A method and device for optimizing performance of a multi-service network based on network performance indicators

Abstract

This application discloses a method and device for multi-service network performance optimization based on network performance indicators, relating to the field of computer technology. The method includes: acquiring network performance indicators for the current operating scenario; obtaining network state values ​​for each indicator through predictive processing; for each service type, weighted summing of the corresponding state values ​​based on matching weights, and combining this with a multiplicative penalty mechanism triggered by a packet loss rate threshold to obtain the final comprehensive network state value for the service; generating multi-service user experience prediction results; and determining anomalies based on the prediction results. If a network anomaly is determined in the current scenario, the target service type and anomaly indicator are identified, and a network state adjustment strategy is generated accordingly to optimize and adjust the current network state. In this way, accurate prediction of the network experience for multi-service users can be achieved, providing an effective decision-making basis for intelligent diagnosis and adaptive optimization of network quality.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method and device for optimizing the performance of a multi-service network based on network performance indicators. Background Technology

[0002] With the development of mobile internet, network speed testing technology has been widely used in user network quality assessment. In these technologies, speed testing systems primarily output Quality of Service (QoS) metrics such as download speed, upload speed, latency, jitter, and packet loss rate. However, QoS metrics are essentially low-level network performance data, and ordinary users find it difficult to directly understand their impact on actual user experience. For example, users cannot judge the impact of a "50 ms latency" on video calls, high bandwidth may mask high packet loss issues, and different application scenarios have different sensitivities to network metrics. Furthermore, these technologies suffer from issues such as a lack of user experience mapping, unreasonable single prediction models (failing to differentiate user scenarios, subjective weight settings, and opaque result interpretation), ignoring business differences (failing to design models for different business types' varying network sensitivity), and high bandwidth easily masking packet loss. These factors lead to inaccurate user experience predictions, making it difficult to provide reliable data support for network quality optimization. Consequently, inaccurate predictions of user network experience result in inaccurate diagnostics of the user's current network quality, making it difficult to promptly detect network anomalies and thus hindering the adoption of appropriate network adjustment strategies to address abnormal network conditions, ultimately leading to a poor user experience. Summary of the Invention

[0003] This application provides a method, device, and medium for predicting multi-service user experience based on network performance indicators, in order to solve the following technical problem: how to accurately predict the network experience of multi-service users and provide an effective decision-making basis for intelligent diagnosis and adaptive optimization of network quality.

[0004] In a first aspect, embodiments of this application provide a method for optimizing the performance of a multi-service network based on network performance metrics, the method comprising: Obtain network performance metrics under the current operating scenario, and perform predictive processing on the network performance metrics to obtain the network state value corresponding to the network performance metrics; For each service type, a first weight matching the service type is determined, and based on the first weight, the network state values ​​corresponding to the network performance indicators matching the service type are weighted and summed to obtain the initial comprehensive network state value of the service type. For each of the aforementioned service types, in response to the network state value corresponding to the packet loss rate of the service type being less than a preset network state threshold, a multiplicative penalty is applied to the initial comprehensive network state value of the service type to obtain the final comprehensive network state value of the service type. Based on the final comprehensive network state value for each of the aforementioned service types, a multi-service user experience prediction result for the current operating scenario is generated. Based on the multi-service user experience prediction results, the current operating scenario is characterized by network status anomalies, and the target service type of the network status anomaly is determined. Based on the target service type and abnormal network performance indicators, a network state adjustment strategy is generated to adjust the network state of the current operating scenario.

[0005] Secondly, embodiments of this application also provide a multi-service network performance optimization device based on network performance indicators. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a multi-service network performance optimization method based on network performance indicators as described above.

[0006] Thirdly, embodiments of this application also provide a computer storage medium storing computer-executable instructions, which, when executed, implement a multi-service network performance optimization method based on network performance indicators as described above.

[0007] This application provides a method, device, and medium for optimizing multi-service network performance based on network performance indicators, which has the following beneficial effects: First, network performance metrics are acquired and predictive processed. This transforms heterogeneous raw physical quantities such as latency, jitter, and bandwidth into unified, quantifiable network state values, achieving an initial mapping from physical layer data to the service perception layer and laying a standardized data foundation for subsequent predictions. Next, matching weights are determined for each service type and a weighted sum is performed. This reflects the objective fact that different services have varying network sensitivities, ensuring that the initial comprehensive network state value accurately reflects the network adaptability under specific service scenarios and avoids the one-sidedness of a "one-size-fits-all" assessment. Finally, a multiplicative penalty mechanism based on packet loss rate thresholds is introduced to generate service-specific data. The final comprehensive network status value can effectively amplify the devastating impact of packet loss, a key bottleneck indicator, on the user experience, preventing high bandwidth from masking the experience collapse caused by low packet loss. This significantly improves the consistency between the assessment results and the actual user experience. Subsequently, multi-service user experience prediction results are generated based on the final network status value of the service. This not only provides a macro-level network profile but also allows for the identification of target service types and abnormal indicators through subsequent anomaly detection logic. This supports the system in automatically generating targeted network status adjustment strategies, achieving a closed loop from "precise diagnosis" to "intelligent optimization," thereby significantly improving the efficiency of network operation and maintenance and user satisfaction. Attached Figure Description

[0008] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart of a multi-service network performance optimization method based on network performance indicators is provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a multi-service network performance optimization device based on network performance indicators, provided as an embodiment of this application. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0010] It is understood that in the embodiments of this application, data related to user information (such as user accounts) is involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0011] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0012] In the following description, the terms “first, second, ...” are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that “first, second, ...” may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0013] With the development of mobile internet, network speed testing technology has been widely used in user network quality assessment. In these technologies, speed testing systems primarily output Quality of Service (QoS) metrics such as download speed, upload speed, latency, jitter, and packet loss rate. However, QoS metrics are essentially low-level network performance data, and ordinary users find it difficult to directly understand their impact on actual user experience. For example, users cannot judge the impact of a 50ms latency on video calls, high bandwidth may mask high packet loss issues, and different application scenarios have different sensitivities to network metrics. Furthermore, these technologies suffer from issues such as a lack of user experience mapping, unreasonable single prediction models (failing to differentiate user scenarios, subjective weight settings, and opaque result interpretation), ignoring business differences (failing to design models for different business types' varying network sensitivity), and high bandwidth easily masking packet loss. These factors lead to inaccurate user experience predictions, making it difficult to provide reliable data support for network quality optimization. Consequently, inaccurate predictions of user network experience result in inaccurate quality diagnostics of the user's current network, making it difficult to promptly detect network anomalies and thus hindering the adoption of appropriate network status adjustment strategies, ultimately leading to a poor user experience.

[0014] Based on this, this application provides a method for optimizing the performance of a multi-service network based on network performance indicators, which can accurately predict the network experience of multi-service users and provide an effective decision-making basis for intelligent diagnosis and adaptive optimization of network quality.

[0015] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0016] Figure 1This application provides a flowchart of a multi-service network performance optimization method based on network performance metrics. This method can be applied to various network performance evaluation scenarios, such as operator network quality assessment and marketing, service optimization for content delivery networks and cloud service providers, enterprise network operation and maintenance and troubleshooting, embedded speed testing of terminal devices and applications, and network adaptation for smart home and Internet of Things (IoT) devices. Certain input parameters or intermediate results in the process can be manually adjusted to help improve accuracy.

[0017] This application provides a method for optimizing multi-service network performance based on network performance indicators. It should be noted that the execution entity in these embodiments can be a server or any terminal device with data processing capabilities. For example, the server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal device can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, in-vehicle terminal, etc., but is not limited to these.

[0018] like Figure 1 As shown in the figure, the multi-service network performance optimization method based on network performance indicators provided in this application embodiment specifically includes the following steps: Step 101: Obtain network performance metrics for the current operating scenario.

[0019] It should be noted that the current operating scenario refers to the technical environment or carrier on which the software runs. In this application, the current operating scenario may include client / APP and web page; network performance indicators refer to the underlying data (i.e., QoS indicators) used to quantitatively describe network transmission quality and service capabilities.

[0020] In some embodiments, network performance metrics include basic network performance metrics and derived network performance metrics. Step 101 described above can be implemented as follows: Obtain the basic network performance metrics for the current operating scenario using speed testing software; use the difference between load latency and idle latency in the basic network performance metrics as the queuing latency in the derived network performance metrics, where the load latency is the maximum value of upload load latency and download load latency in the basic network performance metrics; and use the maximum value among idle jitter, download jitter, and upload jitter in the basic network performance metrics as the effective jitter in the derived network performance metrics.

[0021] Thus, by accurately quantifying the queuing delay in network data transmission using the difference between idle and unloaded latency, and comprehensively capturing the worst-case scenario in bidirectional transmission by taking the maximum value of upload and download load latency, network congestion risks that cannot be detected by looking only at idle latency can be effectively identified. At the same time, defining effective jitter by the maximum value among idle jitter, download jitter, and upload jitter can uniformly reflect the stability risks of the network under different operating states, avoiding the masking of actual transmission fluctuations due to good jitter data in a single dimension. Therefore, by transforming the original, isolated basic network performance indicators into feature indicators that better reflect the real service experience, a solid data foundation is laid for subsequent prediction of network state values ​​in multiple service scenarios, thereby significantly improving the accuracy and objectivity of the prediction results.

[0022] It should be noted that basic network performance metrics refer to raw network data directly detected and collected by speed test software without complex processing, such as download speed, upload speed, idle latency, download load latency, upload load latency, idle jitter, upload jitter, download jitter, packet loss rate, etc.; derived network performance metrics refer to new metrics generated based on basic network performance metrics through specific mathematical operations or logical combinations, such as queuing latency, effective jitter, etc.; speed test software is a dedicated tool or application used to measure network connection performance (e.g., Ookla). Speedtest, etc.); Upload load latency refers to the network round-trip latency measured during data upload; Download load latency refers to the network round-trip latency measured during data download; Idle latency refers to the round-trip latency measured when sending probe packets to the server in the absence of background traffic (i.e., idle state); Idle jitter refers to the variation (variance or standard deviation) of the round-trip latency of data packets continuously measured in an idle network state (no large data transmission); Upload jitter refers to the amount of change in the round-trip latency of data packets measured during data upload; Download jitter refers to the amount of change in the round-trip latency of data packets measured during data download; Packet loss rate refers to the proportion of data packets lost during network transmission, usually expressed as a percentage (%); Queuing latency is used to quantify the additional delay caused by network congestion and router buffer queuing; Effective jitter is used to uniformly describe the most representative jitter risk in the current test.

[0023] As an example, assuming a home WiFi environment, after the speed test software starts, basic network performance indicators are collected. The idle latency is 20ms, the download load latency is 120ms, the upload load latency is 80ms, the idle jitter is 5ms, the download jitter is 30ms, and the upload jitter is 15ms. Then, according to the rules, the download load latency (120ms) and the upload load latency (80ms) are compared, and the maximum value is selected as the load latency, which is 120ms. By subtracting the idle latency (20ms) from the load latency (120ms), the queuing latency of the current scenario can be calculated as 100ms. After that, according to the rules, the maximum value among the idle jitter (5ms), download jitter (30ms), and upload jitter (15ms) is compared, and the effective jitter is obtained as 30ms.

[0024] Step 102: Perform prediction processing on the network performance indicators to obtain the network state values ​​corresponding to the network performance indicators.

[0025] It should be noted that the above prediction processing can be a piecewise linear mapping (e.g., a piecewise function) or a non-linear function mapping (e.g., an exponential function, a logarithmic function, a sigmoid function, etc.), or a lookup table or a threshold-based discrete mapping, or a model mapping based on machine learning or deep learning training, without any specific limitations here.

[0026] As an example, taking a piecewise function, assuming that based on business experience, the network state value corresponding to a latency of 0-5ms is 100, the network state value corresponding to a latency of 5-10ms is 97-100, the network state value corresponding to a latency of 10-20ms is 92-97, the network state value corresponding to a latency of 20-40ms is 82-92, the network state value corresponding to a latency of 40-60ms is 72-82, the network state value corresponding to a latency of 60-100ms is 55-72, the network state value corresponding to a latency of 100-150ms is 35-55, the network state value corresponding to a latency of 150-250ms is 15-35, the network state value corresponding to a latency of 250-400ms is 0-15, and the network state value corresponding to a latency greater than 400ms is 0, when a specific latency value (e.g., 50ms) is obtained, according to the mapping relationship of the piecewise function mentioned above, the corresponding network state value can be obtained as 77.

[0027] In some embodiments, step 102 described above can be implemented in the following way: for each network performance indicator, perform the following processing respectively: determine the target state value interval corresponding to the network performance indicator from a plurality of preset state value intervals; based on the target state value interval, perform mapping processing on the network performance indicator to obtain the network state value corresponding to the network performance indicator.

[0028] In this way, by dividing the indicators into multiple preset intervals, differentiated mapping standards can be set for different levels of network quality, thereby avoiding the limitations of a single linear mapping that is difficult to adapt to complex network environments. Furthermore, performing linear mapping within each interval can ensure the continuity and smoothness of network state value changes, preventing jumps in state values, and enabling network state value prediction results to more accurately reflect the subtle differences of the indicators within the corresponding intervals. This approach not only aligns with the non-linear impact of network quality on actual user experience but also provides a more reasonable and valuable data foundation for subsequent multi-indicator fusion and comprehensive network state value calculation, thereby improving the accuracy and reliability of overall user experience evaluation.

[0029] As an example, taking a piecewise function, assuming that based on business experience, the network state value corresponding to jitter between 0 and 1 ms is 100, between 1 and 5 ms is 95-100, between 5 and 10 ms is 85-95, between 10 and 20 ms is 65-85, between 20 and 30 ms is 50-65, between 30 and 50 ms is 30-50, between 50 and 80 ms is 15-30, between 80 and 150 ms is 0-15, and between jitter greater than 150 ms is 0, when a specific jitter (7 ms) is obtained, the corresponding target state value range can be determined to be 85-95 through matching. Then, linear interpolation is performed within this target state value range to obtain the network state value corresponding to the current jitter as 92.

[0030] In some embodiments, step 102 described above can also be implemented using a machine learning model, such as regression or a neural network, by constructing training data to train the model so that the obtained model can output the corresponding network state value based on the input network performance index.

[0031] In some embodiments, step 102 described above can also be implemented by dynamically fitting historical data, that is, by fitting a curve based on historical mapping data to obtain a suitable fitting curve, and then by mapping the collected network performance indicators through the fitting curve to obtain the corresponding network state value.

[0032] In some embodiments, before performing step 102 above, the following processing may also be performed: performing continuous throughput calculation on the time-series throughput curve data in the network performance metrics to obtain the continuous throughput metric under the current operating scenario; in response to the ratio between the instantaneous download speed in the network performance metrics and the continuous throughput metric being greater than a preset ratio, replacing the instantaneous download speed in the network performance metrics with the continuous throughput metric.

[0033] In this way, continuous throughput calculation can smooth out instantaneous fluctuations, reflect the actual transmission capacity of the network over a period of time, avoid evaluation bias caused by sudden changes in instantaneous speed, and replace the data only when the instantaneous speed is significantly higher than the continuous throughput through the ratio judgment mechanism. This not only preserves the reference value of instantaneous speed in sudden high throughput scenarios, but also prevents abnormal peaks from interfering with the overall evaluation. Furthermore, the dynamic adjustment method makes the network performance indicators more accurately reflect the user's continuous experience, providing a more accurate basis for network quality optimization and decision-making. At the same time, it enhances the adaptability of the evaluation system to the load network environment, thereby effectively improving the stability and reliability of network performance evaluation.

[0034] It should be noted that time-series throughput curve data refers to the change data of network throughput (the amount of data transmitted per unit time) recorded in chronological order over a period of time. It is usually plotted as a curve with time on the horizontal axis and throughput on the vertical axis. Instantaneous download speed can be the average download speed or peak download speed output by the speed test system.

[0035] As an example, assuming a home WiFi scenario, the speed test software records the download speed once per second during a 10-second download test, obtaining time-series throughput curve data (second 1: 50Mbps; second 2: 45Mbps; second 3: 5Mbps; second 4: 48Mbps; second 5: 52Mbps; ...). Subsequently, the system performs continuous throughput calculation on the time-series throughput curve data, which can be referred to in formula (1), and obtains the continuous throughput index sustained as 48Mbps. Afterwards, the system measures the peak download speed of 52Mbps (i.e., instantaneous) in this speed test. The download speed is measured, and the ratio between the instantaneous download speed and the sustained throughput is 1.08, which is less than the preset ratio (e.g., 1.3). This means the ratio does not exceed the preset threshold, so the system determines that the speed of this test is relatively stable and retains the original instantaneous download speed (or takes the average value) as the final indicator. If the system measures a peak download speed of 70Mbps (i.e., instantaneous download speed) in this test, the ratio between the instantaneous download speed and the sustained throughput is 1.45, which is greater than the preset ratio (1.3). Therefore, the instantaneous download speed of 70Mbps is replaced with the sustained throughput indicator of 48Mbps.

[0036] (1) (2) (3) in, It refers to the 50th percentile throughput value, that is, the median of the throughput value during the speed measurement process. For details, please refer to formula (2). It refers to the 90th percentile throughput value, which is the representative value of the higher throughput range during the speed measurement process. For details, please refer to formula (3). It refers to the i-th sample value after all sample values ​​are sorted in ascending order; This represents the number of sampled values.

[0037] For example, if a speed measurement yields 10 sampled values ​​S=[18,20,10,15,22,17,12,19,8,16], i.e., n=10, after sorting in ascending order, we can obtain S_sorted=[8,10,12,15,16,17,18,19,20,22]. Then, according to the above formulas (2) and (3), we know that... P50 sort[5] = 17Mbps, The value of sort[9] is 22Mbps.

[0038] Step 103: For each business type, determine the first weight that matches the business type.

[0039] It should be noted that the business type refers to the specific classification of network application scenarios, such as online games, video calls, audio communication, video playback, web browsing, etc.; the aforementioned first weight can be obtained directly through the preset weight of the corresponding business type, or it can be dynamically determined based on the actual network parameters, without specific limitations here.

[0040] As an example, taking the acquisition of preset weights for corresponding business types as an example, assuming that in the client scenario, the first weights for the network state values ​​corresponding to download speed, queuing latency, download load jitter, and packet loss rate during the video playback business calculation process are 0.5:0.2:0.2:0.1; and for the video call business calculation process, the first weights for the network state values ​​corresponding to upload speed, download speed, latency, upload load jitter, download load jitter, and packet loss rate are 0.5:0.2:0.2:0.1. The first weights corresponding to the state values ​​are 0.3:0.2:0.2:0.15:0.1:0.05; the first weights corresponding to the network state values ​​for latency, idle jitter, packet loss rate, and queuing latency during the online game service calculation process are 0.4:0.25:0.25:0.1; the first weights corresponding to the network state values ​​for experiment, packet loss rate, idle jitter, and upload speed during the voice communication service calculation process are 0.35:0.35:0.2:0.1. Assuming that in a web-based scenario, the first weights for the network state values ​​corresponding to download speed, queuing latency, and download load jitter during video playback calculation are 0.6:0.25:0.15; in a video call scenario, the first weights for the network state values ​​corresponding to upload speed, download speed, latency, upload load jitter, and download load jitter are 0.35:0.25:0.2:0.1:0.1; in an online game scenario, the first weights for the network state values ​​corresponding to latency, idle jitter, and queuing latency are 0.45:0.35:0.2; and in a voice communication scenario, the first weights for the network state values ​​corresponding to upload speed, idle jitter, and upload speed are 0.45:0.35:0.2.

[0041] In some embodiments, the determination of the first weight matching the service type in step 103 above can be achieved by: obtaining the initial weight corresponding to the service type based on a preset weight configuration table; and normalizing the initial weight to obtain the first weight matching the service type in response to the absence of network performance indicators matching the service type in the current operating scenario.

[0042] In this way, by pre-setting the initial weights of the business types and then normalizing the weights when the corresponding indicators are missing, the weight allocation of each indicator can be dynamically adjusted. This avoids distortion or calculation errors in network state value prediction caused by missing data, thus ensuring that even if some performance indicators cannot be obtained, the system can still provide reasonable and stable evaluation results based on the existing data. This enhances the predictive model's adaptability to complex network environments and incomplete data, and ensures the accuracy and reliability of the prediction results.

[0043] As an example, assuming the current operating scenario is a client-side scenario and the service type is video call, the initial weights corresponding to the current service type can be obtained from the preset weight configuration as [latency: jitter: bandwidth: packet loss rate = 0.4:0.3:0.2:0.1]. Subsequently, the system performs missing value detection on the network performance indicators matching this service type in the current operating scenario. For example, during speed testing, due to network fluctuations or device limitations, the packet loss rate indicator may be missing. In this case, the system normalizes the initial weights and proportionally distributes the proportion of the missing packet loss rate indicator to other indicators, resulting in a redistributed weight of [latency: jitter: bandwidth = 0.5:0.3:0.2], which is the first weight matching the service type. Here, the allocation principle can be based on the original weight ratio or according to the preset allocation rules.

[0044] In some embodiments, the determination of the first weight matching the service type in step 103 above can also be achieved by: extracting features from each network performance indicator matching the service type to obtain a first feature vector corresponding to each network performance indicator; performing attention calculation on the first feature vectors corresponding to multiple network performance indicators to obtain a first weight matching the service type.

[0045] In this way, through feature extraction, the original network performance indicators can be transformed into a first feature vector containing deep information such as their statistical characteristics and distribution patterns. This allows subsequent calculations to be based on more representative and discriminative data. At the same time, by introducing attention calculation, the model can automatically learn the importance of each indicator to the current business type, avoiding the limitations of manually setting weights. This allows the model to adapt to the performance evaluation needs of different business scenarios, thereby improving the accuracy and relevance of the prediction results and enhancing the model's adaptability to complex and ever-changing network environments.

[0046] It should be noted that attention computation can employ self-attention mechanisms, multi-head attention mechanisms, dynamic attention mechanisms, graph attention networks, etc., without making specific limitations here.

[0047] As an example, assuming the current operating environment is a client and the service type is video playback, the network performance metrics matching the current service type can be obtained, including download speed, queuing latency, download load jitter, and packet loss rate. Subsequently, feature extraction is performed on each network performance metric to obtain the first feature vector corresponding to the current network performance metric. That is, we obtain four first eigenvectors. Then, for the four first eigenvectors To calculate attention, the model can calculate an attention score for each network performance metric and normalize these four attention scores to obtain the first weight that matches the current video playback service type as [download speed: queuing latency: download load jitter: packet loss rate = 0.3: 0.3: 0.2: 0.2].

[0048] In some embodiments, the above-described attention calculation of the first feature vectors corresponding to multiple network performance indicators to obtain a first weight matching the service type can be implemented in the following way: forming a first matrix from the first feature vectors corresponding to multiple network performance indicators; extracting features from the service type to obtain a second feature vector corresponding to the service type; performing a dot product operation between the second feature vector and the transpose of the first matrix to obtain a second matrix; normalizing each element in the second matrix to obtain a third matrix, and extracting the first weight matching the service type from the third matrix.

[0049] Thus, by assembling the first feature vectors of multiple network performance indicators into a first matrix, the multidimensional features of the current network state can be comprehensively represented. Furthermore, by extracting features from the service type to obtain the second feature vector, the characteristics and demand tendencies of the service itself can be captured. Simultaneously, by performing a dot product operation between the second feature vector and the transpose of the first matrix, the correlation strength between the current service type and various network indicator features can be obtained. After normalization, a third matrix is ​​obtained, where each dimension's data value intuitively reflects the importance of each network indicator to the current service, thereby accurately extracting the first weight suitable for the current service type. This solution eliminates the subjectivity and rigidity of manually setting weights, allowing weight allocation to adaptively adjust according to real-time network conditions and service characteristics, thereby significantly improving the accuracy and intelligence of network experience prediction in multi-service scenarios.

[0050] As an example, assuming the current operating environment is a client and the business type is video playback, by extracting features from the current business type, a second feature vector can be obtained. Furthermore, by extracting features from each network performance metric, we can obtain the first feature vector corresponding to each network performance metric. Assuming the current service type yields three first feature vectors, they are as follows: , , Here, each first eigenvector has a dimension of 1×n, where n is a positive integer; subsequently, the above three first eigenvectors are combined to form a first matrix. for[ , , The dimension is 3×n, and the second feature vector is The dimension is 1×n; then, the second feature vector With the first matrix Performing a dot product on the transpose of the matrix yields the second matrix. The dimension is 1×3, and the second matrix is... The third matrix can be obtained by normalizing using the softmax function. The dimension is 1×3; finally, the value of each element is extracted from the third matrix to obtain the first weight that matches the business type, for example, the third matrix. Given [0.5, 0.3, 0.2], we can obtain... , , The corresponding weights are 0.5, 0.3, and 0.2, respectively.

[0051] Step 104: Based on the first weight, the network state values ​​corresponding to the network performance indicators matching the service type are weighted and summed to obtain the initial comprehensive network state value of the service type.

[0052] As an example, assuming that in a web-based scenario, the first weights for the network state values ​​corresponding to download speed, queuing latency, and download load jitter in a video playback service are 0.6:0.25:0.15, then based on these first weights, the network state value corresponding to download speed... Network state values ​​corresponding to queuing delay and the network status values ​​corresponding to download load jitter. By performing a weighted summation, we can obtain the initial comprehensive network state value for the video playback service. for .

[0053] Step 105: For each of the service types, in response to the network state value corresponding to the packet loss rate in the service type being less than a preset network state threshold, perform a multiplicative penalty on the initial comprehensive network state value of the service type to obtain the final comprehensive network state value of the service type.

[0054] As an example, in a client-side video call scenario, the system calculates an initial comprehensive network state value of 85, while the network state value corresponding to the packet loss rate is 45 (lower than the preset network state value threshold of 60), indicating a significant packet loss problem in the current network. At this point, the system obtains the target penalty coefficient based on a preset mapping relationship (e.g., a penalty coefficient of 0.7 corresponding to a network state value < 60), multiplies the initial comprehensive network state value of 85 by 0.7, and obtains a final comprehensive network state value of 59.5. This effectively avoids high bandwidth masking network quality issues, making the prediction results closer to the actual user experience.

[0055] In some embodiments, the multiplicative penalty processing of the initial integrated network state value of the service type in step 105 above to obtain the final integrated network state value of the service type can be implemented in the following way: based on a preset penalty coefficient mapping relationship associated with the packet loss rate, the packet loss rate is mapped to obtain a target penalty coefficient; the initial integrated network state value of the service type is multiplied by the target penalty coefficient to obtain the final integrated network state value of the service type.

[0056] In this way, by determining the penalty coefficient through a preset mapping relationship, the initial integrated network state value can be significantly reduced, which can effectively prevent misjudgments caused by high bandwidth or low latency masking packet loss problems. Furthermore, the multiplicative penalty mechanism can make the prediction results closer to the real business experience, thereby improving the accuracy and credibility of the prediction model and ensuring the objectivity and rationality of user experience prediction.

[0057] As an example, in a real-time voice call scenario on the client side, even if the network bandwidth is sufficient and the latency is low, if the packet loss rate is high (e.g., exceeding 2%), the voice will be intermittent or have noise, and the user experience will be severely degraded. Therefore, through a preset mapping relationship (e.g., a penalty coefficient of 1.0 for a packet loss rate of 0-1%, 0.8 for 1-3%, and 0.5 for more than 3%), the system will calculate the target penalty coefficient based on the actual packet loss rate (e.g., if the actual packet loss rate is 2.5%, the corresponding target penalty coefficient is 0.8). Then, the initial comprehensive network state value (e.g., the initial comprehensive network state value is 85) is multiplied by the target penalty coefficient (0.8) to obtain the final comprehensive network state value (85 × 0.8 = 68).

[0058] Step 106: Based on the final comprehensive network state value of each of the aforementioned service types, generate the multi-service user experience prediction result for the current operating scenario.

[0059] In some embodiments, step 106 described above can be implemented as follows: generating network status explanation text corresponding to the current operating scenario based on preset prompt words and the final comprehensive network status values ​​of multiple service types; performing verification processing on the network status explanation text based on preset verification rules to obtain a verification result; in response to the verification result being passed, using the network status explanation text as the multi-service user experience prediction result for the current operating scenario; in response to the verification result being failed, regenerating the network status explanation text for the current operating scenario and verifying the regenerated network status explanation text until the verification is passed, and using the regenerated network status explanation text as the multi-service user experience prediction result for the current operating scenario.

[0060] In this way, by utilizing preset prompts, the system can automatically generate explanatory text that fits the business scenario and is easy to understand based on the final network status values ​​of multiple services, thereby improving the readability and user-friendliness of the results. Furthermore, through preset verification rules, it can effectively avoid the risks of generating content that is misleading, ambiguous, or violates regulations. This ensures that the final output of multi-service user experience prediction results accurately reflects the real network situation and meets the system's preset quality standards and security specifications, thereby significantly enhancing the reliability and practicality of the prediction system.

[0061] As an example, assuming that the aforementioned prediction method yields a final comprehensive network state value of 85 for video playback, 72 for video call, 92 for online gaming, and 88 for voice call in the current operating scenario; then, based on preset prompts (e.g., "Based on the following network state values, generate a Chinese explanation of no more than 100 characters, indicating the service type with the worst experience and its main reasons"), the prompts and the final comprehensive network state value for each service type are input into the large model. The large model will then generate corresponding network state explanation text based on the final comprehensive network state value for each service type, such as "Your network performs well in video playback, with sufficient bandwidth and normal latency, allowing for relatively smooth video viewing; in video calls..." Currently, there may be slight latency and stuttering, mainly due to insufficient uplink speed; it performs excellently in online games, with low latency and minimal network jitter, allowing for smooth gameplay of large-scale online games; it performs well in voice calls, with low packet loss, enabling smooth voice calls. Then, according to preset verification rules (such as detecting prohibited words), the current network status explanation text is verified. If the verification result is successful, the current network status explanation text can be used as the multi-service user experience prediction result for the current operating scenario. If the verification fails, the prompt words are automatically adjusted (e.g., adding constraints), and the network status explanation text is regenerated based on the new prompt words. The regenerated network status explanation text is then verified according to preset verification rules until it passes. The verified network status explanation text is then used as the multi-service user experience prediction result for the current operating scenario.

[0062] Step 107: In response to the multi-service user experience prediction result indicating that there is an abnormal network status in the current operating scenario, determine the target service type of the abnormal network status.

[0063] As an example, suppose the multi-service user experience prediction result for the current operating scenario is: "Your network performs well in video playback, with sufficient bandwidth and normal latency, allowing for relatively smooth video viewing; its performance in video calls is average, with possible slight latency and stuttering, mainly due to insufficient uplink speed; its performance in online games is excellent, with low latency and minimal network jitter, allowing for smooth gameplay of large-scale online games; and its performance in voice calls is good, with low packet loss, allowing for smooth voice calls." Detection reveals an abnormal network condition in the current operating scenario, and the target service type with the abnormal network condition in the current operating scenario is identified as video calls.

[0064] Step 108: Based on the target service type and abnormal network performance indicators, generate a network state adjustment strategy to adjust the network state of the current operating scenario based on the network state adjustment strategy.

[0065] Following the example above, the system has identified video calls as the target service type with abnormal network status. By analyzing the corresponding network status value details, it has determined that the main abnormal network performance indicator causing poor experience is insufficient uplink speed. Based on this, the system generates a targeted network status adjustment strategy: "On the one hand, it issues control commands to the router or gateway to mark the user device's video call application as a high-priority queue through QoS (Quality of Service) policies and limits the bandwidth usage of background high-volume download tasks, prioritizing uplink resources; on the other hand, it pushes optimization suggestions to the user through the user terminal interface, prompting them to close cloud backup or video upload programs running in the background to release uplink bandwidth." Subsequently, the system executes this network status adjustment strategy, modifies the network configuration in real time, and re-monitors the uplink speed indicator of the video call after the adjustment to confirm that its network status value has recovered to above the normal threshold, thereby completing the dynamic closed-loop optimization of the network status of the current operating scenario.

[0066] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a multi-service user experience evaluation device based on network performance indicators, the structure of which is as follows: Figure 2 As shown.

[0067] Figure 2 This is a schematic diagram of the internal structure of a multi-service user experience evaluation device based on network performance indicators, provided as an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions that can be executed by at least one processor. The instructions are executed by at least one processor 201 to enable at least one processor 201 to perform the steps of the method corresponding to any of the above embodiments.

[0068] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium stores computer-executable instructions, which are configured to perform the steps of the method corresponding to any of the above embodiments.

[0069] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0070] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0071] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0073] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0075] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0076] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0077] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0078] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0079] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for optimizing performance of a multi-service network based on network performance indicators, characterized in that, The method includes: Obtain network performance metrics under the current operating scenario, and perform predictive processing on the network performance metrics to obtain the network state value corresponding to the network performance metrics; For each service type, a first weight matching the service type is determined, and based on the first weight, the network state values ​​corresponding to the network performance indicators matching the service type are weighted and summed to obtain the initial comprehensive network state value of the service type. For each of the aforementioned service types, in response to the network state value corresponding to the packet loss rate of the service type being less than a preset network state threshold, a multiplicative penalty is applied to the initial comprehensive network state value of the service type to obtain the final comprehensive network state value of the service type. Based on the final comprehensive network state value for each of the aforementioned service types, a multi-service user experience prediction result for the current operating scenario is generated. In response to the multi-service user experience prediction results indicating that there is a network status anomaly in the current operating scenario, the target service type of the network status anomaly is determined. Based on the target service type and abnormal network performance indicators, a network state adjustment strategy is generated to adjust the network state of the current operating scenario.

2. The method according to claim 1, characterized in that, The network performance metrics include basic network performance metrics and derived network performance metrics; The process of obtaining network performance metrics in the current operating scenario includes: Use speed testing software to obtain basic network performance indicators for the current operating scenario; The difference between load latency and idle latency in the basic network performance metrics is used as the queuing latency in the derived network performance metrics, where load latency is the maximum value of upload load latency and download load latency in the basic network performance metrics. The maximum value among the idle jitter, download jitter, and upload jitter in the basic network performance metrics is taken as the effective jitter in the derived network performance metrics.

3. The method according to claim 1, characterized in that, The step of predicting the network performance indicators to obtain the network state values ​​corresponding to the network performance indicators includes: For each of the aforementioned network performance metrics, the following processing is performed: Determine the target state value range corresponding to the network performance index from multiple preset state value ranges; Based on the target state value range, the network performance index is mapped to obtain the network state value corresponding to the network performance index.

4. The method according to claim 1, characterized in that, The determination of the first weight matching the business type includes: Based on a preset weight configuration table, obtain the initial weight corresponding to the business type; In response to the absence of network performance indicators matching the service type in the current operating scenario, the initial weights are normalized to obtain a first weight that matches the service type.

5. The method according to claim 1, characterized in that, The determination of the first weight matching the business type includes: For each network performance metric matched by the service type, feature extraction is performed to obtain a first feature vector corresponding to each network performance metric. Attention is calculated on the first feature vectors corresponding to multiple network performance metrics to obtain a first weight that matches the service type.

6. The method according to claim 5, characterized in that, The step of performing attention calculation on the first feature vectors corresponding to multiple network performance metrics to obtain a first weight that matches the service type includes: The first feature vectors corresponding to multiple network performance indicators are combined into a first matrix; Feature extraction is performed on the business type to obtain a second feature vector corresponding to the business type; The second eigenvector is multiplied by the transpose of the first matrix to obtain the second matrix. Normalize each element in the second matrix to obtain a third matrix, and extract the first weight that matches the business type from the third matrix.

7. The method according to claim 1, characterized in that, The step of performing a multiplicative penalty on the initial integrated network state value of the service type to obtain the final integrated network state value of the service type includes: Based on a preset mapping relationship of penalty coefficients associated with packet loss rate, the packet loss rate is mapped to obtain the target penalty coefficient. The initial integrated network state value of the service type is multiplied by the target penalty coefficient to obtain the final integrated network state value of the service type.

8. The method according to claim 1, characterized in that, Before performing predictive processing on the network performance metric to obtain the network state value corresponding to the network performance metric, the method further includes: Continuous throughput calculation is performed on the time-series throughput curve data in the network performance metrics to obtain the continuous throughput metrics under the current operating scenario. If the ratio between the instantaneous download speed and the sustained throughput in the network performance metrics is greater than a preset ratio, the sustained throughput in the network performance metrics will replace the instantaneous download speed.

9. The method according to claim 1, characterized in that, The process of generating multi-service user experience prediction results for the current operating scenario based on the final comprehensive network state value for each of the aforementioned service types includes: Based on preset prompt words and the final comprehensive network status values ​​of multiple service types, generate network status explanation text corresponding to the current operating scenario; Based on preset verification rules, the network state explanation text is verified to obtain the verification result; In response to the verification result being passed, the network status explanation text is used as the multi-service user experience prediction result for the current operating scenario; In response to the verification result being unsuccessful, the network status explanation text for the current operating scenario is regenerated, and the regenerated network status explanation text is verified until the verification passes. The regenerated network status explanation text is then used as the multi-service user experience prediction result for the current operating scenario.

10. A multi-service network performance optimization device based on network performance indicators, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-9.

Images