Service quality control method and device, electronic equipment and storage medium

Abstract

The invention provides a service quality control method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining multi-modal data corresponding to at least one influence factor of a service quality QoS index; for any influence factor, performing feature extraction on the multi-modal data corresponding to the influence factor to obtain a multi-modal feature; and on the basis of the multi-modal feature and the QoS index corresponding to any influence factor, constructing a causal graph model so as to adjust a QoS strategy according to the causal graph model.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a service quality control method, apparatus, electronic device, and storage medium. Background Technology

[0002] The 6th Generation Mobile Communications (6G) network will give rise to new services and businesses such as intelligent agent interaction, communication sensing, and inclusive intelligence. It will also spur a series of innovative applications, including holographic interaction, digital twins, super-powered transportation, smart industry and agriculture, the metaverse, the tactile internet, and low-altitude connectivity. These services are diverse and have varied requirements. The 6G network needs to enhance existing communications while providing comprehensive, high-quality services across multiple areas, including sensing, computing power, and data, for future 6G services. Therefore, compared to the centralized data control of traditional communication services at a single node, 6G services require distributed control to ensure service quality. Summary of the Invention

[0003] This application aims to at least partially address one of the technical problems in the related art.

[0004] Therefore, the first objective of this application is to propose a service quality control method.

[0005] The second objective of this application is to provide a service quality control device.

[0006] The third objective of this application is to propose an electronic device.

[0007] The fourth objective of this application is to provide a computer-readable storage medium.

[0008] The fifth objective of this application is to provide a computer program product.

[0009] To achieve the above objectives, a service quality control method is proposed in the first aspect of this application, comprising:

[0010] Obtain multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) metric;

[0011] For any of the aforementioned influencing factors, feature extraction is performed on the multimodal data corresponding to the influencing factor to obtain multimodal features;

[0012] Based on the multimodal features and QoS indicators corresponding to any of the aforementioned influencing factors, a causal graph model is constructed to adjust the QoS strategy according to the causal graph model.

[0013] To achieve the above objectives, a second aspect of this application provides a service quality control device, comprising:

[0014] The acquisition module is used to acquire multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) index.

[0015] The extraction module is used to extract features from the multimodal data corresponding to any one of the influencing factors to obtain multimodal features;

[0016] A construction module is used to construct a causal graph model based on the multimodal features corresponding to any of the influencing factors and the QoS indicators, so as to adjust the QoS strategy according to the causal graph model.

[0017] To achieve the above objectives, a third aspect of this application provides an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement a service quality control method as described in the first aspect of this application.

[0018] To achieve the above objectives, a fourth aspect of this application provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement a quality of service control method as described in the first aspect of this application.

[0019] To achieve the above objectives, a fifth aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements a service quality control method as described in the first aspect of this application.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0022] Figure 1 A schematic flowchart illustrating a service quality control method provided in an embodiment of this application;

[0023] Figure 2 A flowchart illustrating a service quality control method provided in another embodiment of this application;

[0024] Figure 3This is a schematic diagram of the structure of a service quality control device provided in an embodiment of this application;

[0025] Figure 4 This is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0027] In related technologies, Bayesian network models are applied to QoS assurance. For example, in Software-Defined Networking (SDN), Bayesian networks are used to predict link congestion probability and dynamically adjust traffic routing; in 5G core networks, Bayesian networks are used to analyze the correlation between network device status (such as router load, link quality, etc.) and QoS indicators, analyze base stations, transmission links, and infer the posterior probability of faulty nodes through probability, and adjust QoS policies in real time; in video streaming services, the encoding bit rate is dynamically adjusted according to device performance and network conditions; and in cloud computing environments, Bayesian networks are used to analyze traffic characteristics and distinguish between normal business peaks and malicious attacks.

[0028] However, Bayesian networks rely on prior probabilities, and inaccurate prior knowledge can negatively impact model performance. Furthermore, 6G distributed QoS assurance requires the collection of large-scale, diverse data, posing a significant challenge to building a prior knowledge base. Additionally, Bayesian-based QoS assurance techniques struggle to meet the time-varying characteristics of multimodal data streams in 6G services.

[0029] To address the aforementioned problems, this application proposes a service quality control method, apparatus, electronic device, and storage medium. The service quality control method, apparatus, electronic device, and storage medium of this application are described below with reference to the accompanying drawings.

[0030] Figure 1 This is a flowchart illustrating a quality of service control method provided in an embodiment of this application. This quality of service control method can be applied to 6G nodes such as base stations.

[0031] like Figure 1 As shown, the service quality control method includes the following steps:

[0032] Step 101: Obtain multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) metric.

[0033] The number of QoS indicators can be one or more. For example, QoS indicators can include network latency, network jitter, packet loss rate, bit error rate, etc.

[0034] Among them, the influencing factors of QoS indicators refer to variables that directly or indirectly change the value of QoS indicators. For example, the influencing factors of network latency include variables such as user density and network bandwidth.

[0035] Multimodal data includes data in at least one modality, such as text data, image data, and numerical data.

[0036] For example, image data may include visual modal data such as base station traffic heatmaps and network topology images generated by cameras or network digital twin systems; text data may include text modal data such as user equipment logs, network equipment logs, QoS requirement instructions, and network status descriptions; numerical data may include numerical modal data such as channel state matrices, real-time bandwidth data, and user density statistics.

[0037] It should be noted that the influencing factors corresponding to the above data may not be the same. After collecting multimodal data, multimodal data corresponding to each influencing factor of each QoS indicator can be obtained from the collected data. For example, the multimodal data corresponding to user density may include base station traffic heatmaps and statistical user density values.

[0038] Step 102: For any influencing factor, extract features from the multimodal data corresponding to the influencing factor to obtain multimodal features.

[0039] Among them, different modal dimensions have corresponding feature extraction algorithms. The feature extraction algorithms corresponding to each modal dimension can be used to extract features from the corresponding modal data to obtain single-modal features. Then, feature fusion is performed based on the extracted single-modal features to obtain multimodal features.

[0040] For example, for image modalities, an encoder (such as a VisionTransformer (ViT) encoder) can be used to extract features from the image data, as shown in the following formula:

[0041] h image =ViT(I)

[0042] Among them, h image The extracted image features are represented by the dimension d, and I represents the image data from which features are to be extracted.

[0043] For text modalities, text encoding networks (such as Bidirectional Encoder Representations from Transformers, BERT) can be used to extract features from text data. The specific formula is as follows:

[0044] h text =BERT(T)

[0045] Among them, h text The extracted text features are represented by the dimension d, and T represents the text data from which features are to be extracted.

[0046] For numerical modes, a multilayer perceptron (MLP) encoder can be used to extract features from the numerical data. The specific formula is as follows:

[0047] h data =MLP(X num )

[0048] Among them, h data This represents the extracted numerical features, with feature dimension d and X. num This represents the numerical data for which feature extraction is to be performed.

[0049] For example, feature fusion can be performed on the extracted single-modal features through residual connections to obtain multimodal features, as shown in the following formula:

[0050] h fused =LayerNorm(h image +h text +h data )

[0051] Where LayerNorm represents layer normalization, h fused This represents the multimodal features obtained after residual connection and layer normalization.

[0052] Step 103: Based on the multimodal characteristics and QoS indicators corresponding to any influencing factor, construct a causal graph model to adjust the QoS strategy according to the causal graph model.

[0053] Specifically, a causal graph model can be constructed based on the causal relationship between influencing factors and QoS indicators, using the multimodal characteristics of the influencing factors and the QoS indicators. This causal graph model is used to output QoS indicator values ​​in real time.

[0054] For example, for any QoS metric, a causal graph model can be constructed based on the influencing factors corresponding to that QoS metric; that is, a causal graph model is associated with a QoS metric.

[0055] For example, for at least two QoS indicators whose influencing factors overlap, a causal graph model is constructed based on the influencing factors corresponding to the at least two QoS indicators; that is, a causal graph model is associated with at least two QoS indicators.

[0056] As an example, QoS policy adjustments based on the cause-effect graph model include: setting threshold values ​​for QoS indicators; and triggering QoS policy adjustments when the QoS indicator value deviates from the threshold value, i.e., when the matching relationship between the QoS indicator value and the threshold value does not meet the set conditions.

[0057] For example, if network latency is detected to be higher than the latency threshold during video playback, resulting in video stuttering, the QoS policy can be adjusted to prioritize video stream bandwidth and reduce bandwidth usage by non-critical applications.

[0058] For example, if network latency is detected to be lower than the latency threshold in a video playback scenario and network bandwidth resources are sufficient, the QoS policy can dynamically trigger a quality optimization mechanism, such as increasing the resolution of the video stream.

[0059] In this embodiment, multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) index is acquired. For any influencing factor, feature extraction is performed on the multimodal data corresponding to the influencing factor to obtain multimodal features. Based on the multimodal features corresponding to any influencing factor and the QoS index, a causal graph model is constructed to adjust the QoS policy according to the causal graph model. By combining multimodal data and QoS index to construct a causal graph model, the reliance on inaccurate prior probabilities and the impact of prior knowledge bias on QoS adjustment performance are avoided. At the same time, the causal graph model can flexibly adapt to the time-varying characteristics of multimodal data streams without the need to build a large prior knowledge base, thus enabling more accurate dynamic adjustment of the QoS policy.

[0060] This embodiment provides another service quality control method. Figure 2 This is a flowchart illustrating a service quality control method provided in an embodiment of this application.

[0061] like Figure 2 As shown, the service quality control method may include the following steps:

[0062] Step 201: Obtain multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) index.

[0063] In one possible embodiment of this application, a cubic spline interpolation algorithm is used to perform timestamp alignment processing on multimodal data.

[0064] For example, the calculation formula for the cubic spline interpolation algorithm is as follows:

[0065]

[0066] Where t is the target timestamp, a n The basis function coefficients are determined by least squares fitting, and x(t) represents the time-aligned data after interpolation. Here, h is the cubic spline basis function, which is the interpolation step size used to control the smoothness, and t is the interpolation step size. n Let N be the timestamp of the nth original data point, and N be the total number of original data points.

[0067] A cubic spline interpolation algorithm is used to perform timestamp alignment on multimodal data, which can achieve timestamp alignment even when the multimodal data is collected asynchronously.

[0068] In one possible embodiment of this application, when the multimodal data includes numerical data, the K-nearest neighbor algorithm is used to fill in missing data in the numerical data.

[0069] For example, the calculation formula for the K-nearest neighbor algorithm is as follows:

[0070]

[0071] Where, x fill The supplementary data is provided, where k represents the number of nearest neighbors. Let be the value of the k-th nearest neighbor.

[0072] By using the K-nearest neighbor algorithm to complete missing data, the missing values ​​can be accurately restored by utilizing the feature information of the nearest samples, providing a complete and reliable data foundation for QoS policy adjustment.

[0073] It should be noted that when multimodal data includes numerical data, the numerical data can be normalized, and feature extraction can be performed based on the normalized data. For example, the calculation formula for normalization is as follows:

[0074]

[0075] Where, x norm This represents the normalized data, where x is the original collected numerical data, μ is the mean of the numerical data, and σ is the standard deviation of the numerical data.

[0076] Step 202: For any influencing factor, extract features from the multimodal data corresponding to the influencing factor to obtain multimodal features.

[0077] In one possible embodiment of this application, the multimodal data corresponds to multiple modal dimensions.

[0078] Modal dimensions include image dimensions, text dimensions, and numerical dimensions.

[0079] In one possible embodiment of this application, feature extraction is performed on modal data under any modal dimension in multimodal data to obtain single-modal features under the corresponding modal dimension; semantic alignment is performed on single-modal features under different modal dimensions; and feature fusion is performed on the semantically aligned single-modal features to obtain multimodal features.

[0080] Specifically, feature extraction algorithms corresponding to each modal dimension can be used to extract features from modal data in multimodal data under each modal dimension, thereby obtaining the corresponding single-modal features. For example, the image dimension corresponds to the ViT encoder, the text dimension corresponds to the BERT encoding model, and the numerical dimension corresponds to the MLP encoder.

[0081] One approach is to use a semantic alignment model. This model performs semantic alignment on single-modal features across different modal dimensions. As an example, the semantic alignment model is trained using contrastive loss, which is calculated as follows:

[0082]

[0083] Among them, L align To compare the losses, h i h represents the single-modal feature corresponding to the i-th modal dimension. j h represents the single-modal feature corresponding to the j-th modal dimension. k is the single-modal feature corresponding to the j-th modal dimension; τ is a hyperparameter used to control the distribution degree; M is the number of negative samples; and sim() is the cosine similarity function.

[0084] As an example, the formula for calculating multimodal features is as follows:

[0085] h fused =LayerNorm(h′) image +h′ text +h′ data )

[0086] Among them, h fused h′ represents the multimodal features obtained after residual connections and layer normalization. image h′ text h′ data These are semantically aligned single-modal features.

[0087] As another example, feature fusion is performed on the semantically aligned single-modal features to obtain fused features; a multi-head attention mechanism is used to perform cross-modal weighted fusion processing on the fused features to obtain multimodal features.

[0088] Among them, the multimodal features h fused Split into Q, K, V matrices.

[0089] Q = h fused W Q

[0090] K = h fused W K

[0091] V = h fused W v

[0092] Among them, W Q W K W is the projection matrix of the query and key. v Let Q be the projection matrix of the values. Split Q, K, and V into A independent heads, each focusing on features of a different subspace. a ,K a V a Let h be the query, key, and value matrix of the a-th head. fused Obtained through linear transformation. The formula for calculating the output of the a-th head is:

[0093]

[0094] The multi-head outputs are concatenated and linearly transformed to map them back to the original dimensions, yielding multimodal features. The calculation formula is as follows:

[0095] MultiHead(Q,K,V)=Concat(head1,...,head A W o

[0096] Where MultiHead(Q,K,V) represents the multimodal features obtained after processing by a multi-head attention mechanism, and W... o The output matrix is ​​learnable, and Concat represents the fusion operation.

[0097] By first fusing semantically aligned single-modal features and then using a multi-head attention mechanism for cross-modal weighted fusion, we can fully integrate information from different modalities, capture deeper-level correlations, improve the expressiveness of multimodal features, and help improve the accuracy of subsequent QoS policy adjustments.

[0098] Step 203: Take the multimodal feature corresponding to any influencing factor as the cause node and the QoS index as the result node; the cause node and the result node are graph nodes; determine the node connection edge between any two graph nodes based on the causal relationship between any two influencing factors and the causal relationship between any influencing factor and the QoS index; obtain a directed acyclic graph based on the graph nodes and the determined node connection edge; determine the target weight corresponding to any node connection edge in the directed acyclic graph to obtain the causal graph model, so as to adjust the QoS strategy according to the causal graph model.

[0099] The result node is the output node of the cause-effect graph model, used to output QoS indicator values.

[0100] In addition to the causal relationship between influencing factors and QoS indicators, there may also be causal relationships among influencing factors themselves. Therefore, when constructing a causal graph model, it is necessary to consider not only the causal relationship between influencing factors and QoS indicators, but also the causal relationship among influencing factors.

[0101] In this graph, nodes with causal relationships are connected by nodes-to-node edges, while nodes without causal relationships are not connected by nodes-to-node edges.

[0102] As an example, the target weights corresponding to the edges connecting nodes can be preset.

[0103] In one possible embodiment of this application, the node connection edges in the directed acyclic graph correspond to reference weights; reference feature data corresponding to any multimodal feature is obtained, and the reference feature data corresponds to a real QoS index value; the reference feature data is used as the input of the cause and result, and the predicted QoS index value output by the result node is determined according to the reference weights; based on the difference between the predicted QoS index value and the real QoS index value, the reference weights are adjusted through a gradient backpropagation algorithm to obtain the target weights.

[0104] The reference weight can refer to the initial weight of the edge connecting the node, and the reference weight can be set manually.

[0105] For example, the formula for obtaining the target weight by adjusting the reference weights using the gradient backpropagation algorithm is as follows:

[0106]

[0107] Among them, W IJ Y represents the weight of the node connection edge from the i-th node to the j-th node in the causal graph model. sis the output of the s-th result node (QoS indicator node) in the causal graph model, and L is the mean squared error loss calculated based on the difference between the predicted QoS indicator value and the actual QoS indicator value.

[0108] Using the above formula, gradient backpropagation is performed until the value of L is less than the set value or the number of weight updates reaches the set number, thus obtaining the target weight.

[0109] The actual QoS index value corresponding to the reference feature data can be obtained by data collection or calculated based on the probability distribution of the influencing factors and the conditional probability of QoS occurrence.

[0110] For example, assuming the QoS metric is network latency, and influencing factors include user density and bandwidth, the QoS metric values ​​under different conditions can be calculated using the following formula:

[0111]

[0112] P(QoS|U=u)=∫P(QoS|B,U=u)P(B)dB

[0113] Wherein, P(QoS|d) o (B=b)) is when a specific intervention is applied to bandwidth B. o Given bandwidth B = b (i.e., bandwidth B is value b), the conditional probability of Quality of Service (QoS) occurring is given by P(QoS|B = b, U). P(U) represents the conditional probability of QoS occurring when bandwidth B is value b and variables such as user density are U. P(U) is a probability distribution estimated from historical data, reflecting the probability of occurrence of variables such as user density under different conditions. o (B = b) indicates that the bandwidth B is reduced relative to the set bandwidth. For example, b = 0.8B', where B' is the set bandwidth, means that the bandwidth is reduced by 20%. Similarly, u = 0.8U' means that the user density is reduced by 20%.

[0114] Among them, P(QoS|d) can be calculated based on the obtained P(QoS|d) o (B=b)0 and P(QoS|U=u), calculate the value of the QoS index under the corresponding conditions, that is, the actual QoS index value.

[0115] To avoid feedback loops in bidirectional causal relationships, time slicing or instrumental variables can be used to determine the causal relationships between nodes, thus breaking the feedback loop. Time slicing discretizes a continuous time series into multiple time segments. In a causal structure with feedback loops, by observing the node state at different time slices, the dynamic feedback relationship is transformed into a staged static relationship. For example, observing the node state at time t and then again at time t+1, the causal direction is determined based on the time sequence, breaking the loop. Instrumental variables are variables that are related to the causal variable but only affect the outcome variable through the causal variable, unaffected by other confounding factors. In a given causal structure, suitable instrumental variables are found, and their association with the causal variable and their indirect influence on the outcome variable are utilized to isolate the causal effect, bypassing the interference of feedback loops and accurately determining the causal relationships between variables.

[0116] For example, regarding the causal relationship between "bandwidth B → network latency D", the "operator backbone link failure indicator" can be used as an instrumental variable. This variable is exogenous to the current causal structure and is related to B; for instance, changes in the failure indicator may affect B, and it only indirectly affects D through B, unaffected by other confounding factors. Using this instrumental variable, the causal effect of B on D can be isolated. Similarly, regarding the relationship between "user density U → B", the "school opening event" can be used as an instrumental variable. The school opening event may affect U, and it only affects B through U. By analyzing the relationship between this instrumental variable and U and B, feedback loop interference can be bypassed, and the causal impact of U on B can be accurately estimated, thus achieving reasonable causal inference in complex causal structures.

[0117] In one possible embodiment of this application, adjusting the QoS policy based on the causal graph model includes: obtaining the real-time QoS indicator value output by the result node in the causal graph model; querying the matching adjustment rule based on the real-time QoS indicator value; and using a near-end policy optimization algorithm to adjust the QoS policy based on the adjustment rule.

[0118] Different QoS indicator values ​​may be matched with different adjustment rules. For example, multiple QoS indicator value ranges can be set, and each QoS indicator value range is associated with a matching adjustment rule.

[0119] Specifically, the QoS indicator value range in which the real-time QoS indicator value is located can be determined. Then, based on the adjustment rules matched by this QoS indicator value range, the Proximal Policy Optimization (PPO) algorithm is used to adjust the QoS policy based on the adjustment rules.

[0120] For example, the formula for the objective function of PPO is as follows:

[0121]

[0122] Where, π θ For the new strategy parameters, π old For the old strategy parameters, R t denoted as the reward value at time step t, and ∈ is a hyperparameter, typically ranging from 0.1 to 0.3.

[0123] In this embodiment, the multimodal features corresponding to any influencing factor are used as cause nodes, and the QoS index is used as result nodes; the cause nodes and result nodes are graph nodes; based on the causal relationship between any two influencing factors, and the causal relationship between any influencing factor and the QoS index, the node connection edges between any two graph nodes are determined; based on the graph nodes and the determined node connection edges, a directed acyclic graph is obtained; the target weight corresponding to any node connection edge in the directed acyclic graph is determined, and a causal graph model is obtained; wherein, constructing a directed acyclic graph with multimodal features as cause nodes and QoS indexes as result nodes can clearly present the causal logic between various factors; determining the edge weights to form a causal graph model can quantify the strength of the causal relationship between graph nodes, accurately analyze the effect of each influencing factor on QoS, and thus provide an intuitive and reliable basis for QoS policy adjustment.

[0124] To implement the above embodiments, this application also proposes a service quality control device.

[0125] Figure 3 This is a schematic diagram of a service quality control device provided in an embodiment of this application.

[0126] like Figure 3 As shown, the service quality control device 300 includes:

[0127] The acquisition module 310 is used to acquire multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) index.

[0128] The extraction module 320 is used to extract features from the multimodal data corresponding to any influencing factor to obtain multimodal features;

[0129] Module 330 is used to construct a causal graph model based on the multimodal characteristics and QoS indicators corresponding to any influencing factor, so as to adjust the QoS strategy according to the causal graph model.

[0130] Optionally, the building module 330 is specifically used for:

[0131] The multimodal features corresponding to any influencing factor are used as cause nodes, and the QoS indicators are used as result nodes; where the cause nodes and result nodes are graph nodes.

[0132] Based on the causal relationship between any two influencing factors, and the causal relationship between any influencing factor and the QoS index, determine the node connection edges between any two graph nodes;

[0133] By connecting the nodes of the graph with edges that define the nodes, a directed acyclic graph is obtained.

[0134] By determining the target weight of the edge connecting any node in a directed acyclic graph, a causal graph model is obtained.

[0135] Optionally, the nodes connecting edges in the directed acyclic graph have corresponding reference weights, and the construction module 330 is specifically used for:

[0136] Obtain reference feature data corresponding to any multimodal feature. The reference feature data corresponds to the actual QoS indicator value.

[0137] The reference feature data is used as the input to the cause and result, and the predicted QoS index value of the result node is determined according to the reference weight.

[0138] Based on the difference between the predicted QoS indicator value and the actual QoS indicator value, the reference weights are adjusted using a gradient backpropagation algorithm to obtain the target weights.

[0139] Optionally, the QoS policy can be adjusted based on the cause-effect graph model, including:

[0140] Obtain the real-time QoS metric values ​​output by the result nodes in the cause-effect graph model;

[0141] Based on real-time QoS metric values, the matching adjustment rules are queried, and a near-end policy optimization algorithm is used to adjust the QoS policy based on the adjustment rules.

[0142] Optionally, the extraction module 320 is specifically used for the extraction of multiple modal dimensions corresponding to the multimodal data:

[0143] Feature extraction is performed on modal data under any modal dimension in multimodal data to obtain single-modal features under the corresponding modal dimension;

[0144] Semantic alignment of single-modal features across different modal dimensions;

[0145] Multimodal features are obtained by fusing the semantically aligned single-modal features.

[0146] Optionally, the extraction module 320 is specifically used for:

[0147] The semantically aligned single-modal features are fused to obtain fused features.

[0148] A multi-head attention mechanism is used to perform cross-modal weighted fusion processing on the fused features to obtain multimodal features.

[0149] Optionally, the device further includes a first processing module for:

[0150] A cubic spline interpolation algorithm is used to perform timestamp alignment on multimodal data.

[0151] Optionally, the device further includes a second processing module for:

[0152] When multimodal data includes numerical data, the K-nearest neighbor algorithm is used to fill in missing data in the numerical data.

[0153] In this embodiment, multimodal data corresponding to at least one influencing factor of the Quality of Service (QoS) index is acquired. For any influencing factor, feature extraction is performed on the multimodal data corresponding to the influencing factor to obtain multimodal features. Based on the multimodal features corresponding to any influencing factor and the QoS index, a causal graph model is constructed to adjust the QoS policy according to the causal graph model. By combining multimodal data and QoS index to construct a causal graph model, the reliance on inaccurate prior probabilities and the impact of prior knowledge bias on QoS adjustment performance are avoided. At the same time, the causal graph model can flexibly adapt to the time-varying characteristics of multimodal data streams without the need to build a large prior knowledge base, thus enabling more accurate dynamic adjustment of the QoS policy.

[0154] It should be noted that the foregoing explanation of the service quality control method embodiment also applies to the service quality control device of this embodiment, and will not be repeated here.

[0155] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 400 in this embodiment is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0156] like Figure 4 As shown, the above-mentioned electronic device 400 includes:

[0157] The memory 401 and the processor 402 are connected by a bus 403, which connects the different components (including the memory 401 and the processor 402). The memory 401 stores a computer program, and when the processor 402 executes the program, it implements the service quality control method of the present application embodiment.

[0158] Bus 403 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0159] Electronic device 400 typically includes a variety of electronic device readable media. These media can be any available media that can be accessed by electronic device 400, including volatile and non-volatile media, removable and non-removable media.

[0160] Memory 401 may also include computer system readable media in the form of volatile memory, such as random access memory (RAM) 404 and / or cache memory 405. Electronic device 400 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 406 may be used to read and write non-removable, non-volatile magnetic media (… Figure 4 Not shown; usually referred to as a "hard drive"). Although Figure 4 As not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 403 via one or more data media interfaces. Memory 401 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of this application.

[0161] A program / utility 408 having a set (at least one) of program modules 407 may be stored, for example, in memory 401. Such program modules 407 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 407 typically perform the functions and / or methods described in the embodiments of this application.

[0162] Electronic device 400 can also communicate with one or more external devices 409 (e.g., keyboard, pointing device, display 411, etc.), and with one or more devices that enable a user to interact with the electronic device 400, and / or with any device that enables the electronic device 400 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 412. Furthermore, electronic device 400 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 413. Figure 4 As shown, network adapter 413 communicates with other modules of electronic device 400 via bus 403. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with electronic device 400, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0163] The processor 402 performs various functional applications and data processing by running programs stored in the memory 401.

[0164] It should be noted that the implementation process and technical principles of the electronic device in this embodiment are explained in the foregoing description of the service quality control method of this application embodiment, and will not be repeated here.

[0165] To implement the above embodiments, this application also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.

[0166] To implement the above embodiments, this application also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.

[0167] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this application comply with relevant laws and regulations and do not violate public order and good morals.

[0168] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.

[0169] This application is intended to provide an implementation scheme for users to selectively prevent the use or access to their personal information data. Specifically, this disclosure is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information is de-identified to protect user privacy.

[0170] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0171] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0172] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0173] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0174] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0175] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0176] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0177] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A quality of service control method, characterized by, The method comprises the following steps: obtaining multi-modal data corresponding to at least one influencing factor of a quality of service (QoS) indicator; performing feature extraction on the multi-modal data corresponding to any influencing factor to obtain multi-modal features; constructing a causal graph model based on the multi-modal features corresponding to any influencing factor and the QoS indicator, and adjusting a QoS strategy according to the causal graph model.

2. The method of claim 1, wherein, The method of constructing a causal graph model based on the multi-modal features corresponding to any influencing factor and the QoS indicator comprises: taking the multi-modal features corresponding to any influencing factor as cause nodes and the QoS indicator as a result node, wherein the cause nodes and the result node are graph nodes; determining node connection edges between any two graph nodes according to the causal relationship between any two influencing factors and the causal relationship between any influencing factor and the QoS indicator; obtaining a directed acyclic graph according to the graph nodes and the determined node connection edges; determining a target weight corresponding to any node connection edge in the directed acyclic graph to obtain the causal graph model.

3. The method of claim 2, wherein, The node connection edges in the directed acyclic graph correspond to reference weights, and the method of determining a target weight corresponding to any node connection edge in the directed acyclic graph comprises: obtaining reference feature data corresponding to any multi-modal feature, wherein the reference feature data corresponds to a real QoS indicator value; taking the reference feature data as input of the cause result and determining a predicted QoS indicator value output by the result node according to the reference weight; adjusting the reference weight by a gradient backpropagation algorithm according to the difference between the predicted QoS indicator value and the real QoS indicator value to obtain the target weight.

4. The method of claim 2, wherein, The method of adjusting a QoS strategy according to the causal graph model comprises: obtaining a real-time QoS indicator value output by the result node in the causal graph model; querying a matched adjustment rule according to the real-time QoS indicator value and adjusting a QoS strategy based on the adjustment rule by using a proximal policy optimization algorithm.

5. The method of claim 1, wherein, The method of performing feature extraction on the multi-modal data corresponding to any influencing factor to obtain multi-modal features comprises: performing feature extraction on modal data under any modal dimension in the multi-modal data to obtain single-modal features under the corresponding modal dimension; performing semantic alignment on the single-modal features under different modal dimensions; performing feature fusion on the single-modal features after semantic alignment to obtain the multi-modal features.

6. The method of claim 5, wherein, The method of performing feature fusion on the single-modal features after semantic alignment to obtain the multi-modal features comprises: performing feature fusion on the single-modal features after semantic alignment to obtain fused features; performing cross-modal weighted fusion processing on the fused features by using a multi-head attention mechanism to obtain the multi-modal features.

7. The method of claim 1, wherein, The method further comprises: performing timestamp alignment processing on the multi-modal data by using a cubic spline interpolation algorithm.

8. The method of claim 1, wherein, The method further comprises: In a case where the multi-modal data comprises numerical data, K-Nearest Neighbor algorithm is adopted to fill in missing data of the numerical data.

9. A quality of service control apparatus characterized by comprising: The method comprises the steps of: an acquisition module, configured to acquire multi-modal data corresponding to at least one influencing factor of a quality of service (QoS) indicator; an extraction module, configured to, for any influencing factor, perform feature extraction on the multi-modal data corresponding to the influencing factor to obtain multi-modal features; a construction module, configured to construct a causal graph model based on the multi-modal features corresponding to any influencing factor and the QoS indicator, and to perform adjustment of a QoS strategy according to the causal graph model.

10. An electronic device, comprising: The method comprises the steps of: a processor, and a memory connected to the processor in communication; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1-8.

11. A computer readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by the processor to implement the method according to any one of claims 1-8.

12. A computer program product, characterised in that, The computer program is executed by the processor to implement the method according to any one of claims 1-8.

Images