dataset; The weighted calculation results are normalized by standard deviation, and the normalized values ​​of all dimensions are integrated to generate the global feature space matrix.

7. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 1, characterized in that, The steps for generating a network resource strategy that includes channel allocation parameters and terminal scheduling sequences based on the analysis of the spatiotemporal correlation of devices using the global feature space matrix include: Obtain the geographic location coordinate data reported by the terminal device, and calculate the Euclidean distance matrix between the terminal devices based on the geographic location coordinate data; Construct a spatiotemporal correlation matrix based on the Euclidean distance matrix and the device service type; Cluster analysis of the spatiotemporal correlation matrix is ​​performed to identify communication device groups, and a channel bandwidth ratio is allocated to each group based on the communication device groups; Generate a terminal scheduling priority sequence based on the urgency of the service; The network resource strategy is generated by integrating the channel bandwidth allocation ratio and the terminal scheduling priority sequence.

8. The joint optimization method for wireless network resources based on federated feature sparsity as described in claim 3, characterized in that, The method further includes: Calculate the cosine similarity between the network resource strategies for a second consecutive predetermined number of decision periods; When the cosine similarity is less than a preset threshold, a sparsity adjustment instruction is generated based on the cosine similarity and sent to the terminal device; The terminal device updates the sliding window size in the feature importance analysis according to the sparsity adjustment instruction.

9. A joint optimization system for wireless network resources based on federated feature sparsity, characterized in that, The system includes: a memory, a processor, and a federated feature sparsity-based wireless network resource joint optimization program stored in the memory and executable on the processor, the federated feature sparsity-based wireless network resource joint optimization program being configured to implement the steps of the federated feature sparsity-based wireless network resource joint optimization method as described in any one of claims 1 to 8.

Images