Intelligent router traffic management method for home scene

Abstract

The invention relates to the technical field of network communication and route management, in particular to an intelligent router flow management method for a home scene, which comprises the following steps that: a home router collects network activity data of equipment, including data flow size, connection frequency and equipment type, performs cleaning and formatting processing, and stores the data flow size, the connection frequency and the equipment type; formatted device behavior data is generated. According to the method for analyzing the behavior data of the home network equipment based on the long and short-term memory network, the prediction precision of the periodicity and dependency of user behaviors is improved, and the allocation of bandwidth resources is more prospective and dynamic. Specifically, resources can be automatically adjusted according to predicted data requirements, and particularly in the peak period of requirements, it is ensured that key applications such as video conferences and online education preferentially obtain necessary network resources. According to the flow management technology, the overall operation efficiency of the network is improved by optimizing bandwidth utilization and reducing network congestion.

Description

Technical Field

[0001] The present invention relates to the technical field of network communication and routing management, and in particular to a smart router traffic management method for home scenarios. Background Art

[0002] The field of network communication and routing management technology involves the design, optimization and implementation of various devices and protocols to ensure the effective transmission of data in the network. This field includes packet routing, flow control, bandwidth allocation, and connection stability and security management. Routing management is the core of network engineering, focusing on how to efficiently and reliably forward information between various network nodes.

[0003] However, most existing technologies rely on static rules and limited data processing capabilities, making it difficult to adapt to the rapid changes in user behavior and irregular high-traffic demands, resulting in poor results in real-time data processing and dynamic allocation of bandwidth resources. Especially in home network scenarios, the concurrent use of multiple devices often leads to unbalanced resource allocation, affecting network performance and stability. This inflexible bandwidth management method is particularly inadequate when dealing with emergencies, and is prone to insufficient bandwidth and reduced service quality. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose a smart router traffic management method for home scenarios.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a smart router traffic management method for home scenarios, comprising the following steps:

[0006] The home router collects the network activity data of the device, including the data flow size, connection frequency and device type, cleans and formats it, and generates formatted device behavior data;

[0007] Based on the formatted device behavior data, a long short-term memory network is used to process time series data, learn the periodicity and time dependency of user behavior, and generate a user behavior pattern; based on the user behavior pattern, the network bandwidth demand for each time period in the future is analyzed to generate a bandwidth demand prediction result;

[0008] Based on the bandwidth demand prediction results, a bandwidth allocation strategy is designed to generate a preliminary bandwidth allocation plan; the preliminary bandwidth allocation plan is optimized according to usage adjustments to give priority to meeting the needs of video conferencing and online education;

[0009] Continuously monitor the network environment, including latency and packet loss rate indicators, to obtain network performance results.

[0010] Preferably, the steps of obtaining the formatted device behavior data are:

[0011] The router scans all connected devices in the home network and records the network activity of each device, including the data flow size, connection frequency and device type of each device, forming a preliminary network activity record;

[0012] Based on the preliminary network activity records, perform data verification and time stamping, remove invalid and erroneous data records, and reorganize and classify the data according to device type and connection frequency to generate organized network activity data;

[0013] Based on the sorted network activity data, the data is standardized and arranged according to device identification numbers and time sequence to obtain formatted device behavior data.

[0014] Preferably, the steps of acquiring the user behavior pattern are:

[0015] Using a long short-term memory network model, identifying time series features in the formatted device behavior data, including periodicity and time dependency of user behavior, to obtain time series data after feature extraction;

[0016] Based on the time series data after feature extraction, the user behavior prediction score at each time point is calculated using the following formula:

[0017] S t =α·h t-1 +β·x t

[0018] Among them, S t is the user behavior score at time step t, h t-1 is the hidden state score of the previous time step, x t is the input feature of the current time step, α and β are adjustment coefficients;

[0019] Based on the user behavior scores, high-score intervals are marked as high-frequency activity time periods, and user behavior patterns are divided and identified.

[0020] Preferably, the steps of obtaining the bandwidth demand prediction result are:

[0021] Extracting features from the user behavior patterns, including peak activity periods and persistent behavior trends, to obtain key behavior feature data;

[0022] Based on the key behavior feature data, the bandwidth demand in each time period is calculated using the following formula:

[0023]

[0024] Among them, D t represents the predicted bandwidth demand in time period t, ai represents the influence coefficient of the ith behavior on the bandwidth demand, b it is the predicted activity of the ith behavior in time period t, c i is the normalized coefficient of the ith behavior, and N is the total number of behavior types;

[0025] Based on the predicted bandwidth demand, the data is integrated and the bandwidth peak and valley values ​​are analyzed to obtain a bandwidth demand prediction result.

[0026] Preferably, the steps for obtaining the preliminary bandwidth allocation scheme are:

[0027] Analyze the bandwidth demand forecast results, determine the bandwidth demand peak and valley values ​​in each time period, and obtain demand peak and valley data;

[0028] Based on the demand peak and valley data, the bandwidth allocation weight for each time period is calculated using the following formula:

[0029]

[0030] Among them, W t Assign a weight to the bandwidth in time period t, R max is the predicted maximum bandwidth requirement, R t is the bandwidth requirement for time period t, ∈ is a constant to prevent the denominator from being zero;

[0031] Based on the bandwidth allocation weight of each time period, a bandwidth allocation strategy is designed, network resources are allocated according to the weight, and a preliminary bandwidth allocation plan is generated.

[0032] Preferably, the steps of optimizing the preliminary bandwidth allocation scheme according to usage adjustment to give priority to meeting the needs of video conferencing and online education are:

[0033] Analyze the preliminary bandwidth allocation plan, identify the bandwidth consumption patterns of video conferencing and online education, and obtain bandwidth usage data of high-demand services;

[0034] Adjusting bandwidth allocation based on the bandwidth usage data of the high-demand service to obtain an adjusted and optimized bandwidth allocation plan;

[0035] Based on the adjusted and optimized bandwidth allocation scheme, a simulation test is performed to verify the effect of the new strategy.

[0036] Preferably, the steps of obtaining the network performance results are:

[0037] Collect network delay and packet loss rate data in real time to obtain real-time network monitoring data;

[0038] Based on the real-time network monitoring data, the real-time network monitoring data is sorted and analyzed, the overall performance of the network is evaluated, and a comprehensive network performance analysis result is obtained.

[0039] Preferably, the step of obtaining the network performance results also includes: based on the comprehensive network performance analysis results, identifying performance bottlenecks and optimization points, formulating improvement measures, and continuously tracking the improvement effects to obtain network performance results.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are:

[0041] The present invention is a method for analyzing the behavior data of home network devices based on long short-term memory networks, which improves the prediction accuracy of the periodicity and dependence of user behavior, making the allocation of bandwidth resources more forward-looking and dynamic. Specifically, it can automatically adjust resources according to the predicted data demand, especially during peak demand periods, to ensure that key applications such as video conferencing and online education have priority access to necessary network resources. This traffic management technology improves the overall operating efficiency of the network by optimizing bandwidth utilization and reducing network congestion. In addition, continuous monitoring of network performance and analysis of data can timely adjust and optimize network settings to keep network performance in the best state, thereby extending equipment life and reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of the steps of the present invention;

[0043] Figure 2 A schematic diagram of the steps for obtaining formatted device behavior data in the present invention;

[0044] Figure 3 The present invention is a schematic diagram of the steps of optimizing the preliminary bandwidth allocation scheme according to usage adjustment to give priority to meeting the needs of video conferencing and online education. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0046] See also Figure 1 The present invention provides a technical solution, a smart router traffic management method for home scenarios, comprising the following steps:

[0047] Home routers collect network activity data from devices, including data traffic size, connection frequency, and device type, clean and format them, and generate formatted device behavior data.

[0048] Based on the formatted device behavior data, the long short-term memory network is used to process the time series data, learn the periodicity and time dependence of user behavior, and generate user behavior patterns. Based on the user behavior patterns, the network bandwidth demand in each future time period is analyzed to generate bandwidth demand prediction results.

[0049] Based on the bandwidth demand prediction results, a bandwidth allocation strategy is designed to generate a preliminary bandwidth allocation plan; the preliminary bandwidth allocation plan is optimized according to usage adjustments to give priority to meeting the needs of video conferencing and online education.

[0050] Continuously monitor the network environment, including latency and packet loss rate indicators, to obtain network performance results.

[0051] See also Figure 2 , the steps to obtain formatted device behavior data are:

[0052] The router scans all connected devices in the home network and records the network activity of each device, including the data flow size, connection frequency and device type of each device, forming a preliminary network activity record;

[0053] Based on the preliminary network activity records, perform data verification and time stamping, remove invalid and erroneous data records, and reorganize and classify the data according to device type and connection frequency to generate organized network activity data;

[0054] Based on the sorted network activity data, the data is standardized and arranged according to the device identification number and time sequence to obtain formatted device behavior data.

[0055] Specifically, the router scans all devices in the home network and captures the network activity of each device in real time, including information on data traffic size, connection frequency, and device type. This information serves as the basic data for device identification and network behavior analysis. It continuously tracks the devices and records their network usage, including the active time points and traffic peaks of each device. At the same time, it captures and classifies data packets, marks the data source and timestamp, and forms a preliminary record of network activity.

[0056] Based on the preliminary network activity records, data verification is performed, including checking the integrity and format correctness of the data, removing invalid data caused by network noise or transmission errors, and time-stamping the data to ensure that each data record has an exact time reference. The data is classified according to device type and connection frequency. This classification helps distinguish different types of devices and usage patterns, such as the device behavior distinction between regular browsing and large data downloads. Furthermore, the data is reorganized according to these classification results into a more orderly structure to generate sorted network activity data.

[0057] Based on the sorted network activity data, various types of data are standardized according to device identification numbers and time sequence, and the data are converted into a unified format to obtain formatted device behavior data.

[0058] The steps to obtain user behavior patterns are:

[0059] Use the long short-term memory network model to identify the time series features in the formatted device behavior data, including the periodicity and time dependency of user behavior, and obtain the time series data after feature extraction;

[0060] Based on the time series data after feature extraction, the user behavior prediction score at each time point is calculated using the following formula:

[0061] S t =α·h t-1 +β·x t

[0062] Among them, S t is the user behavior score at time step t, h t-1 is the hidden state score of the previous time step, x t is the input feature of the current time step, α and β are adjustment coefficients;

[0063] Based on the user behavior scores, high-score intervals are marked as high-frequency activity time periods to divide and identify user behavior patterns.

[0064] Specifically, a long short-term memory network model is used to identify time series features in formatted device behavior data. The hourly visit data in the past 30 days are recorded in a time sequence in an array. The time series is formed by continuously extracting the visit data of each period within 24 hours. The visit difference of these periods is calculated one by one and stored in a difference set. Then, the average value M1 of the visit data in the same period of the past 30 days is calculated. The difference between the visit data of each period and M1 constitutes a sequence D. The cycle threshold T0 is determined according to the mean and variance of D. The time period marks exceeding T0 are recorded in a list. The time differences between adjacent time periods in the list are sorted to obtain the cycle length distribution, from which the cycle length with an appearance ratio exceeding 10% is selected as the main cycle length. The data of these time periods are then matched one by one with the previous main cycle period and the deviation value sequence is calculated. The time dependence threshold T1 is determined according to the distribution range of the deviation sequence. The corresponding time points of the deviation exceeding T1 are marked and recorded, and finally a set of time series data containing cycle features and time dependence features is obtained.

[0065] The benefit of the formula is that it organically combines the hidden state score of the previous time step with the input features of the current time step through the weight coefficients α and β, thereby fully reflecting the joint effect of historical state and real-time input data when calculating the user behavior prediction score.

[0066] The steps for obtaining the α parameter are as follows: normalize the state of each time point in the user's historical behavior data within 30 consecutive days, use the relationship between the recorded state value and the target score, and find the α that can minimize the prediction error through least squares fitting. After parameter retrieval and comparison, it is confirmed that α = 0.62; the steps for obtaining the β parameter are as follows: select multiple sets of candidate β values ​​from the same data set, substitute them into the calculation process and compare the size of the prediction deviation, and select the β that can effectively reduce the prediction error. After analysis, it is determined that β = 0.38; h t-1 The steps to obtain the parameters are as follows: extract the user behavior state value from the hidden layer output record at time step t-1, normalize and compare the activity frequency data of the same period in the past 7 days, and obtain h t-1 =0.75; x t The steps to obtain the parameters are as follows: read the device connection frequency data of the period from the record file of the router at the current time step t, calculate the average value of the data of the same period in the last 24 hours and normalize it to obtain x t =0.58.

[0067] Calculation process:

[0068] Substitute the parameters into the formula:

[0069] S t =0.62 0.75 + 0.38 0.58;

[0070] First, calculate 0.62 0.75 = 0.465;

[0071] Then calculate 0.38 0.58 = 0.2204;

[0072] Add together to get S t =0.465+0.2204=0.6854.

[0073] The result shows that the user behavior prediction score at the current time step is 0.6854. When the value exceeds 0.68, it can be considered that the user's behavior in this time period is relatively active. This result value will provide a data basis for the subsequent identification of high-frequency activity time periods.

[0074] Based on user behavior scores, the score data of each time point in the past 90 days are recorded in an array by time period. After sorting all the scores, the median m2 is extracted. The difference between the score of each time point and m2 is calculated and stored in the list W. The upper quartile Q3 is obtained from the numerical distribution of W. Q3 is used as the high score threshold Z. The time point numbers exceeding Z are recorded in the index table. Then, the adjacent numbers in the index table are continuously combined to form several high score intervals. The high-frequency activity time period is confirmed by counting the time intervals between adjacent high-score points, and the start and end time point numbers of these high-frequency activity time periods are recorded in the result file.

[0075] The steps to obtain the bandwidth demand prediction results are as follows:

[0076] Extract features from user behavior patterns, including peak activity periods and persistent behavior trends, to obtain key behavior feature data;

[0077] Based on the key behavior feature data, the bandwidth demand in each time period is calculated using the following formula:

[0078]

[0079] Among them, D t represents the predicted bandwidth demand in time period t, a i represents the influence coefficient of the ith behavior on the bandwidth demand, b it is the predicted activity of the ith behavior in time period t, c i is the normalized coefficient of the ith behavior, and N is the total number of behavior types;

[0080] Based on the predicted bandwidth demand, the data is integrated and the bandwidth peak and valley values ​​are analyzed to obtain the bandwidth demand prediction results.

[0081] Specifically, the initial step of extracting features from user behavior patterns is to retrieve hourly connection records for 30 consecutive days from the recorded user access logs, and to sort the access record numbers in a sequence with hourly granularity. The time distribution of the corresponding behavior is calculated based on the device connection frequency and data exchange volume in each hour, and the time distribution data is counted and a counting table is created. In the counting table, the records of each behavior type are arranged and several hour segments with the highest count values ​​are found. The count values ​​of these hour segments are compared with the average count values ​​of all hour segments. If the number of hour segments with count values ​​significantly higher than the average exceeds the time segment threshold calculated by referring to multiple groups of historical data (the time segment threshold is a fixed value obtained by adding two times the standard deviation to the average value of the statistical results of multiple days), such hour segments are marked and recorded in a list. The behavior distribution with a high level of connection values ​​for multiple consecutive hour segments is selected from the list as the activity peak period. The acquisition of continuous behavior trends is obtained by comparing the number of occurrences of similar behaviors in multiple adjacent hour segments. When the range of quantity changes is lower than the critical value calculated by summarizing the variance of each day's data and taking a weighted average, these relatively stable behaviors are recorded in the list, and finally the key behavior feature data is obtained.

[0082] The formula is beneficial in that it performs weighted normalization calculations on the impact coefficients of different types of behaviors and the amount of activity, which helps balance the bandwidth impact of different types of behaviors and makes the prediction results closer to the actual demand level;

[0083] The steps for obtaining the N parameter are to classify and count the behavior types through the user access log. When the records confirm the presence of four typical behaviors: video streaming, file downloading, online games, and web browsing, N is set to 4.

[0084] a i The steps for obtaining parameters are: monitoring user access data, performing linear regression analysis on bandwidth consumption and overall demand of each behavior type, selecting the value that minimizes the prediction deviation from historical data, and confirming a1=0.85, a2=0.65, a3=0.55, a4=0.40 through parameter retrieval;

[0085] b it The steps to obtain the parameters are to query the activity records of the current time period t, count the activity volume (Mbps) of each behavior type in this time period from the multi-day trend data, and obtain b through weighted processing. 1t =5Mbps,b 2t =6Mbps,b 3t =4Mbps,b 4t =2Mbps;

[0086] c iThe steps of obtaining parameters are as follows: summarize the basic bandwidth consumption of each behavior type in long-term monitoring, select quantitative values ​​that can make the results more stable from the reference day statistics through data analysis, and determine c1=3, c2=

[0087] 2, c3 = 1, c4 = 2;

[0088] Calculation process:

[0089]

[0090] Calculate item by item:

[0091]

[0092] Sum:

[0093] 1.12890625+1.69+1.21+0.0711111=4.10001735

[0094] Take the square root:

[0095]

[0096] The result shows that the predicted bandwidth demand for the current period is about 2.0248 Mbps. When the value is greater than 2.0 Mbps, it indicates that the current superposition of multiple types of behaviors has a high bandwidth demand, and when the value is lower than 1.0 Mbps, it indicates that the bandwidth demand is low. If there are consecutive periods exceeding 3.0 Mbps in the statistics, corresponding adjustments can be made in the resource configuration. This value is closely related to the subsequent bandwidth allocation steps.

[0097] Based on the predicted bandwidth demand, the bandwidth values ​​corresponding to each time period are selected from the data recorded in the previous stage, and these bandwidth values ​​are loaded into a list in chronological order. Statistical processing is performed on the values ​​of each time period in the list, and the global average and standard deviation are calculated. The global average plus one standard deviation is used as a reference standard for judging the peak value. When the bandwidth value is greater than this reference standard, the time period number is recorded in the peak index table. When the peak index table is tested, isolated single abnormal values ​​are screened out, and the remaining continuous multiple peak periods are marked in the peak list. When the bandwidth value is lower than the global average minus one standard deviation, the time period number is marked in the valley list. The frequency of occurrence and duration are compared one by one according to the peak list and the valley list, and the data of the comparison results are summarized in the integrated table. Then, the peak and valley distribution information is read item by item from the integrated table, and the relevant statistical data are extracted sequentially and written into the result file to finally obtain the bandwidth demand prediction result.

[0098] The steps to obtain the preliminary bandwidth allocation plan are:

[0099] Analyze the bandwidth demand forecast results, determine the bandwidth demand peak and valley values ​​in each time period, and obtain demand peak and valley data;

[0100] Based on the demand peak and valley data, the bandwidth allocation weight for each time period is calculated using the following formula:

[0101]

[0102] Among them, W t Assign a weight to the bandwidth in time period t, R max is the predicted maximum bandwidth requirement, R t is the bandwidth requirement for time period t, ∈ is a constant to prevent the denominator from being zero;

[0103] Based on the bandwidth allocation weight of each time period, a bandwidth allocation strategy is designed, network resources are allocated according to the weight, and a preliminary bandwidth allocation plan is generated.

[0104] Specifically, the bandwidth demand forecast results are analyzed, and the bandwidth forecast data for each hour in 7 consecutive days are numbered, sorted and stored in a data sequence. The values ​​of each time period are selected from the data sequence, added and the global average value M1 and the global standard deviation σ are calculated. In order to set the peak threshold, M1 plus one times σ is used as the peak judgment standard in the detection process. The σ is obtained by calculating the variance of the bandwidth data of all time periods and performing a square root operation on the result. The peak judgment standard is compared with the values ​​of each time period. When the bandwidth value of the time period is greater than M1 plus σ, the time period number is marked in the peak index. The marked peak period is checked to filter out discontinuous single-point peaks. After elimination, the continuous peak period list is recorded in the peak result set. The valley threshold is set by M1 minus σ. When the bandwidth value of the time period is lower than M1 minus σ, the time period number is marked in the valley list. The same processing steps are performed on the valley list to ensure the continuity of the valley segment. The number information of the peak list and the valley list is compared and recorded in parallel. The processing results are written into the data file to obtain the demand peak and valley data.

[0105] The benefit of the formula is that by proportionalizing the predicted maximum bandwidth demand with the bandwidth demand of the current time period, the bandwidth allocation weights of each time period are more balanced;

[0106] R max The parameter acquisition step is to collect statistics on bandwidth usage data recorded every hour for 30 consecutive days, and select the maximum bandwidth demand value as R max ,In this example, the maximum value of the historical data detected is 15.0Mbps;

[0107] R tThe parameter acquisition step is to select the predicted bandwidth demand value of the current time period t from the prediction results obtained in the previous stage, and the query record is 5.5Mbps;

[0108] The steps for obtaining the ∈ parameter are as follows: It is defined according to the minimum bandwidth measurement accuracy. To prevent the denominator from being zero, the minimum step value is 0.01Mbps;

[0109] Calculation process:

[0110] Substituting the parameters into the formula:

[0111]

[0112] calculate:

[0113]

[0114] The result shows that the bandwidth allocation weight in the current time period t is approximately 2.7236. When the weight value is greater than 2, it means that the relative demand in this time period is low (because the weight decreases when the denominator is larger, and the weight increases when the denominator is smaller). When it is less than 1, it means that the demand in this time period is relatively high. By calculating the weights of different time periods, a series of weight values ​​can be obtained for the subsequent formulation of bandwidth resource allocation plans.

[0115] Based on the bandwidth allocation weight of each time period, the bandwidth allocation weight of each time period within 24 consecutive hours is recorded in the weight array in strips, and each bandwidth allocation weight is read from the weight array in time sequence and accumulated to obtain the weight sum W_sum. The recorded total available bandwidth B_total = 100Mbps is selected from the data file, and the allocable bandwidth B_t of each time period is calculated one by one by the method of B_t = (bandwidth allocation weight / W_sum)*B_total. In the processing process, the abnormal time period is identified by comparing the recorded historical average bandwidth of each time period with the current allocated bandwidth. When the deviation value exceeds the threshold value obtained by previous statistics (the threshold value is determined by the average value of multi-day data deviation plus one standard deviation), the current W_t is readjusted, and the adjusted bandwidth allocation weight is re-submitted into the B_t formula to calculate the allocated bandwidth. After completing the bandwidth allocation for all time periods, the result is stored in the file to generate a preliminary bandwidth allocation plan.

[0116] See also Figure 3 , the steps to adjust and optimize the initial bandwidth allocation plan based on usage, giving priority to meeting the needs of video conferencing and online education are:

[0117] Analyze preliminary bandwidth allocation plans, identify bandwidth consumption patterns for video conferencing and online education, and obtain bandwidth usage data for high-demand services;

[0118] Adjust bandwidth allocation based on bandwidth usage data of high-demand services to obtain an adjusted and optimized bandwidth allocation plan;

[0119] Based on the adjusted and optimized bandwidth allocation scheme, a simulation test is conducted to verify the effectiveness of the new strategy.

[0120] Specifically, when analyzing bandwidth allocation plans, special attention is paid to bandwidth usage data for video conferencing and online education services, and bandwidth usage during key time periods is recorded, including service start and end times, number of participating users, and peak bandwidth consumption. For each service, the average bandwidth usage and maximum bandwidth demand for each time period are calculated, and these data are compared with bandwidth supply capacity to determine the time periods when bandwidth demand approaches or exceeds 70% of the current supply. This threshold is determined based on the recommendations of network operators and historical data analysis as a key indicator that bandwidth allocation may need to be optimized.

[0121] The bandwidth usage data collected for high-demand services is used to adjust the bandwidth allocation plan. First, the bandwidth priority of each service is set, with video conferencing and online education being given the highest priority. Services are then sorted according to their average bandwidth demand during critical periods. The bandwidth allocation ratio for each period is then adjusted based on the peak bandwidth demand of the service to ensure that critical services can obtain at least 120% of their average bandwidth demand as a safety buffer during peak hours. This safety buffer ratio is determined based on previous service satisfaction feedback and network stability results. Through multiple adjustments and comparisons of the allocation effects of each period, an adjusted and optimized bandwidth allocation plan is formed.

[0122] The actual effect test of the adjusted and optimized bandwidth allocation plan was carried out. The network traffic during normal and peak hours was simulated, and the network usage mode was set for each service, such as the large-volume data transmission of video conferencing and the high-frequency interaction needs of online education. The response time, packet loss rate and user connection stability of the service under various simulation conditions were monitored. Special attention was paid to the performance of the service in the security threshold test, that is, when the bandwidth usage reached 90%. The actual effect of the bandwidth allocation strategy was verified through these test data, and it was confirmed that the new strategy can meet the performance standards of all services at the highest demand. The test results support the formal implementation of the adjusted bandwidth allocation strategy.

[0123] The steps to obtain network performance results are:

[0124] Collect network delay and packet loss rate data in real time to obtain real-time network monitoring data;

[0125] Based on real-time network monitoring data, organize and analyze the real-time network monitoring data, evaluate the overall performance of the network, and obtain comprehensive network performance analysis results;

[0126] Based on the comprehensive network performance analysis results, identify performance bottlenecks and optimization points, formulate improvement measures, and continuously track the improvement effects to obtain network performance results.

[0127] Specifically, monitoring devices deployed at network nodes continuously collect delay and packet loss information for each data packet, record the average delay and packet loss rate per minute, ensure that emergencies and persistent problems can be tracked, and update these monitoring data to the central database in real time. The data is timestamped and source recorded, enabling network administrators to respond quickly when abnormal indicators are observed. For example, if the delay exceeds the safety threshold of 100 milliseconds or the packet loss rate is higher than the critical point of 0.1%, an alarm is triggered immediately. These thresholds are set based on the network's design standards and past performance history, ensuring that the network operates within an acceptable performance range, thereby maintaining service continuity and quality.

[0128] The collected real-time network monitoring data is used to evaluate the overall health and performance of the network, especially the data during peak periods. For example, the changing trends of latency and packet loss rate in the same time period are analyzed. At the same time, the time periods that deviate from normal performance are identified by calculating the hourly average and standard deviation. These analyses help identify potential network bottlenecks, such as router configuration errors or insufficient bandwidth, and then recommend network optimization measures, such as increasing the bandwidth of specific links or optimizing network routing strategies.

[0129] After comprehensively analyzing the results of network performance, we identified and located performance bottlenecks, such as the long response time of a specific server or frequent packet loss on a certain link. For these problems, we recorded and analyzed the specific reasons for the performance degradation and formulated corresponding improvement measures, such as replacing hardware or adjusting network traffic management strategies. After implementing these measures, we continued to monitor their effects and compared the network performance data before and after the adjustment, such as the numerical changes in latency and packet loss rate, to ensure that the improvement measures can effectively solve the problems.

[0130] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A smart router traffic management method for home scenarios, characterized in that: The following steps are involved: The home router collects the network activity data of the device, including the data flow size, connection frequency and device type, cleans and formats it, and generates formatted device behavior data; Based on the formatted device behavior data, a long short-term memory network is used to process time series data, learn the periodicity and time dependency of user behavior, and generate a user behavior pattern; Based on the user behavior pattern, analyzing the network bandwidth demand in each future time period and generating a bandwidth demand prediction result; Based on the bandwidth demand prediction result, design a bandwidth allocation strategy and generate a preliminary bandwidth allocation plan; Optimizing the preliminary bandwidth allocation plan based on usage adjustments, giving priority to meeting the needs of video conferencing and online education; Continuously monitor the network environment, including latency and packet loss rate indicators, to obtain network performance results.

2. The intelligent router traffic management method for home scenarios according to claim 1 is characterized in that: The steps for obtaining the formatted device behavior data are as follows: The router scans all connected devices in the home network and records the network activity of each device, including the data flow size, connection frequency and device type of each device, forming a preliminary network activity record; Based on the preliminary network activity records, perform data verification and time stamping, remove invalid and erroneous data records, and reorganize and classify the data according to device type and connection frequency to generate organized network activity data; Based on the sorted network activity data, the data is standardized and arranged according to device identification numbers and time sequence to obtain formatted device behavior data.

3. The intelligent router traffic management method for home scenarios according to claim 1 is characterized in that: The steps of acquiring the user behavior pattern are: Using a long short-term memory network model, identifying time series features in the formatted device behavior data, including periodicity and time dependency of user behavior, to obtain time series data after feature extraction; Based on the time series data after feature extraction, the user behavior prediction score at each time point is calculated using the following formula: S t =a·h t-1 +β·x t Among them, S t is the user behavior score at time step t, h t-1 is the hidden state score of the previous time step, x t is the input feature of the current time step, α and β are adjustment coefficients; Based on the user behavior scores, high-score intervals are marked as high-frequency activity time periods, and user behavior patterns are divided and identified.

4. The intelligent router traffic management method for home scenarios according to claim 1 is characterized in that: The steps for obtaining the bandwidth demand prediction result are: Extracting features from the user behavior patterns, including peak activity periods and persistent behavior trends, to obtain key behavior feature data; Based on the key behavior feature data, the bandwidth demand in each time period is calculated using the following formula: Among them, D t represents the predicted bandwidth demand in time period t, a i represents the influence coefficient of the ith behavior on the bandwidth demand, b it is the predicted activity of the ith behavior in time period t, c i is the normalized coefficient of the ith behavior, and N is the total number of behavior types; Based on the predicted bandwidth demand, the data is integrated and the bandwidth peak and valley values ​​are analyzed to obtain a bandwidth demand prediction result.

5. The intelligent router traffic management method for home scenarios according to claim 1 is characterized in that: The steps for obtaining the preliminary bandwidth allocation plan are: Analyze the bandwidth demand forecast results, determine the bandwidth demand peak and valley values ​​in each time period, and obtain demand peak and valley data; Based on the demand peak and valley data, the bandwidth allocation weight for each time period is calculated using the following formula: Among them, W t Assign a weight to the bandwidth in time period t, R max is the predicted maximum bandwidth requirement, R t is the bandwidth requirement for time period t, ∈ is a constant to prevent the denominator from being zero; Based on the bandwidth allocation weight of each time period, a bandwidth allocation strategy is designed, network resources are allocated according to the weight, and a preliminary bandwidth allocation plan is generated.

6. The intelligent router traffic management method for home scenarios according to claim 1, characterized in that: The steps for optimizing the preliminary bandwidth allocation scheme based on usage adjustment to prioritize the needs of video conferencing and online education are as follows: Analyze the preliminary bandwidth allocation plan, identify the bandwidth consumption patterns of video conferencing and online education, and obtain bandwidth usage data of high-demand services; Adjusting bandwidth allocation based on the bandwidth usage data of the high-demand service to obtain an adjusted and optimized bandwidth allocation plan; Based on the adjusted and optimized bandwidth allocation scheme, a simulation test is performed to verify the effect of the new strategy.

7. The intelligent router traffic management method for home scenarios according to claim 1, characterized in that: The steps for obtaining the network performance results are: Collect network delay and packet loss rate data in real time to obtain real-time network monitoring data; Based on the real-time network monitoring data, the real-time network monitoring data is sorted and analyzed, the overall performance of the network is evaluated, and a comprehensive network performance analysis result is obtained.

8. The intelligent router traffic management method for home scenarios according to claim 7, characterized in that: The step of obtaining the network performance results also includes: based on the comprehensive network performance analysis results, identifying performance bottlenecks and optimization points, formulating improvement measures, and continuously tracking the improvement effects to obtain network performance results.

Images