Network state analysis method and device, equipment and medium

Abstract

The invention provides a network state analysis method and device, equipment and a medium, and relates to the technical field of deep learning. The method comprises the following steps: firstly, acquiring historical internet surfing data of each piece of hung terminal equipment of target network equipment; secondly, performing feature extraction on the historical internet surfing data through a feature extraction unit included in the target network state analysis model, and outputting historical internet surfing semantic features corresponding to the historical internet surfing data; and then, analyzing the historical internet surfing semantic features through a feature analysis unit, and outputting network state data corresponding to the target network equipment. Based on the content, the problem of relatively low reliability of network state analysis in the prior art can be improved.

Description

Technical Field

[0001] The present application relates to the field of deep learning technology, and specifically to a network status analysis method and apparatus, equipment, and medium. Background Art

[0002] With the rapid development of the internet and smart home devices, FTTR (Fiber to the Room) technology has been widely used in home and business networks. FTTR equipment connects directly to devices in each room via optical fiber, providing high-speed, stable network connections that meet the needs of high bandwidth and low latency. However, as the number of connected devices in the network increases, traditional network status monitoring and analysis methods are no longer able to effectively cope with complex and changing network environments. Existing network status analysis methods mostly rely on static rules or traditional signal strength and bandwidth measurements. While these methods can provide a certain level of network health assessment, their reliance on simple statistical data prevents them from effectively extracting and utilizing the deep semantic information contained in historical network data. This means that network status analysis can only rely on surface data and is unable to cope with complex network environments and changes in device behavior. Summary of the Invention

[0003] In view of this, the purpose of the present application is to provide a network status analysis method and apparatus, equipment and medium to improve the problem of relatively low reliability of network status analysis in the prior art.

[0004] To achieve the above objectives, this application adopts the following technical solutions: A network status analysis method, comprising: Obtaining historical Internet access data of each terminal device connected to the target network device, wherein the historical Internet access data is used to at least reflect the device Internet access time period of each terminal device connected to the target network device and the bandwidth usage during the corresponding device Internet access time period; Performing feature extraction on the historical online data by a feature extraction unit included in a target network state analysis model, and outputting historical online semantic features corresponding to the historical online data, wherein the target network state analysis model is a trained neural network model, and the target network state analysis model further includes a feature analysis unit, and in the process of performing feature extraction, extracting associated semantic information between the historical online data corresponding to each connected terminal device; The historical Internet access semantic features are analyzed by the feature analysis unit, and the network status data corresponding to the target network device is output, wherein the network status data is used to reflect the bandwidth requirements of each of the downstream terminal devices for the target network device, and the network status data serves as the basis for the target network device to dynamically allocate bandwidth.

[0005] In a preferred embodiment of the present application, in the above-mentioned network status analysis method, the step of extracting features from the historical online data by a feature extraction unit included in the target network status analysis model and outputting historical online semantic features corresponding to the historical online data includes: The semantic embedding subunit in the feature extraction unit included in the target network state analysis model processes the historical Internet access data of each connected terminal device to obtain a historical Internet access embedding feature corresponding to each historical Internet access data, wherein the historical Internet access embedding feature includes multiple sub-features, each sub-feature corresponding to a device Internet access time period; The historical Internet embedded features corresponding to each historical Internet data are aggregated through the association extraction subunit in the feature extraction unit, and the historical Internet semantic features corresponding to the historical Internet data are output. In the aggregation process, at least two levels of association semantic information extraction are performed, and the at least two levels include the association semantic information extraction between sub-features within the historical Internet embedded features and the association semantic information extraction between the historical Internet embedded features.

[0006] In a preferred embodiment of the present application, in the above-mentioned network status analysis method, the step of aggregating the historical online embedded features corresponding to each historical online data through the association extraction subunit in the feature extraction unit and outputting the historical online semantic features corresponding to the historical online data includes: For each historical Internet embedded feature corresponding to the historical Internet data, clustering is performed on the multiple sub-features included in the historical Internet embedded feature, and related semantic information is extracted based on the central sub-feature formed by the clustering process to obtain the corresponding internal related semantic feature; The associated semantic information of each of the internal associated semantic features is extracted by the associated extraction subunit in the feature extraction unit, and the corresponding historical online semantic features are output.

[0007] In a preferred embodiment of the present application, in the above-mentioned network status analysis method, the steps of clustering the historical online embedded features corresponding to each historical online data, performing clustering processing on the multiple sub-features included in the historical online embedded features, and extracting associated semantic information based on the central sub-feature formed by the clustering processing to obtain the corresponding internal associated semantic features include: For each historical Internet access embedding feature corresponding to the historical Internet access data, clustering is performed on multiple sub-features included in the historical Internet access embedding feature to obtain multiple central sub-features; Sorting the plurality of sub-features according to a corresponding time sequence, and segmenting the sequence formed by the sorting based on the plurality of central sub-features to form a plurality of segmented sub-sequences corresponding to the plurality of central sub-features, wherein the last sub-feature in each of the segmented sub-sequences is the corresponding central sub-feature; For each central sub-feature, based on the other sub-features in the segmentation sub-sequence corresponding to the central sub-feature, the associated semantic information of the central sub-feature is extracted to obtain the local associated semantic feature corresponding to the central sub-feature; The local associated semantic features corresponding to each of the central sub-features are concatenated to obtain the internal associated semantic features corresponding to the historical network embedding features.

[0008] In a preferred embodiment of the present application, in the above-mentioned network state analysis method, for each central sub-feature, based on other sub-features in the segmented sub-sequence corresponding to the central sub-feature, the step of extracting associated semantic information of the central sub-feature to obtain the local associated semantic feature corresponding to the central sub-feature includes: For each central sub-feature, the other sub-features in the segmentation sub-sequence corresponding to the central sub-feature are sequentially spliced ​​to form a spliced ​​sub-feature corresponding to the central sub-feature, and the spliced ​​sub-feature is compressed to form a compressed sub-feature with the same size as the central sub-feature; Based on the compressed sub-feature, cross-attention processing is performed on the central sub-feature to form a corresponding cross-attention sub-feature, and the cross-attention sub-feature and the central sub-feature are spliced, and the spliced ​​feature is compressed to form a local associated semantic feature with the same size as the central sub-feature.

[0009] In a preferred embodiment of the present application, in the above-mentioned network status analysis method, the step of extracting associated semantic information from each of the internal associated semantic features through the associated extraction subunit in the feature extraction unit and outputting the corresponding historical online semantic features includes: splicing the internal associated semantic features to form a spliced ​​associated semantic feature; performing a first compression process on the concatenated associated semantic feature based on a first compression parameter carried by the associated extraction subunit in the feature extraction unit to form a first compressed semantic feature having the same size as the internal associated semantic feature, wherein the first compression process is used to perform a first sliding window on the concatenated associated semantic feature and combine mean parameters obtained by performing mean processing on feature parameters included in each sliding window to form the first compressed semantic feature; Based on a second compression parameter carried by the association extraction subunit in the feature extraction unit, a second compression process is performed on the spliced ​​association semantic feature to form a second compressed semantic feature having the same size as the internal association semantic feature, wherein the second compression process is used to perform a second sliding window on the spliced ​​association semantic feature and combine standard deviation parameters obtained by performing standard deviation calculation on feature parameters included in each sliding window to form the second compressed semantic feature, and the window size of the second sliding window is equal to the window size of the first sliding window; splicing the first compressed semantic feature and the second compressed semantic feature to form a corresponding spliced ​​compressed semantic feature, and performing feature expansion on the spliced ​​compressed semantic feature to form a corresponding spliced ​​expanded semantic feature, wherein the size of the spliced ​​expanded semantic feature is the same as the size of the spliced ​​associated semantic feature; The splicing extended semantic feature and the splicing associated semantic feature are added together to output a corresponding historical online semantic feature.

[0010] In a preferred embodiment of the present application, in the above-mentioned network status analysis method, the network status analysis method further includes: Obtaining historical Internet access data of the sample and a corresponding network status data tag, wherein the historical Internet access data of the sample is used to at least reflect the device Internet access period of the corresponding connected terminal device and the bandwidth usage during the corresponding device Internet access period; Performing feature extraction on the sample historical online data by a feature extraction unit included in the candidate network state analysis model, and outputting sample historical online semantic features corresponding to the sample historical online data, wherein the candidate network state analysis model is a neural network model that needs to be trained, and the candidate network state analysis model also includes a feature analysis unit, and in the process of performing feature extraction, extracting associated semantic information between the historical online data corresponding to each connected terminal device; Analyzing the historical online semantic features of the sample through the feature analysis unit and outputting corresponding sample network status data; Based on the difference between the sample network state data and the network state data label, a network training loss index of the candidate network state analysis model is determined, and based on the network training loss index, the parameters of the candidate network state analysis model are optimized to form the target network state analysis model.

[0011] The present application also provides a network status analysis device, comprising: An online data acquisition module is used to acquire historical online data of each terminal device connected to the target network device, wherein the historical online data is used to at least reflect the device online time period of each terminal device connected to the target network device and the bandwidth usage during the corresponding device online time period; a semantic feature extraction module for extracting features from the historical online data using a feature extraction unit included in a target network state analysis model, and outputting historical online semantic features corresponding to the historical online data, wherein the target network state analysis model is a trained neural network model and further includes a feature analysis unit, and during the feature extraction process, extracting associated semantic information between the historical online data corresponding to each downstream terminal device; The semantic feature analysis module is used to analyze the historical Internet access semantic features through the feature analysis unit and output the network status data corresponding to the target network device, wherein the network status data is used to reflect the bandwidth requirements of each of the downstream terminal devices for the target network device, and the network status data serves as the basis for the target network device to dynamically allocate bandwidth.

[0012] Based on the above, the present application further provides an electronic device, including: memory for storing computer programs; The processor connected to the memory is used to execute the computer program stored in the memory to implement the above-mentioned network status analysis method.

[0013] On the basis of the above, the present application further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is run, each step of the above-mentioned network status analysis method is executed.

[0014] The network status analysis method, apparatus, device and medium provided by the present application first obtain the historical Internet access data of each downstream terminal device of the target network device; secondly, the feature extraction unit included in the target network status analysis model extracts the features of the historical Internet access data and outputs the historical Internet access semantic features corresponding to the historical Internet access data; then, the feature analysis unit analyzes the historical Internet access semantic features and outputs the network status data corresponding to the target network device. Based on the above content, on the one hand, the powerful learning ability of the neural network model can be used to extract the potential semantic information of the historical Internet access data, so that semantic features with richer semantics can be obtained. Therefore, compared with the conventional bandwidth prediction scheme based on simple statistical features, it can have a more accurate bandwidth prediction capability. On the other hand, since the associated semantic information between the historical Internet access data corresponding to each downstream terminal device is extracted during the feature extraction process, the extracted semantic features have higher reliability on the basis of richness, thereby improving the problem of relatively low reliability of network status analysis in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

[0016] Figure 1 This is a structural block diagram of an electronic device provided in an embodiment of the present application.

[0017] Figure 2 A flowchart of a network status analysis method provided in an embodiment of the present application.

[0018] Figure 3 A schematic diagram of the internal association semantic information extraction provided in an embodiment of the present application.

[0019] Figure 4 A schematic diagram of the extraction of external associated semantic information provided in an embodiment of the present application.

[0020] Figure 5 A block diagram of a network status analysis device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Generally, the components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0023] like Figure 1 As shown, an embodiment of the present application provides an electronic device, wherein the electronic device may include a memory, a processor, and a network status analysis device.

[0024] Specifically, the memory and the processor are electrically connected, directly or indirectly, to enable data transmission or interaction. For example, the memory and the processor may be electrically connected via one or more communication buses or signal lines. The network status analysis device includes at least one software function module stored in the memory in the form of software or firmware. The processor is configured to execute an executable computer program stored in the memory, such as the software function module and computer program included in the network status analysis device, to implement the network status analysis method provided in the embodiments of the present application.

[0025] Optionally, the memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0026] Furthermore, the processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), a system on chip (SoC), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0027] I understand. Figure 1 The structure shown is only for illustration, and the electronic device may also include Figure 1 More or fewer components than shown, or with Figure 1 The different configurations shown may, for example, further include a communication unit for exchanging information with other devices.

[0028] Combine Figure 2 , the embodiment of the present application also provides a network status analysis method applicable to the above electronic device. Among them, the method steps defined in the process related to the network status analysis method can be implemented by the electronic device. Figure 2 The specific process shown is explained in detail.

[0029] Step S110: Obtain historical Internet access data of each terminal device connected to the target network device.

[0030] In an embodiment of the present application, the electronic device can obtain historical internet access data for each downstream terminal device of the target network device. This historical internet access data at least reflects the device internet access time periods and bandwidth usage of each downstream terminal device during those time periods, such as 100 Mbps bandwidth at 10:00, 95 Mbps bandwidth at 10:05, 90 Mbps bandwidth at 10:10, and 80 Mbps bandwidth at 10:15. Furthermore, the target network device can be a gateway device, and the corresponding application scenario can be FTTR-B (Business FTTR), an all-optical networking solution for businesses and enterprises. This Wi-Fi solution, designed specifically for businesses, utilizes an all-optical networking solution combining optical fiber access, optical fiber cables, and Wi-Fi 6 to extend Wi-Fi coverage to every corner of the enterprise, providing network services for employees' online activities, such as live streaming and online conferencing. Furthermore, the target network device can also be used to provide IPTV (Internet Protocol Television) services.

[0031] Step S120 , performing feature extraction on the historical Internet access data through a feature extraction unit included in the target network status analysis model, and outputting historical Internet access semantic features corresponding to the historical Internet access data.

[0032] In an embodiment of the present application, after obtaining the historical Internet access data, the electronic device can perform feature extraction on the historical Internet access data through the feature extraction unit included in the target network state analysis model, and output the historical Internet access semantic features corresponding to the historical Internet access data. Wherein, the target network state analysis model is a trained neural network model, and the target network state analysis model also includes a feature analysis unit, and in the process of feature extraction, the associated semantic information between the historical Internet access data corresponding to each downstream terminal device is extracted. That is to say, in the process of feature extraction, it is necessary not only to extract the semantic information of each historical Internet access data itself, but also to extract the associated semantic information between each historical Internet access data.

[0033] Step S130: Analyze the historical Internet access semantic features through the feature analysis unit and output network status data corresponding to the target network device.

[0034] In an embodiment of the present application, after extracting the historical Internet access semantic features, the electronic device can analyze the historical Internet access semantic features through the feature analysis unit and output the network status data corresponding to the target network device. The network status data is used to reflect the bandwidth requirements of each of the downstream terminal devices for the target network device, and the network status data serves as the basis for the target network device to dynamically allocate bandwidth (for example, when the bandwidth requirement reflected by the gateway status data is 1800Mbps and the bandwidth capacity of the target network device is 1200Mbps, due to the problem of insufficient bandwidth, dynamic bandwidth allocation can be performed, such as giving priority to high-priority traffic (such as live broadcasts, online meetings, etc.) and reducing bandwidth allocation for low-priority traffic). In addition, the feature analysis unit can include a fully connected network layer and a linear regression function. After processing the historical Internet access semantic features, the fully connected network layer obtains the corresponding fully connected semantic features. Then, based on the linear regression function, the fully connected semantic features can be processed to obtain the corresponding bandwidth requirements. The fully connected semantic feature may include a parameter, and the linear regression function may map the parameter, such as identity mapping, to obtain a bandwidth value.

[0035] Based on the above content, on the one hand, the powerful learning ability of the neural network model can be used to extract the potential semantic information of the historical Internet access data, so that semantic features with richer semantics can be obtained. Therefore, compared with the conventional bandwidth prediction scheme based on simple statistical features, it can have a more accurate bandwidth prediction capability. On the other hand, since the associated semantic information between the historical Internet access data corresponding to each downstream terminal device will be extracted during the feature extraction process, the extracted semantic features will have higher reliability on the basis of richness, thereby improving the problem of relatively low reliability of network status analysis in the existing technology.

[0036] First, it should be noted that, with respect to step S120 , the specific method of extracting features from the historical Internet access data is not limited and can be selected according to actual needs.

[0037] For example, in an alternative embodiment, in order to ensure the reliability of the extracted historical Internet usage semantic features, the above-mentioned step S120 may further include step S121 and step S122, and the specific content of each step is described as follows.

[0038] In step S121, the historical Internet access data of each connected terminal device is processed respectively by the semantic embedding subunit in the feature extraction unit included in the target network status analysis model to obtain the historical Internet access embedding feature corresponding to each historical Internet access data.

[0039] In an embodiment of the present application, the historical Internet access data of each downstream terminal device can be processed separately through the semantic embedding subunit in the feature extraction unit included in the target network status analysis model to obtain the historical Internet access embedding feature corresponding to each historical Internet access data. Wherein, the historical Internet access embedding feature includes multiple sub-features, and each sub-feature corresponds to a device Internet access time period. Exemplarily, the semantic embedding subunit may include a word embedding model, so that each word in the historical Internet access data can be embedded through the word embedding model to obtain the corresponding word vector, and then, the word vectors belonging to the same device Internet access time period can be spliced ​​or added together to obtain the corresponding sub-feature. For example, for "the bandwidth is 1000Mbps", word segmentation can obtain "bandwidth", "for", "1000", "Mbps", and embedding can obtain: "bandwidth": [0.2, 0.1, -0.3, 0.7, 0.5, -0.6, 0.9, -0.1, 0.4,...,-0.2]; "for": [0.1, 0.0, 0.2, -0.1, 0.3, -0.5, 0.6, 0.0, -0.4,..., 0.5]; "1000": [0.8, -0.3, 0.2, 0.4, 0.7, 0.1, 0.9, -0.6, 0.2,..., 0.0]; "Mbps": [0.6, 0.8, -0.1, 0.9, -0.3, 0.4, 0.5, -0.2, 0.1,..., -0.7].

[0040] Step S122 : aggregating the historical Internet access embedded features corresponding to each historical Internet access data through the association extraction sub-unit in the feature extraction unit, and outputting the historical Internet access semantic features corresponding to the historical Internet access data.

[0041] In an embodiment of the present application, after obtaining the historical Internet embedded features corresponding to each historical Internet data, the historical Internet embedded features corresponding to each historical Internet data can be aggregated by the association extraction subunit in the feature extraction unit to output the historical Internet semantic features corresponding to the historical Internet data. In the aggregation process, at least two levels of association semantic information extraction are performed, including the association semantic information extraction between sub-features within the historical Internet embedded features and the association semantic information extraction between the historical Internet embedded features. That is, not only the association relationship between sub-features is focused on, but also the association relationship between different historical Internet embedded features can be focused on, so that more effective semantic information can be captured.

[0042] It is understandable that in the above-mentioned step S122, the specific method of aggregating the historical online embedded features corresponding to each historical online data is not limited. For example, in an alternative embodiment, in order to ensure the accuracy and efficiency of the aggregation, the above-mentioned step S122 may further include step S122a and step S122b. The specific content of each step is as follows.

[0043] Step S122a: for each historical Internet embedded feature corresponding to the historical Internet data, cluster the multiple sub-features included in the historical Internet embedded feature, and extract associated semantic information based on the central sub-feature formed by the clustering process to obtain the corresponding internal associated semantic feature.

[0044] In the embodiment of the present application, for each historical online embedding feature corresponding to the historical online data, clustering processing is performed on the multiple sub-features included in the historical online embedding feature, and associated semantic information is extracted based on the central sub-feature formed by the clustering processing to obtain the corresponding internal associated semantic feature. For example, the clustering processing can be performed based on a K-means clustering algorithm.

[0045] Step S122b: extracting associated semantic information from each of the internal associated semantic features through the associated extraction subunit in the feature extraction unit, and outputting corresponding historical online semantic features.

[0046] In an embodiment of the present application, after obtaining the internal associated semantic features, the associated extraction subunit within the feature extraction unit can be used to extract associated semantic information from each of the internal associated semantic features, outputting the corresponding historical online semantic features. Based on this, since the associated semantic information of the internal subfeatures is extracted before the external associated semantic information is extracted, the semantic representation of the object extracted from the external associated semantic information is more accurate, and the reliability of the external associated semantic information extraction is also improved, avoiding interference from a large number of irrelevant semantic features.

[0047] It is understood that in the above step S122a, the specific manner of clustering the multiple sub-features included in the historical network embedding feature and extracting the associated semantic information based on the central sub-feature formed by the clustering process is not limited. For example, in an improved embodiment, the above step S122a may further include step a1, step a2, step a3 and step a4. The specific contents of each step are as follows (combined with Figure 3 shown).

[0048] Step a1: for each historical Internet access embedded feature corresponding to the historical Internet access data, clustering is performed on multiple sub-features included in the historical Internet access embedded feature to obtain multiple central sub-features.

[0049] In the embodiment of the present application, for each historical online embedded feature corresponding to the historical online data, a clustering process can be performed on the multiple sub-features included in the historical online embedded feature to obtain multiple central sub-features, such as Figure 3 Subfeatures 3 and 7 in .

[0050] Step a2: sorting the plurality of sub-features according to their corresponding time sequence, and segmenting the sequence formed by the sorting based on the plurality of central sub-features to form a plurality of segmented sub-sequences corresponding to the plurality of central sub-features.

[0051] In an embodiment of the present application, after obtaining the multiple central sub-features, the multiple sub-features can be sorted according to the corresponding time sequence (i.e., the sequence between the corresponding device Internet access time periods) (such as time from early to late), and the sequence formed by the sorting can be segmented based on the multiple central sub-features to form multiple segmented sub-sequences corresponding to the multiple central sub-features. Among them, the last sub-feature in each of the segmented sub-sequences is the corresponding central sub-feature, such as the last sub-feature in the segmented sub-sequence corresponding to sub-feature 3 is sub-feature 3, and the last sub-feature in the segmented sub-sequence corresponding to sub-feature 7 is sub-feature 7. In addition, it should be noted that when the last sub-feature in the sequence formed by sorting does not belong to the central sub-feature formed by clustering, the last sub-feature can also be used as a central sub-feature, for example, in Figure 3 Among the sub-features shown, sub-feature 11 may also be used as the central sub-feature, so that three segmentation sub-sequences corresponding to sub-feature 3, sub-feature 7 and sub-feature 11 may be formed.

[0052] Step a3: for each central sub-feature, based on other sub-features in the segmentation sub-sequence corresponding to the central sub-feature, extract the associated semantic information of the central sub-feature to obtain the local associated semantic feature corresponding to the central sub-feature.

[0053] In an embodiment of the present application, after forming the segmented subsequences, for each central sub-feature, associated semantic information extraction can be performed on the central sub-feature based on the other sub-features in the segmented subsequence corresponding to the central sub-feature, thereby obtaining a local associated semantic feature corresponding to the central sub-feature. Based on this, since associated semantic information extraction is performed only on the central sub-feature, the number of associated semantic information extractions can be reduced, thereby improving efficiency and reducing data processing volume. In addition, since the sequence is formed according to a temporal sequence, the sub-features in the segmented subsequences formed by segmentation have a temporal correlation. Therefore, performing associated semantic information extraction within the segmented subsequences can also ensure its reliability.

[0054] Step a4: splicing the local associated semantic features corresponding to each of the central sub-features to obtain the internal associated semantic features corresponding to the historical online embedded features.

[0055] In an embodiment of the present application, after obtaining the local associated semantic features corresponding to each of the central sub-features, the local associated semantic features corresponding to each of the central sub-features can be concatenated to obtain the internal associated semantic features corresponding to the historical network embedded features. In this way, since the number of central sub-features is generally less than the number of sub-features, the number of parameters of the concatenated internal associated semantic features is smaller than the number of parameters of the semantic features obtained by directly concatenating the sub-features. This can further reduce the amount of data processing in subsequent processing, thereby improving processing efficiency.

[0056] It is understood that, in the above step a3, the specific manner of extracting associated semantic information of the central sub-feature based on other sub-features in the segmented sub-sequence corresponding to the central sub-feature is not limited. For example, in an alternative embodiment, in order to achieve attention to global semantic information while reducing the amount of computation, the above step a3 may further include the following: First, for each central sub-feature, the other sub-features in the segmentation sub-sequence corresponding to the central sub-feature are sequentially spliced ​​to form a spliced ​​sub-feature corresponding to the central sub-feature, and the spliced ​​sub-feature is compressed to form a compressed sub-feature with the same size as the central sub-feature. Exemplarily, compression can refer to downsampling, for example, which can be achieved through a convolutional network layer. In this way, on the basis of achieving size compression, more high-level and abstract features can be captured. Secondly, based on the compressed sub-features, the central sub-features can be cross-attention processed to form corresponding cross-attention sub-features (capturing the correlation features between other sub-features and the central sub-features, so that attention can be paid to important semantic features), and the cross-attention sub-features and the central sub-features can be spliced ​​(to avoid the complete loss of some non-correlated important semantic information and cause semantic distortion), and the spliced ​​features are compressed to form local correlation semantic features with the same size as the central sub-feature.

[0057] Based on this, through splicing, compression, cross-attention and other processing, the associated information of each sub-feature can be effectively utilized, and compression is used to ensure that the feature dimension will not be too large, maintain efficient calculation, and achieve the fusion of multi-level context information.

[0058] It is understood that in the above step S122b, the specific method of extracting the associated semantic information from each of the internal associated semantic features is not limited. For example, in order to effectively improve the expressiveness and detail capture capabilities of the features, the above step S122b may further include the following contents (combined with Figure 4 shown): First, the internal correlation semantic features can be spliced ​​together to form a spliced ​​correlation semantic feature, which is used to represent the global semantic information of each historical online data; Secondly, based on the first compression parameters carried by the association extraction subunit in the feature extraction unit (for example, the window size and step size of the sliding window, which can be specifically formed during the model training process), a first compression process can be performed on the spliced ​​association semantic feature to form a first compressed semantic feature having the same size as the internal association semantic feature. The first compression process is used to perform a first sliding window on the spliced ​​association semantic feature and combine the mean parameters obtained by performing mean processing on the feature parameters included in each sliding window to form the first compressed semantic feature. Based on this, feature fluctuations can be smoothed, noise can be removed, and global trend information can be retained. This can effectively reduce feature redundancy and retain the most important statistical information during the compression process. Then, based on the second compression parameters carried by the association extraction subunit in the feature extraction unit (for example, the window size and step size of the sliding window, which can be specifically formed during the model training process), a second compression process can be performed on the spliced ​​association semantic feature to form a second compressed semantic feature having the same size as the internal association semantic feature. The second compression process is used to perform a second sliding window on the spliced ​​association semantic feature and combine standard deviation parameters obtained by calculating the standard deviation of the feature parameters included in each sliding window to form the second compressed semantic feature, and the window size of the second sliding window is equal to the window size of the first sliding window. Based on this, the variation amplitude of the local feature can be captured, indicating the degree of fluctuation of the feature within the local window, and can effectively describe the discreteness or variation trend of the data, which helps to capture the distribution characteristics of the data, thereby providing more differentiated information for the model. Furthermore, the first compressed semantic feature and the second compressed semantic feature can be concatenated to form a corresponding concatenated compressed semantic feature, and the concatenated compressed semantic feature can be feature expanded (e.g., upsampling, specifically, can be achieved through transposed convolution) to form a corresponding concatenated expanded semantic feature, wherein the size of the concatenated expanded semantic feature is the same as the size of the concatenated associated semantic feature. Based on this, the information generated by the two different compression methods of mean and standard deviation is combined, thereby integrating the feature information of both stability (mean) and volatility (standard deviation) of the data, and providing a more comprehensive and comprehensive feature description. Finally, the spliced ​​extended semantic features and the spliced ​​associated semantic features can be added to output the corresponding historical online semantic features, or, in other implementations, a mean calculation can also be performed; based on this, semantic information at different levels can be comprehensively considered to avoid the problem of semantic loss during the compression process, thereby improving the comprehensiveness and accuracy of the semantic features.

[0059] Secondly, it should be noted that for the above-mentioned target network state analysis model, in order to ensure that the target network state analysis model can achieve reliable feature extraction and feature analysis, the network state analysis method may further include a step of training to form the target network state analysis model. In an alternative embodiment, the following may be specifically included: First, historical Internet access data of the sample and the corresponding network status data tag (i.e., actual bandwidth demand) can be obtained, wherein the historical Internet access data of the sample is used to at least reflect the device Internet access time period of the corresponding connected terminal device and the bandwidth usage during the corresponding device Internet access time period; Secondly, the feature extraction unit included in the candidate network state analysis model can be used to extract features from the sample historical online data and output sample historical online semantic features corresponding to the sample historical online data (refer to the explanation of step S120 above). The candidate network state analysis model is a neural network model that needs to be trained, and the candidate network state analysis model also includes a feature analysis unit. During the feature extraction process, semantic information associated with the historical online data corresponding to each connected terminal device is extracted. Then, the feature analysis unit analyzes the historical online semantic features of the sample and outputs corresponding sample network status data (refer to the explanation of step S130 above); Finally, based on the difference between the sample network state data and the network state data label, the network training loss index (such as mean square error, etc.) of the candidate network state analysis model can be determined, and based on the network training loss index, the parameters of the candidate network state analysis model can be optimized to form the target network state analysis model. For example, the parameters of the candidate network state analysis model can be updated and adjusted in the direction of reducing the network training loss index (the specific process can refer to the relevant existing technology and is not specifically limited here) until the network training loss index converges, thereby obtaining the target network state analysis model.

[0060] Combine Figure 5The present application also provides a network status analysis device applicable to the above electronic device. The network status analysis device may include an Internet access data acquisition module, a semantic feature extraction module, and a semantic feature analysis module.

[0061] The Internet access data acquisition module can be used to obtain the historical Internet access data of each terminal device connected to the target network device, wherein the historical Internet access data is at least used to reflect the device Internet access period of each terminal device connected to the target network device and the bandwidth usage during the corresponding device Internet access period. Figure 2 As shown in step S110, for the relevant content of the Internet data acquisition module, reference can be made to the above description of step S110.

[0062] The semantic feature extraction module can be used to extract features from the historical Internet access data through the feature extraction unit included in the target network status analysis model, and output the historical Internet access semantic features corresponding to the historical Internet access data, wherein the target network status analysis model is a trained neural network model, and the target network status analysis model also includes a feature analysis unit, and in the process of feature extraction, the associated semantic information between the historical Internet access data corresponding to each downstream terminal device is extracted. In the embodiment of the present application, the semantic feature extraction module can be used to perform Figure 2 As shown in step S120, for the relevant content of the semantic feature extraction module, reference can be made to the above description of step S120.

[0063] The semantic feature analysis module can be used to analyze the historical Internet semantic features through the feature analysis unit and output the network status data corresponding to the target network device, wherein the network status data is used to reflect the bandwidth requirements of each of the downstream terminal devices for the target network device, and the network status data is used as the basis for the target network device to dynamically allocate bandwidth. In the embodiment of the present application, the semantic feature analysis module can be used to perform Figure 2 As shown in step S130, for the relevant content of the semantic feature analysis module, reference can be made to the above description of step S130.

[0064] In an embodiment of the present application, corresponding to the above-mentioned network status analysis method applied to the electronic device, a computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is run, each step of the network status analysis method is executed.

[0065] The steps executed when the aforementioned computer program is running will not be described in detail here, and reference may be made to the above explanation of the network status analysis method.

[0066] In summary, the network status analysis method, apparatus, device and medium provided by the present application first obtain the historical Internet access data of each downstream terminal device of the target network device; secondly, the feature extraction unit included in the target network status analysis model extracts features of the historical Internet access data and outputs the historical Internet access semantic features corresponding to the historical Internet access data; then, the feature analysis unit analyzes the historical Internet access semantic features and outputs the network status data corresponding to the target network device. Based on the above content, on the one hand, the powerful learning ability of the neural network model can be used to extract the potential semantic information of the historical Internet access data, so that semantic features with richer semantics can be obtained. Therefore, compared with the conventional bandwidth prediction scheme based on simple statistical features, it can have a more accurate bandwidth prediction capability. On the other hand, since the associated semantic information between the historical Internet access data corresponding to each downstream terminal device is extracted during the feature extraction process, the extracted semantic features have higher reliability on the basis of richness, thereby improving the problem of relatively low reliability of network status analysis in the prior art.

[0067] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0068] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0069] If the functions are implemented in the form of software modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. It should be noted that, in this document, the terms "comprise," "include," or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. Without further constraints, an element defined by the phrase "comprises a..." does not preclude the existence of additional identical elements in the process, method, article or apparatus that includes the element.

[0070] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A network status analysis method, characterized in that: include: Obtaining historical Internet access data of each terminal device connected to the target network device, wherein the historical Internet access data is used to at least reflect the device Internet access time period of each terminal device connected to the target network device and the bandwidth usage during the corresponding device Internet access time period; Performing feature extraction on the historical online data by a feature extraction unit included in a target network state analysis model, and outputting historical online semantic features corresponding to the historical online data, wherein the target network state analysis model is a trained neural network model, and the target network state analysis model further includes a feature analysis unit, and in the process of performing feature extraction, extracting associated semantic information between the historical online data corresponding to each connected terminal device; The historical Internet access semantic features are analyzed by the feature analysis unit, and the network status data corresponding to the target network device is output, wherein the network status data is used to reflect the bandwidth requirements of each of the downstream terminal devices for the target network device, and the network status data serves as the basis for the target network device to dynamically allocate bandwidth.

2. The network status analysis method according to claim 1, characterized in that: The step of extracting features from the historical online data by a feature extraction unit included in the target network status analysis model and outputting historical online semantic features corresponding to the historical online data comprises: The semantic embedding subunit in the feature extraction unit included in the target network state analysis model processes the historical Internet access data of each connected terminal device to obtain a historical Internet access embedding feature corresponding to each historical Internet access data, wherein the historical Internet access embedding feature includes multiple sub-features, each sub-feature corresponding to a device Internet access time period; The historical Internet embedded features corresponding to each historical Internet data are aggregated through the association extraction subunit in the feature extraction unit, and the historical Internet semantic features corresponding to the historical Internet data are output. In the aggregation process, at least two levels of association semantic information extraction are performed, and the at least two levels include the association semantic information extraction between sub-features within the historical Internet embedded features and the association semantic information extraction between the historical Internet embedded features.

3. The network status analysis method according to claim 2, characterized in that: The step of aggregating the historical Internet access embedded features corresponding to each historical Internet access data through the association extraction subunit in the feature extraction unit and outputting the historical Internet access semantic features corresponding to the historical Internet access data includes: For each historical Internet embedded feature corresponding to the historical Internet data, clustering is performed on the multiple sub-features included in the historical Internet embedded feature, and related semantic information is extracted based on the central sub-feature formed by the clustering process to obtain the corresponding internal related semantic feature; The associated semantic information of each of the internal associated semantic features is extracted by the associated extraction subunit in the feature extraction unit, and the corresponding historical online semantic features are output.

4. The network status analysis method according to claim 3, characterized in that: The step of clustering the multiple sub-features included in the historical Internet embedding feature corresponding to each historical Internet data, and extracting associated semantic information based on the central sub-feature formed by the clustering process to obtain the corresponding internal associated semantic feature includes: For each historical Internet access embedding feature corresponding to the historical Internet access data, clustering is performed on multiple sub-features included in the historical Internet access embedding feature to obtain multiple central sub-features; Sorting the plurality of sub-features according to a corresponding time sequence, and segmenting the sequence formed by the sorting based on the plurality of central sub-features to form a plurality of segmented sub-sequences corresponding to the plurality of central sub-features, wherein the last sub-feature in each of the segmented sub-sequences is the corresponding central sub-feature; For each central sub-feature, based on the other sub-features in the segmentation sub-sequence corresponding to the central sub-feature, the associated semantic information of the central sub-feature is extracted to obtain the local associated semantic feature corresponding to the central sub-feature; The local associated semantic features corresponding to each of the central sub-features are concatenated to obtain the internal associated semantic features corresponding to the historical network embedding features.

5. The network status analysis method according to claim 4, characterized in that: The step of extracting associated semantic information of each central sub-feature based on other sub-features in the segmented sub-sequence corresponding to the central sub-feature to obtain a local associated semantic feature corresponding to the central sub-feature includes: For each central sub-feature, the other sub-features in the segmentation sub-sequence corresponding to the central sub-feature are sequentially spliced ​​to form a spliced ​​sub-feature corresponding to the central sub-feature, and the spliced ​​sub-feature is compressed to form a compressed sub-feature with the same size as the central sub-feature; Based on the compressed sub-feature, cross-attention processing is performed on the central sub-feature to form a corresponding cross-attention sub-feature, and the cross-attention sub-feature and the central sub-feature are spliced, and the spliced ​​feature is compressed to form a local associated semantic feature with the same size as the central sub-feature.

6. The network status analysis method according to claim 3, characterized in that: The step of extracting associated semantic information from each of the internal associated semantic features through the associated extraction subunit in the feature extraction unit and outputting the corresponding historical online semantic features includes: splicing the internal associated semantic features to form a spliced ​​associated semantic feature; performing a first compression process on the concatenated associated semantic feature based on a first compression parameter carried by the associated extraction subunit in the feature extraction unit to form a first compressed semantic feature having the same size as the internal associated semantic feature, wherein the first compression process is used to perform a first sliding window on the concatenated associated semantic feature and combine mean parameters obtained by performing mean processing on feature parameters included in each sliding window to form the first compressed semantic feature; Based on a second compression parameter carried by the association extraction subunit in the feature extraction unit, a second compression process is performed on the spliced ​​association semantic feature to form a second compressed semantic feature having the same size as the internal association semantic feature, wherein the second compression process is used to perform a second sliding window on the spliced ​​association semantic feature and combine standard deviation parameters obtained by performing standard deviation calculation on feature parameters included in each sliding window to form the second compressed semantic feature, and the window size of the second sliding window is equal to the window size of the first sliding window; splicing the first compressed semantic feature and the second compressed semantic feature to form a corresponding spliced ​​compressed semantic feature, and performing feature expansion on the spliced ​​compressed semantic feature to form a corresponding spliced ​​expanded semantic feature, wherein the size of the spliced ​​expanded semantic feature is the same as the size of the spliced ​​associated semantic feature; The splicing extended semantic feature and the splicing associated semantic feature are added together to output a corresponding historical online semantic feature.

7. The network status analysis method according to any one of claims 1 to 6, characterized in that: The network status analysis method further includes: Obtaining historical Internet access data of the sample and a corresponding network status data tag, wherein the historical Internet access data of the sample is used to at least reflect the device Internet access period of the corresponding connected terminal device and the bandwidth usage during the corresponding device Internet access period; Performing feature extraction on the sample historical online data by a feature extraction unit included in the candidate network state analysis model, and outputting sample historical online semantic features corresponding to the sample historical online data, wherein the candidate network state analysis model is a neural network model that needs to be trained, and the candidate network state analysis model also includes a feature analysis unit, and in the process of performing feature extraction, extracting associated semantic information between the historical online data corresponding to each connected terminal device; Analyzing the historical online semantic features of the sample through the feature analysis unit and outputting corresponding sample network status data; Based on the difference between the sample network state data and the network state data label, a network training loss index of the candidate network state analysis model is determined, and based on the network training loss index, the parameters of the candidate network state analysis model are optimized to form the target network state analysis model.

8. A network status analysis device, characterized in that: include: An online data acquisition module is used to acquire historical online data of each terminal device connected to the target network device, wherein the historical online data is used to at least reflect the device online time period of each terminal device connected to the target network device and the bandwidth usage during the corresponding device online time period; a semantic feature extraction module for extracting features from the historical online data using a feature extraction unit included in a target network state analysis model, and outputting historical online semantic features corresponding to the historical online data, wherein the target network state analysis model is a trained neural network model and further includes a feature analysis unit, and during the feature extraction process, extracting associated semantic information between the historical online data corresponding to each downstream terminal device; The semantic feature analysis module is used to analyze the historical Internet access semantic features through the feature analysis unit and output the network status data corresponding to the target network device, wherein the network status data is used to reflect the bandwidth requirements of each of the downstream terminal devices for the target network device, and the network status data serves as the basis for the target network device to dynamically allocate bandwidth.

9. An electronic device, characterized in that: include: memory for storing computer programs; A processor connected to the memory is used to execute the computer program stored in the memory to implement the network status analysis method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which executes the network status analysis method according to any one of claims 1 to 7 when the computer program is run.

Images