A Smart Service Area Traffic Transmission Monitoring Method Based on Big Data Analysis

Abstract

This invention relates to the field of telecommunications technology, specifically to a service area traffic intelligent transmission monitoring method based on big data analysis. The method includes: forming a sequence of the same network performance indicators for the same service type at different times within the same area of ​​the service area; each network performance indicator within the sequence is considered a sample; obtaining the feature vector of the sample; establishing first and second prediction models corresponding to each sequence; obtaining a current window containing the current time in a sequence, then combining the two prediction models to obtain a predicted network performance indicator sequence corresponding to the current window, and further obtaining the network performance indicator threshold corresponding to the current window; and determining whether there is traffic transmission anomaly for that service type in the area based on the network performance indicators at the current time in different sequences corresponding to a service type in a given area and the network performance indicator thresholds corresponding to the current windows in different sequences. This application can accurately monitor the traffic transmission status of a service area.

Description

Technical Field

[0001] This invention relates to the field of telecommunications technology, specifically to a method for intelligent transmission monitoring of service area traffic based on big data analysis. Background Technology

[0002] Highway service areas are important traffic nodes, and their networks carry various critical services such as video surveillance, toll collection systems, and internet access for users. Due to the strong spatiotemporal fluctuations in traffic flow at service areas (e.g., traffic volume varies drastically with vehicle traffic, peak and off-peak hours, seasons, and holidays), the diversity of services (e.g., different services such as video surveillance, transactions, and internet access have vastly different requirements for bandwidth, latency, and packet loss rate), and the significant environmental complexity (affected by weather conditions), service area networks are highly complex. Existing traffic transmission monitoring methods are insufficient to accurately assess the traffic transmission status of service area networks, resulting in unsatisfactory monitoring results.

[0003] Existing methods for monitoring traffic flow in highway service areas often employ simple threshold alarms or detection based on fixed models, assessing the network traffic flow status based on static and single thresholds. However, service area traffic exhibits strong spatiotemporal fluctuations and significant business variations. Assessing the network traffic flow status based on a single and static threshold results in substantial discrepancies between the assessment results and the actual situation, failing to guarantee effective monitoring of service area traffic flow. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a service area traffic intelligent transmission monitoring method based on big data analysis. The specific technical solution adopted is as follows:

[0005] One embodiment of the present invention provides a service area traffic intelligent transmission monitoring method based on big data analysis, the method comprising:

[0006] Obtain different network performance indicators and service types within the service area; within the same area and under the same service type, the same network performance indicator at different times forms a sequence; each network performance indicator within the sequence is a sample.

[0007] The temporal characteristics of a sample are obtained based on the time of its collection; the spatial correlation characteristics of a sample are obtained based on the adjacent regions of the region where the sample is located; and the environmental characteristics of a sample are obtained based on the weather conditions at the time of its collection.

[0008] Based on the temporal features, spatial correlation features, environmental features, and the sample itself, the feature vector of the sample is obtained; a first prediction model and a second prediction model are established for each sequence.

[0009] The time window is slid across a sequence, and the time window that reaches the current moment is recorded as the current window. The feature vectors of the samples in the current window are input into the corresponding first and second prediction models respectively to obtain the prediction network performance index sequence. The network performance index threshold is obtained by using the prediction network performance index sequence corresponding to a current window.

[0010] Based on the network performance indicators at the current moment of different sequences corresponding to a service type in a region, and the network performance indicator thresholds corresponding to each current window in different sequences, it is determined whether there is an abnormal traffic transmission for that service type in that region.

[0011] Preferably, different network performance indicators and service types include:

[0012] Different network performance metrics include bandwidth, latency, and packet loss rate; different service types include video services, pay-per-view services, and internet access services.

[0013] Preferably, obtaining the temporal characteristics of a sample based on its collection time includes:

[0014] A first period and a second period are set. Based on the collection time of a sample, sine and cosine coding are used to encode the first sine and first cosine values ​​using the first period. Based on the collection time of the sample, sine and cosine coding are used to encode the second sine and second cosine values ​​using the second period. A holiday factor is set for the sample according to whether the collection time of the sample belongs to a holiday and the length of the holiday. The first sine, second cosine, second sine, second cosine and holiday factor corresponding to a sample are combined to form the time feature of the sample.

[0015] Preferably, the spatial correlation features of a sample are obtained based on the adjacent regions of the region where the sample is located, including:

[0016] The bandwidth of a sample in a region is used to form the bandwidth time series of the region under the Internet access service type at each time point. The adjacent regions of the region are obtained, which are the regions that are geographically adjacent to the region. The Pearson correlation coefficients of the bandwidth time series of the region and the bandwidth time series of each of the adjacent regions are calculated and summed to obtain the spatial correlation characteristics of all samples in the region.

[0017] Preferably, the environmental characteristics of a sample are obtained based on the weather conditions at the time of sample collection, including:

[0018] The environmental characteristics of a sample are encoded as the weather conditions at the time of sample collection.

[0019] Preferably, the sequence of predicted network performance metrics is obtained, including:

[0020] The feature vector of the sample in the current window of a sequence is input into the first prediction model corresponding to the sequence to obtain the performance index of each first prediction network. The performance indexes of each first prediction network are arranged in time sequence to form the first prediction network performance index sequence. Similarly, the feature vector of the sample in the current window of the sequence is input into the second prediction model corresponding to the sequence to obtain the second prediction network performance index sequence.

[0021] The mean squared error of the first prediction model is obtained based on the network performance indicators at each time point before the current window in the sequence and the first prediction network performance indicator predicted by the first prediction model. The mean squared error of the second prediction model is obtained based on the network performance indicators at each time point before the current window in the sequence and the second prediction network performance indicator predicted by the second prediction model. The mean squared error of the first prediction model is divided by the sum of the mean squared errors of the first and second prediction models to obtain the first feature value. The difference between the preset value and the first feature value is the weight of the first prediction model corresponding to the sequence. Similarly, the weight of the second prediction model corresponding to the sequence is obtained.

[0022] By using the weights of the first prediction model and the weights of the second prediction model, the first and second prediction network performance indicators at the same time in the first and second prediction network performance indicator sequences are weighted and summed to obtain the prediction network performance indicators at each time, thus forming the prediction network performance indicator sequence.

[0023] Preferably, obtaining the network performance indicator threshold using the predicted network performance indicator sequence corresponding to a current window includes:

[0024] For a sequence of predicted network performance metrics corresponding to the current window, obtain the standard deviation and mean of the predicted network performance metrics sequence. The difference between the mean and the standard deviation of a preset multiple is the lower limit value corresponding to the current window, and the sum of the standard deviations of the mean and the preset multiple is the upper limit value corresponding to the current window. If the predicted network performance metric is bandwidth, the lower limit value corresponding to the current window is the network performance metric threshold value. If the predicted network performance metric is latency, the upper limit value corresponding to the current window is the network performance metric threshold value. If the predicted network performance metric is packet loss rate, the upper limit value corresponding to the current window is the network performance metric threshold value.

[0025] Preferably, determining whether there is traffic transmission anomaly for that service type in a region based on the network performance indicators at the current time on different sequences corresponding to a service type in a region and the network performance indicator thresholds corresponding to each current window on different sequences includes:

[0026] If the sequence of network performance indicators corresponding to a service type in a region is a sequence composed of bandwidth, and the network performance indicator at the current moment in the sequence is less than or equal to the network performance indicator threshold corresponding to the current window in the sequence, then there is an anomaly in the traffic transmission of that service type in that region. If the sequence of network performance indicators corresponding to a service type in a region is a sequence composed of latency, and the network performance indicator at the current moment in the sequence is greater than or equal to the network performance indicator threshold corresponding to the current window in the sequence, then there is an anomaly in the traffic transmission of that service type in that region. If the sequence of network performance indicators corresponding to a service type in a region is a sequence composed of packet loss rate, and the network performance indicator at the current moment in the sequence is greater than or equal to the network performance indicator threshold corresponding to the current window in the sequence, then there is an anomaly in the traffic transmission of that service type in that region.

[0027] The embodiments of this invention have at least the following beneficial effects: This application collects different network performance indicators for each service type and region within each service area. The same network performance indicators for the same service type within the same region at different times are grouped into a sequence. Analyzing each sequence allows for a more precise determination of which region and service type has an abnormal traffic transmission status. Furthermore, it obtains the temporal characteristics, spatial correlation characteristics, and environmental characteristics of each sample, and combines these with the sample itself to obtain the sample's feature vector. By comprehensively considering the influence of multiple factors on network performance indicators, the accuracy of subsequent predictions is improved. Then, a first prediction model and a second prediction model are established for each sequence. A time window is used to slide across a sequence, and the time window that slides to the current time is denoted as... In the current window, the feature vectors of the samples within the current window are input into the corresponding first and second prediction models to obtain a predicted network performance index sequence. Using both models for prediction and then combining them improves the accuracy of network performance index prediction. Finally, a network performance index threshold is obtained based on the predicted network performance index sequence corresponding to a current window. Then, based on the network performance index at the current moment on different sequences corresponding to a service type in a region and the network performance index thresholds corresponding to each current window on different sequences, it is determined whether there is an anomaly in traffic transmission for that service type in that region. This application adapts to the characteristics of strong spatiotemporal fluctuations and large service differences in service area traffic, greatly improving the accuracy of service area network traffic transmission status assessment and ensuring the effectiveness of service area traffic transmission monitoring. Attached Figure Description

[0028] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 A flowchart illustrating a service area traffic intelligent transmission monitoring method based on big data analysis, provided as an embodiment of the present invention. Detailed Implementation

[0030] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a service area traffic intelligent transmission monitoring method based on big data analysis proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0031] 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 invention pertains.

[0032] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent transmission monitoring method for service area traffic based on big data analysis provided by this invention.

[0033] Example:

[0034] The main application scenarios of this invention are:

[0035] Due to the strong spatiotemporal fluctuations and diverse service characteristics of traffic in service areas, a single, static threshold cannot guarantee the accuracy of service area network traffic transmission status assessment. Therefore, it is necessary to dynamically predict thresholds for various service types in order to achieve accurate assessment of service area network traffic transmission status based on precise thresholds, thereby ensuring the effectiveness of service area traffic transmission monitoring.

[0036] Please see Figure 1 The diagram illustrates a method flowchart for intelligent transmission monitoring of service area traffic based on big data analysis, according to an embodiment of the present invention. The method includes the following steps:

[0037] Step S1: Obtain different network performance indicators for the service area; the same network performance indicators for the same service type in the same area at different times form a sequence; each network performance indicator in the sequence is a sample.

[0038] The main purpose of this application is to dynamically predict the thresholds of various network performance indicators for different service types in the service area network, so as to improve the accuracy of the assessment of the traffic transmission status of the service area network.

[0039] Because the service area contains multiple functional zones with varying network usage, it is necessary to divide the service area into zones. Specifically, zones are divided according to their functional type and location, such as parking areas, restroom areas, restaurant areas, accommodation areas, etc. The specific division needs to be determined based on the specific structure and distribution of the service area. Zones with the same function but different locations cannot be grouped together. After the zones are divided, they need to be coded, and the zone code is denoted as GID. For example, the GID of parking lot 1 is Z01, the GID of parking lot 2 is Z02, and the GID of restaurant area 1 is C01. The specific coding can be determined by the implementer.

[0040] Simultaneously, it is necessary to determine the service types. The service types of network traffic in highway service areas mainly include video services, toll transaction services, and user internet access services, which are also key monitoring targets. Therefore, in this embodiment, video services, toll transaction services, and user internet access services are analyzed, and these three service types are coded as 01, 02, and 03, respectively. Furthermore, the network service situation in different areas within the service area is also related to weather conditions. For example, people are more likely to be indoors on rainy or snowy days, so environmental data also needs to be collected. Bandwidth, latency, and packet loss rate are commonly used indicators of network performance; therefore, bandwidth, latency, and packet loss rate are used as three network performance indicators for analysis.

[0041] The weather conditions at a given moment are coded based on environmental data. For example, sunny days are coded as 1, rainy days as 2, and snowy days as 3. It should be noted that here, only sunny, cloudy, rainy, and snowy days are used as the standard for weather condition coding. Implementers can also make more detailed divisions, such as adding temperature as part of the weather condition coding.

[0042] Therefore, by setting a sampling interval, three different network performance indicators can be collected at various times within a region and for a single service type. For a given time within a region and for a single service type, three different network performance indicators can be collected, represented by bandwidth, etc. Delay and packet loss rate Where t is the data collection time, GID is the region code, k is the service type code, and W is the weather condition code.

[0043] To facilitate subsequent analysis, the same network performance indicators at different times under the same service type in the same area within the service area are grouped into a sequence. Thus, one service type in one area corresponds to three sequences. There are a total of 3 service types, so 9 different sequences will be collected for one area, and one performance indicator in the sequence is used as a sample.

[0044] Step S2: Obtain the temporal characteristics of a sample based on the collection time of the sample; obtain the spatial correlation characteristics of a sample based on the adjacent regions of the region where the sample is located; obtain the environmental characteristics of a sample based on the weather conditions at the time of collection.

[0045] After obtaining the sequence corresponding to each network performance index for each service type in each area within the service area, it is necessary to extract features from each sample in the sequence.

[0046] Specifically, the first step is to extract the temporal features of a sample. Because service area traffic data exhibits strong spatiotemporal fluctuations, network performance indicators also fluctuate over time. Therefore, by extracting temporal features from network performance indicators, we can quantify their temporal periodic characteristics, improve the model's ability to capture patterns, and thus enhance the model's prediction accuracy.

[0047] Because network performance metrics exhibit periodicity, sine and cosine coding are used to encode these metrics over a first period of 24 hours and a second period of 7 days, respectively. Specifically, sine and cosine coding are used to encode the first sine and first cosine values ​​based on the collection time of a sample within the first period, and the second sine and second cosine values ​​are obtained based on the collection time of a sample within the second period, thus better representing the periodic relationship. Furthermore, traffic fluctuations in service areas during holidays lead to significant fluctuations in network performance metrics; therefore, based on the network performance metrics... The timestamp determines whether the network performance metric falls within a holiday period and the duration of that holiday (as stipulated by the state). A holiday factor is set based on the length of the holiday (excluding weekends). The longer the holiday period, the larger the holiday factor. For example, a non-holiday holiday is 1, a single-day holiday is 1.1, a three-day holiday is 1.3, a five-day holiday is 1.5, and a holiday exceeding five days is 1.7. In other words, the holiday factor for a sample is set based on whether the sample's collection time falls within a holiday period and the length of the holiday. The time feature of a sample is composed of the first sine, second cosine, second sine, second cosine, and the holiday factor. The time feature is represented as follows: , To represent the temporal characteristics of a sample, and These are the first sine and first cosine values ​​obtained by converting over a 24-hour period, respectively. and These are the second sine and second cosine values ​​obtained by converting over a 7-day period, respectively. This represents the holiday factor. From this, we can obtain the time characteristics of each sample, which are actually a vector.

[0048] Furthermore, since traffic between regions can influence each other (e.g., when vehicles and people move within a service area, their network usage behavior propagates along their movement paths between adjacent areas), and changes in network bandwidth directly reflect changes in traffic, spatial correlation features are extracted based on the bandwidth correlation between regions.

[0049] Since the core objective of spatial correlation features is to capture the movement patterns of people within a service area (e.g., people moving from the parking lot to the restaurant, then to the restroom, and then to the convenience store), it is necessary to select services that can reflect individual movement throughout the entire process. Because internet access services have full coverage and strong mobility, with users' mobile phones connected to WiFi throughout (from entry to exit) and mobile devices automatically switching access points (seamless roaming), the bandwidth sequence corresponding to internet access services can completely record the user's trajectory. Other services, however, have fixed locations or limited mobility, such as fixed-location video surveillance cameras and payment systems that only operate within fixed power grids. Therefore, it is necessary to analyze the bandwidth of internet access behavior in each area, constructing a bandwidth time series for each area based on the bandwidth of each internet access service type at different times. This allows us to obtain the bandwidth time series for each area.

[0050] Next, the adjacent regions of a sample's region are obtained. The adjacent regions are those that are geographically adjacent to the sample's region. The Pearson correlation coefficients of the bandwidth time series of the sample's region and the bandwidth time series of each of its adjacent regions are calculated and summed to obtain the spatial correlation characteristics of the sample's region. The spatial correlation characteristics of a region are also the spatial correlation characteristics of all samples within that region.

[0051] The Pearson correlation coefficient is defined as the product of the covariance of two sets of data divided by their respective standard deviations. It quantifies the similarity of bandwidth change trends between two regions. A Pearson correlation coefficient closer to 1 indicates a stronger positive correlation in bandwidth changes between the two regions. A Pearson correlation coefficient closer to -1 indicates a stronger negative correlation in bandwidth changes between the two regions. A Pearson correlation coefficient close to 0 indicates no linear correlation in traffic changes between the two regions. Therefore, by calculating the Pearson correlation coefficient of the bandwidth time series of a region and its adjacent regions, the degree of linear correlation between them can be quantified. If the bandwidth changes of two regions (such as a parking lot and a restaurant) are highly positively correlated (Pearson correlation coefficient close to 1), it indicates that the two regions are synchronous in bandwidth changes and may have a strong spatial correlation (for example, after vehicles arrive at the parking lot, people will go to the restaurant, causing the bandwidth of both regions to increase simultaneously).

[0052] If the spatial correlation feature of a sample within a region is positive and the larger the value, the greater the positive correlation influence of the adjacent regions on the traffic transmission in that region. If it is negative and the smaller the value, the greater the negative correlation influence of the adjacent regions on the traffic transmission in that region.

[0053] Meanwhile, the weather conditions at the time of sampling also have a certain impact on traffic transmission. Therefore, it is necessary to obtain the environmental characteristics of a sample based on the weather conditions at the time of sample collection. The environmental characteristics of a sample are encoded as the weather conditions at the time of sample collection. Thus, the temporal characteristics, spatial correlation characteristics, and environmental characteristics of each sample can be obtained.

[0054] Step S3: Obtain the feature vector of a sample based on its temporal features, spatial correlation features, environmental features, and the sample itself; establish the first prediction model and the second prediction model corresponding to each sequence.

[0055] The above steps obtain the temporal features, spatial correlation features, and environmental features of each sample, which can then be used to construct the feature vector of each sample. Specifically, the feature vector of each sample is constructed by sequentially combining the temporal features, spatial correlation features, environmental features, and the sample itself.

[0056] Because network performance indicators representing traffic data transmission in highway service areas include both linear components (such as daily periodicity) and nonlinear components (such as sudden traffic surges and weather effects), a combination of VARIMA and GRU models is used for prediction. Specifically, the first prediction model is a VARIMA model, and the second is a GRU model. The VARIMA model excels at capturing linear patterns (such as trends and seasonality), while the GRU model excels at capturing complex nonlinear patterns (such as sudden traffic surges during holidays and nonlinear fluctuations), and has good modeling capabilities for long-term dependencies. A dynamic weighting mechanism is also established, allocating weights based on the recent performance (prediction error) of the two models when combining their predictions. This adapts to changes in traffic patterns (such as the transition from a stable period to holidays), allowing the prediction model to adjust adaptively to changes in service area traffic patterns and ensuring prediction accuracy.

[0057] It should be noted that there are three types of services in a region, and each type of service has three network performance indicators. That is, there are nine sequences in a region. Due to the different types of services and network performance indicators, for a sequence, it is necessary to construct the corresponding first prediction model and second prediction model.

[0058] Then, samples from a sequence are used to train the first and second prediction models corresponding to that sequence. The training methods for the first and second prediction models are the same for each sequence, only the samples used during training differ. The input to each prediction model is the feature vector of that sample. The structure and training methods of the first and second prediction models are existing technologies and will not be elaborated upon here. Furthermore, to ensure the determination of the thresholds for each subsequent network performance metric during training, the input to both prediction models is data from a period prior to the current time, and the number of output prediction data is the same as the number of input data.

[0059] Step S4: Slide a time window across a sequence, and record the time window that has slid to the current moment as the current window. Input the feature vectors of the samples in the current window into the corresponding first and second prediction models respectively to obtain the prediction network performance index sequence. Use the prediction network performance index sequence corresponding to a current window to obtain the network performance index threshold.

[0060] The above process yields the first and second prediction models for each sequence, which can then be used to predict network performance metrics at subsequent time points.

[0061] Specifically, since predictions need to be made based on subsequent network performance metrics, a time window needs to be set and slid across the sequence. When the time window reaches the current moment, it is recorded as the current window. In this application, the length of the current window is 2 minutes. Implementers can adjust the length of the window according to the required accuracy. In addition, the sliding step size is the same as the time length of the window.

[0062] Furthermore, the feature vectors of samples within the current window of a sequence are input into the first prediction model corresponding to the sequence to obtain the performance indicators of each first prediction network. The performance indicators of each first prediction network are arranged in time sequence to form a first prediction network performance indicator sequence. Similarly, the feature vectors of samples within the current window of the sequence are input into the second prediction model corresponding to the sequence to obtain a second prediction network performance indicator sequence.

[0063] After obtaining the first and second predicted network performance index sequences, the first and second predicted network performance indices at the same time in the two sequences need to be weighted and summed to obtain the predicted network performance index.

[0064] The mean squared error (MSE) of the first prediction model is obtained based on the network performance metrics at each time point before the current window in a sequence and the first prediction network performance metrics predicted by the first prediction model. The MSE of the second prediction model is obtained based on the network performance metrics at each time point before the current window in the sequence and the second prediction network performance metrics predicted by the second prediction model. The MSE of the first prediction model is divided by the sum of the MSEs of the first and second prediction models to obtain a first feature value. The difference between the preset value and the first feature value is the weight of the first prediction model for that sequence. Similarly, the weight of the second prediction model for that sequence is obtained, and the calculation method for the weight of the second prediction model is the same as that for the first prediction model. The preset value is 1; the larger the MSE of a model, the smaller its corresponding weight.

[0065] By using the weights of the first and second prediction models, the predicted network performance indicators at the same time point in both the first and second prediction models are weighted and summed to obtain the predicted network performance indicators at each time point, forming a predicted network performance indicator sequence. Essentially, the model with higher prediction accuracy has a larger weight, and then the network performance indicators predicted by the two models at a given time point are weighted and summed to obtain the comprehensive predicted network performance indicator.

[0066] For a sequence of predicted network performance metrics corresponding to the current window, obtain the standard deviation and mean of the predicted network performance metrics sequence. The difference between the mean and the standard deviation of a preset multiple is the lower limit value corresponding to the current window, and the sum of the standard deviations of the mean and the preset multiple is the upper limit value corresponding to the current window. If the predicted network performance metric is bandwidth, the lower limit value corresponding to the current window is the network performance metric threshold value. If the predicted network performance metric is latency, the upper limit value corresponding to the current window is the network performance metric threshold value. If the predicted network performance metric is packet loss rate, the upper limit value corresponding to the current window is the network performance metric threshold value.

[0067] The preset multiplier is 1.96. The method used to obtain the upper and lower limits for the current window is a commonly used method in statistics. Different network performance indicators have different physical meanings, so the threshold values ​​are selected differently. For bandwidth, when real-time bandwidth is lower than the normal bandwidth range, it indicates that the real-time bandwidth is below the service requirements, which will lead to increased channel contention and cause abnormal traffic transmission for various service types. Therefore, the lower limit is used as the threshold for bandwidth. When real-time latency is higher than the normal latency range, it will cause response timeouts for various services. Therefore, the upper limit is used as the threshold for latency. When the packet loss rate is higher than the normal packet loss rate range, video surveillance will show partial screen distortion, and the billing system will experience single-transaction packet loss. Therefore, the upper limit is used as the threshold for packet loss rate.

[0068] Step S5: Determine whether there is traffic transmission anomaly for this service type in this region based on the network performance indicators at the current time of different sequences corresponding to a service type in a region and the network performance indicator thresholds corresponding to each current window in different sequences.

[0069] The above steps obtain the network performance index threshold corresponding to the current window on the sequence of different network performance indices under a region and a business type.

[0070] Therefore, anomaly detection can be performed based on network performance indicator thresholds. The system determines whether there are traffic transmission anomalies for a particular service type in a given region based on the network performance indicators at the current moment across different sequences and the network performance indicator thresholds for each current window within those sequences. Specifically, if the sequence of network performance indicators for a particular service type in a given region is composed of bandwidth, then if the network performance indicator at the current moment is less than or equal to the network performance indicator threshold for the current window, then there are traffic transmission anomalies for that service type in that region. If the sequence of network performance indicators for a particular service type in a given region is composed of latency, then if the network performance indicator at the current moment is greater than or equal to the network performance indicator threshold for the current window, then there are traffic transmission anomalies for that service type in that region. If the sequence of network performance indicators for a particular service type in a given region is composed of packet loss rate, then if the network performance indicator at the current moment is greater than or equal to the network performance indicator threshold for the current window, then there are traffic transmission anomalies for that service type in that region.

[0071] In other words, if one of the three network performance indicators in a region and for a specific service type is abnormal and does not meet the requirements, then there is an abnormal traffic transmission for that region and service type. In this case, staff need to carry out maintenance to ensure normal traffic transmission.

[0072] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0073] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0074] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A service area traffic intelligent transmission monitoring method based on big data analysis, characterized in that, The method includes: Obtain different network performance indicators and service types within the service area; within the same area and under the same service type, the same network performance indicator at different times forms a sequence; each network performance indicator within the sequence is a sample. The temporal characteristics of a sample are obtained based on the time of its collection; the spatial correlation characteristics of a sample are obtained based on the adjacent regions of the region where the sample is located; and the environmental characteristics of a sample are obtained based on the weather conditions at the time of its collection. The feature vector of a sample is obtained based on its temporal features, spatial correlation features, environmental features, and the sample itself; a first prediction model and a second prediction model are established for each sequence; the first prediction model and the second prediction model are the VARIMA model and the GRU model, respectively. The time window is slid across a sequence, and the time window that reaches the current moment is recorded as the current window. The feature vectors of the samples in the current window are input into the corresponding first and second prediction models respectively to obtain the prediction network performance index sequence. The network performance index threshold is obtained by using the prediction network performance index sequence corresponding to a current window. Based on the network performance indicators at the current moment of different sequences corresponding to a service type in a region and the network performance indicator thresholds corresponding to each current window in different sequences, it is determined whether there is an abnormal traffic transmission for that service type in that region. The step of obtaining the temporal characteristics of a sample based on its acquisition time includes: A first period and a second period are defined. Using sine and cosine coding, a sample's acquisition time is encoded using the first period to obtain a first sine value and a first cosine value. Similarly, the sample's acquisition time is encoded using the second period to obtain a second sine value and a second cosine value. A holiday factor is set for the sample based on whether its acquisition time falls on a holiday and the length of the holiday. The first sine value, second cosine value, second sine value, second cosine value, and holiday factor corresponding to a sample constitute the sample's temporal feature. The method of obtaining the spatial correlation features of a sample based on the adjacent regions of the region where the sample is located includes: The bandwidth of a sample in a region is used to form the bandwidth time series of the region under the Internet access service type at each time point; the adjacent regions of the region are obtained, which are regions that are geographically adjacent to the region; the Pearson correlation coefficients of the bandwidth time series of the region and the bandwidth time series of each of the adjacent regions are calculated and summed to obtain the spatial correlation characteristics of all samples in the region; The obtained sequence of predicted network performance metrics includes: The feature vector of the sample in the current window of a sequence is input into the first prediction model corresponding to the sequence to obtain the performance index of each first prediction network. The performance indexes of each first prediction network are arranged in time sequence to form the first prediction network performance index sequence. Similarly, the feature vector of the sample in the current window of the sequence is input into the second prediction model corresponding to the sequence to obtain the second prediction network performance index sequence. The mean squared error of the first prediction model is obtained based on the network performance indicators at each time point before the current window in the sequence and the first prediction network performance indicator predicted by the first prediction model. The mean squared error of the second prediction model is obtained based on the network performance indicators at each time point before the current window in the sequence and the second prediction network performance indicator predicted by the second prediction model. The mean squared error of the first prediction model is divided by the sum of the mean squared errors of the first and second prediction models to obtain the first feature value. The difference between the preset value and the first feature value is the weight of the first prediction model corresponding to the sequence. Similarly, the weight of the second prediction model corresponding to the sequence is obtained. Using the weights of the first prediction model and the weights of the second prediction model, the first and second prediction network performance indicators at the same time in the first and second prediction network performance indicator sequences are weighted and summed to obtain the prediction network performance indicators at each time, thus forming the prediction network performance indicator sequence. The step of obtaining the network performance indicator threshold using the predicted network performance indicator sequence corresponding to a current window includes: For a sequence of predicted network performance metrics corresponding to the current window, obtain the standard deviation and mean of the predicted network performance metrics sequence. The difference between the mean and the standard deviation of a preset multiple is the lower limit value corresponding to the current window, and the sum of the standard deviations of the mean and the preset multiple is the upper limit value corresponding to the current window. If the predicted network performance metric is bandwidth, the lower limit value corresponding to the current window is the network performance metric threshold value. If the predicted network performance metric is latency, the upper limit value corresponding to the current window is the network performance metric threshold value. If the predicted network performance metric is packet loss rate, the upper limit value corresponding to the current window is the network performance metric threshold value.

2. The service area traffic intelligent transmission monitoring method based on big data analysis according to claim 1, characterized in that, The different network performance metrics and service types include: Different network performance metrics include bandwidth, latency, and packet loss rate; different service types include video services, pay-per-view services, and internet access services.

3. The service area traffic intelligent transmission monitoring method based on big data analysis according to claim 1, characterized in that, The process of obtaining the environmental characteristics of a sample based on the weather conditions at the time of sample collection includes: The environmental characteristics of a sample are encoded as the weather conditions at the time of sample collection.

4. The service area traffic intelligent transmission monitoring method based on big data analysis according to claim 1, characterized in that, The step of determining whether there is traffic transmission anomaly for a service type in a region based on the network performance indicators at the current time on different sequences corresponding to a service type in a region and the network performance indicator thresholds corresponding to each current window on different sequences includes: If the sequence of network performance indicators corresponding to a service type in a region is a sequence composed of bandwidth, and the network performance indicator at the current moment in the sequence is less than or equal to the network performance indicator threshold corresponding to the current window in the sequence, then there is an anomaly in the traffic transmission of that service type in that region. If the sequence of network performance indicators corresponding to a service type in a region is a sequence composed of latency, and the network performance indicator at the current moment in the sequence is greater than or equal to the network performance indicator threshold corresponding to the current window in the sequence, then there is an anomaly in the traffic transmission of that service type in that region. If the sequence of network performance indicators corresponding to a service type in a region is a sequence composed of packet loss rate, and the network performance indicator at the current moment in the sequence is greater than or equal to the network performance indicator threshold corresponding to the current window in the sequence, then there is an anomaly in the traffic transmission of that service type in that region.

Images