5G intelligent optimization digital platform and method based on tidal effect

Abstract

The invention relates to the technical field of mobile communication, and discloses a 5G intelligent optimization digital platform and method based on a tidal effect. The method comprises the steps of fine-grained crowd movement trajectory analysis, regional crowd activity-network flow correlation modeling, time-varying network topology feature extraction, topology elasticity scoring and resource optimization allocation, and adaptive network topology reconstruction execution. The method comprises the following steps: analyzing anonymized user position data to identify a crowd movement mode, establishing a correlation model of regional activity and network traffic, extracting network topology characteristics to identify a traffic spatial-temporal change rule, calculating a topology elastic score to realize optimal resource allocation, executing adaptive network topology reconstruction to ensure that a network structure pre-adapts to traffic change, and finally, performing adaptive network topology reconstruction to ensure that the network structure pre-adapts to traffic change. The tidal effect in the 5G network is intelligently sensed and predicted, the network topology is adjusted in advance to adapt to the flow change, the network resource utilization efficiency is remarkably improved, the network congestion probability is reduced, the user experience is improved, and the technical advantage and the application value are obvious.

Description

Technical Field

[0001] The present invention relates to the field of mobile communication technology, and more specifically, to a 5G intelligent optimization digital platform and method based on tidal effect. Background Art

[0002] With the widespread deployment and application of 5G networks, network traffic has shown significant tidal effect characteristics. The so-called tidal effect refers to the phenomenon that network traffic shows periodic fluctuations in space and time due to regular changes in people's activities (such as commuting, large-scale events, etc.). This tidal effect poses a huge challenge to network resource allocation: static network configuration is difficult to cope with dynamically changing traffic demands, resulting in an unbalanced state where some areas have excess resources and other areas have shortages.

[0003] Currently, the technology used to solve the tidal effect problem of 5G networks mainly relies on post-response mechanisms, that is, adjusting resources after network congestion occurs. This passive optimization method leads to a decline in user experience and low network efficiency. Although some studies have begun to focus on predictive network optimization, most methods still have the following shortcomings: The correlation analysis between crowd activities and network traffic is coarse-grained and lacks the perception of the movement patterns of micro-crowds; Network topology optimization is mainly based on historical traffic data and lacks consideration of external social factors; Resource adjustment strategies lack foresight and cannot respond to traffic changes in advance; Most optimization algorithms are static algorithms with limited adaptability and learning capabilities.

[0004] Therefore, existing technologies are difficult to effectively cope with the challenges of tidal effects in 5G networks. An intelligent optimization method is needed that can perceive changes in crowd activities, predict network traffic distribution, adjust network topology in advance to adapt to traffic changes, and achieve efficient use of network resources. Summary of the invention

[0005] The present invention provides a 5G intelligent optimization digital platform and method based on the tidal effect to solve the technical problem in the related technology that the network topology configuration is too static and cannot effectively cope with the rapid changes in network traffic distribution caused by the tidal effect.

[0006] The present invention provides a 5G intelligent optimization method based on tidal effect, comprising the following steps: Fine-grained crowd movement trajectory analysis: Analyze the movement patterns of people in the area based on anonymized user location data and generate a movement trajectory prediction model; Regional crowd activity-network traffic correlation modeling: predict regional crowd activity changes based on multi-source data and generate network traffic prediction results; Time-varying network topology feature extraction: Extract topology features based on the time-varying network graph model and graph embedding algorithm to identify the temporal and spatial variation patterns of traffic distribution; Topology elasticity scoring and resource optimization allocation: Based on the topology scoring system, the priority and resource allocation ratio are calculated and adjusted to achieve forward-looking network resource layout; Adaptive network topology reconstruction execution: Based on the prediction results and scoring system, network topology adjustments are performed in real time to ensure that the network structure is pre-adapted to traffic changes.

[0007] Furthermore, the fine-grained crowd movement trajectory analysis step specifically includes: Collect anonymized user location data at multiple time points in the region, perform data cleaning, remove outliers, and extract spatiotemporal features; Divide the area into grids, count the crowd density of each grid at different times, and construct a regional mobility heat map; Apply spatiotemporal trajectory clustering algorithm to analyze user movement trajectories and identify typical movement patterns and key paths; Based on the trajectory clustering results, a path flow prediction function is established, and the flow of each key path at a specific time point is predicted by combining the time factor.

[0008] Furthermore, the regional crowd activity-network traffic association modeling step specifically includes: Obtain POI information, traffic data, and event calendar data in the area, and perform data processing and integration; Based on multi-source data, a regional activity feature vector is constructed, which contains information on activity intensity, activity type, and number of participants at a specific time point; Build a dual-mode topology pre-adaptive system including normal mode and event mode, and define event impact factors to trigger mode switching; Based on historical traffic data and regional activity characteristics, the traffic-activity association model is trained to output the predicted network traffic distribution.

[0009] Furthermore, the time-varying network topology feature extraction step specifically includes: Based on historical traffic data, a time-varying network graph model is constructed to represent the dynamic characteristics of network topology changing over time; Apply the time-series graph embedding algorithm to convert the time-varying network graph into a low-dimensional feature representation, optimizing the reconstruction error and temporal consistency; Based on the graph embedding results, the typical spatiotemporal patterns of traffic distribution are analyzed to extract periodic patterns and abnormal patterns. Identify the critical link set in the network, establish a link load timing prediction model, and generate a future network topology load distribution prediction map.

[0010] Furthermore, the topology elasticity scoring and resource optimization allocation steps specifically include: Establish a network topology resilience scoring system to evaluate the adaptability of network topology to traffic changes; Based on the network topology and traffic prediction results, identify the critical link set and redundant path set; Calculate the adjustment priority score of each link in the topology based on the predicted traffic changes; Based on the link adjustment priority score, the optimal allocation ratio of network resources is calculated to solve the constrained optimization problem.

[0011] Furthermore, the adaptive network topology reconstruction execution step specifically includes: Build a network resource redistribution control system to convert resource allocation plans into specific resource allocation instructions; Based on resource allocation instructions, the link bandwidth is dynamically adjusted, and a smooth transition strategy is adopted to avoid service interruption caused by sudden changes; Based on the link bandwidth adjustment results, the network routing strategy is optimized to find a set of routing paths to optimize the overall network performance; After implementing topology reconstruction, a real-time monitoring and feedback mechanism is established to trigger the adaptive adjustment process based on performance deviations.

[0012] Furthermore, the traffic-activity association model adopts a multi-layer perceptron structure, including: The input layer receives the regional activity feature vector; The hidden layer contains two sub-layers and introduces a temporal attention mechanism to assign different weights to activity features at different time points; The output layer outputs the predicted traffic distribution, and the number of neurons is equal to the number of network area divisions.

[0013] Furthermore, the network topology resilience score is composed of the following indicators: Link utilization balance, which evaluates the balance of load distribution on each link in the network; The number of redundant paths, which evaluates the availability of alternative paths in the network in the event of a link failure; Key node connectivity, evaluates the connection status of key nodes in the network.

[0014] Furthermore, the resource allocation instruction includes the following types: A bandwidth adjustment instruction, used to adjust the bandwidth allocation of the key link identified by the time-varying network topology feature extraction step; Routing update instructions, used to update network routing tables and forwarding rules; Load balancing instructions, used to distribute traffic among multiple paths; Resource reclaim instructions are used to reclaim resources from low-load areas.

[0015] A 5G intelligent optimization digital platform based on tidal effect, including: Data collection module, used to obtain anonymized user location data and regional activity data; Trajectory analysis module, used to analyze crowd movement patterns and generate movement trajectory prediction models; Traffic prediction module, used to establish a correlation model between regional crowd activities and network traffic and predict traffic distribution; Topology analysis module, used to extract network topology features and identify the temporal and spatial variation patterns of traffic distribution; Resource optimization module, used to calculate the topology elasticity score and the optimal allocation of network resources; Topology reconstruction module, used to perform network topology adjustment and conduct real-time monitoring and feedback; The central control module is used to coordinate the work of each functional module and provide a unified management interface.

[0016] The beneficial effects of the present invention are: Accurate network traffic prediction has been achieved: the system is based on fine-grained crowd movement trajectory analysis and regional crowd activity-network traffic correlation modeling to accurately predict future network traffic distribution, with a prediction accuracy rate that is approximately 25% higher than traditional methods.

[0017] Improved network resource utilization efficiency: Through topology elasticity scoring and resource optimization allocation, accurate scheduling and optimized allocation of network resources are achieved, network capacity utilization is increased by 40%, and resource waste and congestion problems are avoided.

[0018] Significantly reduced network congestion: The predictive topology adjustment system can sense and respond to traffic peaks in advance, reducing the probability of congestion in high-load areas by 65%, greatly improving user experience.

[0019] Enhanced network resilience: The adaptive network topology reconstruction execution system can dynamically respond to changes in the network environment, especially for unconventional events such as large-scale activities, and complete network topology adjustments 13 to 15 minutes in advance, significantly enhancing network resilience.

[0020] Improved operation and maintenance efficiency: The system's adaptive adjustment capability changes network maintenance from passive response to active prevention, reducing the frequency of manual intervention, improving network operation and maintenance efficiency by about 35%, and reducing operating costs accordingly. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a 5G intelligent optimization method based on tidal effect of the present invention. DETAILED DESCRIPTION

[0022] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0023] Implementation method 1, a 5G intelligent optimization method based on tidal effect, such as Figure 1 As shown, the following steps are included: Step 1: Fine-grained crowd movement trajectory analysis; This step builds a fine-grained crowd movement trajectory analysis framework, analyzes the movement patterns of people in the area based on anonymized user location data, and generates a movement trajectory prediction model. Specifically, it includes the following steps: Anonymous user location data collection and preprocessing: Collect anonymized user location data at multiple time points in the area:

[0024] in Indicates The location data set of each user includes location coordinates and timestamps. The collected data is preprocessed, including data cleaning, outlier removal, and spatiotemporal feature extraction, to form a standardized location data set. .

[0025] Regional mobile heat map construction: based on preprocessed location data , dividing the area into Grid, for each grid In time Count the crowd density and build a regional mobile heat map ,in Indicates at time Grid By analyzing the changes in the heat map at different times, the main characteristics and patterns of crowd flow in the area can be identified.

[0026] Trajectory clustering algorithm implementation: Apply spatiotemporal trajectory clustering algorithm to analyze user movement trajectory and identify typical movement patterns and key paths. Set trajectory similarity calculation function ,in and Respectively represent and Based on the trajectory similarity, the density clustering algorithm is used to classify similar trajectories into the same category to form a trajectory category set:

[0027] in Indicates A collection of class tracks.

[0028] Path flow prediction model construction: Based on the trajectory clustering results, a path flow prediction function is established: in Indicates the path, represents the prediction time point, Indicates the number of users, Indicates at time user Select Path The prediction function analyzes historical mobile trajectory data and combines time factors (working days / non-working days, peak hours / non-peak hours, etc.) to output the expected traffic flow of each key path at a specific time point.

[0029] In the specific implementation, the path flow prediction function is calculated based on the sequence prediction algorithm, in which the user selects the path probability This is achieved through a conditional probability model. For each path First, we count the frequency of users choosing this route at different time points (such as morning rush hour on weekdays, holidays, etc.) in historical data to form a basic probability; then we combine the feature vector at the current time point: (including time type, weather conditions, surrounding activities and other factors), and obtain the real-time selection probability through weighted calculation: in is the basic probability, is the time type, is the weight coefficient of each feature.

[0030] Step 2: Regional crowd activity-network traffic correlation modeling; This step builds a correlation model between regional crowd activities and network traffic, uses multi-source data to predict regional crowd activity changes, and generates network traffic prediction results. Specifically, it includes the following steps: Multi-source data acquisition and fusion: Acquire multi-source data such as POI (point of interest) information, traffic data, event calendar, etc. in the area, and process and fuse the data. POI information includes:

[0031] in Indicates Information about points of interest, including location, type, capacity and other attributes; Traffic data includes:

[0032] in Indicates Information about transportation routes; The events calendar includes:

[0033] in Indicates The information of a planned activity includes time, location, estimated scale and other attributes.

[0034] Regional activity feature vector construction: Based on multi-source data, regional activity feature vector is constructed ,in Indicates the area, Represents a time point. The feature vector contains information such as the activity intensity, activity type, and number of participants in the area at a specific time point, and is formally expressed as: in Indicates The value of the activity feature dimension.

[0035] Dual-mode topology pre-adaptation system construction: Build a dual-mode topology pre-adaptation system that includes normal mode and event mode. In normal mode, traffic changes are predicted based on historical rules; in event mode, topology adjustments are triggered by event impact factors. Define event impact factors: in Indicates The weight coefficient of each activity, Indicates the scale of the activity, Indicates the duration of the activity. The value exceeds the preset threshold When the system switches from normal mode to event mode.

[0036] Traffic-activity correlation model training: Based on historical traffic data and regional activity characteristics, the traffic-activity correlation model is trained The model adopts a multi-layer perceptron structure, and the input is the regional activity feature vector , the output is the predicted network traffic distribution The model training objective is to minimize the mean square error between the predicted flow and the actual flow: in Indicates area In time The actual network traffic.

[0037] In the specific implementation, the traffic-activity association model adopts a three-layer structure: the input layer receives the regional activity feature vector , contains 32 neurons; the hidden layer contains two sub-layers, with 64 and 32 neurons respectively, using the ReLU activation function; the output layer outputs the predicted traffic distribution , the number of neurons is equal to the number of network area divisions. In order to improve the model's ability to process time series data, a temporal attention mechanism is also introduced in the hidden layer to assign different weights to the activity features at different time points. The calculation formula is: in , , are query, key and value matrices respectively, which are obtained by linear transformation of input features. The dimension of the key.

[0038] Step 3: Extract time-varying network topology features; This step constructs a time-varying network graph model, uses a graph embedding algorithm to extract topological features, identifies the temporal and spatial variation patterns of traffic distribution, and provides a decision-making basis for network topology reconstruction. Specifically, it includes the following steps: Time-varying network graph model construction: Based on historical traffic data, a time-varying network graph model is constructed ,in Represents a collection of network nodes, Indicates time The edge set of Indicates time The edge weight matrix of . Indicates time node and nodes The model can represent the dynamic characteristics of network topology changing over time and capture the spatiotemporal distribution pattern of network traffic under the tidal effect.

[0039] Application of graph embedding algorithm: Applying the time-series graph embedding algorithm to embed the time-varying network graph Convert to low-dimensional feature representation. For each time point , calculate the node embedding matrix ,in Representation Node In time The embedding vector of .

[0040] The optimization goal of the embedding algorithm is to minimize the reconstruction error: While maximizing timing consistency: Traffic spatiotemporal pattern analysis and extraction: Based on the graph embedding results, clustering algorithms are used to analyze the typical spatiotemporal patterns of traffic distribution. The time series is divided into time windows, and calculate the feature matrix for each window ,The dynamic time warping algorithm is applied to calculate the similarity of features in different time windows, and extract the periodic and abnormal patterns.,Finally, the temporal and spatial variation of traffic distribution is obtained,,including peak period distribution and tidal effect characteristics.

[0041] Identification of key links and prediction of change trends: Based on the results of spatiotemporal pattern analysis, identify the key link set in the network:

[0042] These links play a key role in traffic distribution changes. For each key link, a link load timing prediction model is established. , predicting the future Combine the prediction results of step 1 and step 2 to generate a comprehensive prediction model and output the prediction diagram of future network topology load distribution. .

[0043] Step 4: Topology elasticity scoring and resource optimization allocation; This step designs a topology elasticity scoring system, calculates the topology adjustment priority and resource allocation ratio based on the analysis results of the previous steps, and implements forward-looking network resource layout. Specifically, it includes the following steps: Topology resilience scoring system construction: Establish a network topology resilience scoring system and define a topology resilience scoring function , which is used to evaluate the adaptability of network topology to traffic changes. The topology resilience score is composed of multiple indicators, including link utilization balance , Number of redundant paths , key node connectivity etc. The topological elasticity score calculation formula is: in , , is the weight coefficient, and .

[0044] Identification of critical links and redundant paths: Based on the network topology and traffic prediction results, identify the critical link set:

[0045] Redundant path set:

[0046] For each key link , calculate its importance score , which is based on factors such as the traffic carrying capacity of the link, the importance of the connected nodes, and the number of alternative paths; for each redundant path , calculate its availability score ,This score takes into account factors such as path length, current load, and available bandwidth.

[0047] Topology adjustment priority score calculation: Calculate the adjustment priority score of each link in the topology based on the predicted traffic changes. Define the link adjustment priority score function: in represents the predicted future link traffic, Indicates the current link traffic. represents the link capacity, Indicates the importance of the link. The higher the priority score, the more the link needs to be adjusted.

[0048] Resource allocation ratio optimization: Calculate the optimal allocation ratio of network resources based on link adjustment priority scores.

[0049] Define the resource allocation problem as a constrained optimization problem: maximize ,in Indicates the adjusted network topology.

[0050] Constraints include total resource limits , and the capacity requirements of each link .

[0051] The Lagrange multiplier method is used to solve the optimization problem and obtain the optimal resource allocation solution. ,in (i∈[1,n]) represents the The optimal resource allocation for each link.

[0052] Step 5: Adaptive network topology reconstruction is performed.

[0053] This step is based on the prediction results and scoring system of the previous step, and performs real-time network topology adjustment to ensure that the network structure can pre-adapt to the traffic changes caused by the tidal effect. It specifically includes the following steps: Implementation of network resource redistribution control system: Build a network resource redistribution control system that receives the resource allocation plan calculated in step 4 , and convert it into specific resource allocation instructions. Define the resource allocation instruction set:

[0054] Each instruction Contains information such as target link, adjustment type, and adjustment parameters.

[0055] Link bandwidth dynamic adjustment algorithm: Based on resource allocation instructions, dynamic adjustment of link bandwidth is achieved.

[0056] Define bandwidth adjustment function ,in represents the target link, Indicates the bandwidth adjustment amount.

[0057] The bandwidth adjustment process takes into account the current load status of the link and adopts a smooth transition strategy to avoid service interruption caused by sudden changes. The mathematical expression of the adjustment process is: in To smooth the function, ensure the gradualness of bandwidth adjustment.

[0058] Routing strategy optimization algorithm: Optimize network routing strategy based on link bandwidth adjustment results.

[0059] Define the routing strategy optimization problem: Find a set of routing paths under the updated topology structure , making the overall network performance optimal.

[0060] The optimization objectives include minimizing end-to-end delay, maximizing throughput, balancing load, etc. A multi-objective optimization algorithm is used to solve the problem and obtain the optimized routing strategy. .

[0061] Topology reconstruction feedback and adaptive adjustment: After topology reconstruction is implemented, a real-time monitoring and feedback mechanism is established to collect actual network performance data Expected performance Deviation of performance. Define the performance deviation vector , when the deviation exceeds the threshold The adaptive adjustment process is triggered when the adaptive adjustment process is triggered. The adaptive adjustment process is based on the reinforcement learning algorithm, which learns the mapping relationship between the environment state and the optimal adjustment strategy and continuously optimizes the topology reconstruction decision model.

[0062] Through the above sub-steps, the adaptive network topology reconstruction execution system realizes the intelligent allocation of network resources and dynamic optimization of routing strategies. It can accurately capture the characteristics of tidal effects and make topology adjustments in advance, ensuring that network performance remains stable and efficient during traffic changes.

[0063] The technical effects of this implementation are as follows: Accurate network traffic prediction capability: Through fine-grained crowd movement trajectory analysis and regional crowd activity-network traffic correlation modeling, the system can accurately predict future network traffic distribution. The prediction accuracy is about 25% higher than traditional methods, providing a reliable basis for network topology optimization.

[0064] Efficient network resource utilization: Based on topology elasticity scoring and resource optimization allocation, accurate scheduling and optimized allocation of network resources are achieved, network capacity utilization is increased by 40%, and resource waste and congestion problems are avoided.

[0065] Significant congestion reduction effect: Through the predictive topology adjustment system, the system can sense and respond to traffic peaks in advance, reducing the probability of congestion in high-load areas by 65%, greatly improving the user experience.

[0066] Enhanced network resilience: The adaptive network topology reconstruction execution system can dynamically respond to changes in the network environment, especially for unconventional events such as large-scale activities. The system can complete network topology adjustments 13 to 15 minutes in advance, significantly enhancing network resilience.

[0067] Intelligent operation and maintenance efficiency improvement: The system's adaptive adjustment capability enables network maintenance to change from passive response to active prevention, reducing the frequency and degree of manual intervention. The network operation and maintenance efficiency is improved by about 35%, and the operating costs are reduced accordingly.

[0068] An application example of implementation mode 1 is as follows: This method has been put into practical application in the optimization of 5G networks around an international exhibition center. The exhibition center covers an area of ​​120,000 square meters, with 8 main exhibition halls and multiple conference centers. It is an important venue for large-scale exhibitions, international conferences and cultural performances. The 5G network in this area faces the following challenges: Traffic distribution is highly dynamic: different exhibitions and events attract different types of people, and the concentration of traffic changes rapidly with the event schedule, with typical tidal effect characteristics.

[0069] Diversity of traffic demand: data transmission and video conferencing are the main activities during commercial exhibitions; high-definition live broadcasts and short video sharing are the main activities during cultural performances; and online conferencing and data synchronization are the main activities during international conferences. Different activities have different requirements for network service quality.

[0070] Peak forecasting is difficult: In the short period before and after a large-scale event, regional traffic will increase suddenly, and traditional static network topology cannot effectively cope with it, resulting in a decline in user experience.

[0071] This application scenario covers the following network structure: a total of 32 5G base stations and 102 cells are deployed to form a network covering the exhibition center and surrounding areas. During peak hours, the number of users connected simultaneously can reach 150,000, and the network traffic peak can reach 48Gbps. Traditional network optimization methods experience multiple congestions during large-scale events, and the user complaint rate is high, so there is an urgent need to adopt new intelligent optimization methods.

[0072] Deployment process: In response to the actual needs of the convention and exhibition center, we deployed and implemented the system according to the five core steps of implementation method 1.

[0073] Fine-grained crowd movement trajectory analysis: Data collection and preprocessing: 15 data collection points were set up in the exhibition center area to collect anonymous user location data based on WiFi probes and base station location information. To ensure user privacy, hash coding and data desensitization technology were used to process user identification. The collection cycle was 10 seconds per time, and data was collected for 3 months as a training set, including crowd trajectory information during different types of activities.

[0074] Construction of mobile heat map of the exhibition center area: The exhibition center area is divided into 50×50 grid units, each covering about 50 square meters. Based on the collected location data, a heat map of crowd density in different time periods is constructed. An example of the heat map is shown in Table 1: Table 1: Example of heat map of crowd density in the exhibition center at different time periods (partial data) Trajectory clustering and key path identification: Based on the collected data, 8 main pedestrian flow paths are identified, including the path from the subway station to each exhibition hall, the connection path between exhibition halls, and the path from the exhibition hall to the dining area. An identifier (P1-P8) is set for each path, and the path selection probability at different time periods is analyzed. The key path flow statistics are shown in Table 2: Table 2: Traffic statistics of key paths at different time periods (weekdays)

[0075] Application of path traffic prediction model: The path traffic prediction model is trained based on historical data. The model comprehensively considers time factors, activity types, and historical traffic patterns. The model adopts a tree structure. It first determines the type of the current time period (weekday / weekend, peak / off-peak), then further determines whether there are special activities, and finally predicts the traffic of each path based on historical data under corresponding conditions. Especially for large exhibitions, a mapping relationship between exhibition types and crowd flow distribution is established. For example, the crowd flow distribution difference between the International Consumer Electronics Show and the Medical Equipment Show reaches 35%.

[0076] Regional crowd activity-network traffic correlation modeling: Multi-source data acquisition and fusion: Acquire the exhibition schedule, ticketing data, booth distribution map and surrounding traffic information of the convention and exhibition center to establish a multidimensional database. At the same time, extract network traffic data during different activities from the historical network monitoring system to establish an activity-traffic mapping relationship. The data example is shown in Table 3: Table 3: Example of activity and network traffic correlation data

[0077] Activity feature vector construction: construct a feature vector for each type of activity, including information such as time, scale, type, and crowd attributes. For example, the feature vector of a technology exhibition is: [1 (working day), 3 (duration), 35,000 (scale), 0.7 (proportion of business people), 0.2 (proportion of technical personnel), 0.1 (others)].

[0078] Dual-mode topology pre-adaptive system deployment: A dual-mode system is deployed on the network management platform. In normal mode, traffic prediction and fine-tuning are performed at 15-minute intervals; in event mode, more frequent predictions (5 minutes / time) and larger adjustments are made based on event impact factors. The system sets the event impact factor threshold to 0.65 and automatically switches to event mode when a large event (such as the opening of an exhibition with more than 15,000 participants) is predicted.

[0079] Traffic-activity association model training and validation: A neural network model was trained based on historical data. The input layer contains 32 nodes (corresponding to activity features), two hidden layers (64 nodes and 32 nodes), and the output layer is 102 nodes (corresponding to traffic predictions for each cell). After 5,000 rounds of iterative training, the model achieved a prediction accuracy of 88.5% on the validation set. The model can identify network traffic characteristics of different exhibition types. For example, the data download volume of medical exhibitions is 35% higher than that of consumer electronics exhibitions, while the real-time streaming transmission of consumer electronics exhibitions is 52% higher than that of medical exhibitions.

[0080] Time-varying network topology feature extraction: Construction of time-varying network graph model: Based on network topology information and historical traffic data, a time-varying network graph model of the exhibition center area is constructed. The model contains 32 nodes (base stations) and 86 edges (links), and the edge weights (link loads) are updated every 15 minutes.

[0081] Graph embedding and feature extraction: Apply the time-series graph neural network algorithm to convert the network topology graph into a low-dimensional feature representation (dimension is 16). Based on the extracted features, four typical network status modes are identified: ordinary working day mode, exhibition opening day mode, exhibition regular day mode and large-scale cultural and artistic event mode.

[0082] Traffic spatiotemporal pattern analysis: Analyze the traffic distribution during different activities within three months and extract the spatiotemporal variation patterns. It was found that the traffic growth rate at the south entrance reached 300% from 30 minutes before the opening to 60 minutes after the opening; and the traffic growth rate in the exit area and transportation hub direction reached 250% from 30 minutes before the end of the exhibition to 45 minutes after the end of the exhibition. Based on these patterns, targeted pre-adjustment strategies were designed.

[0083] Identification of key links and prediction of change trends: 12 key links were identified, which bear the main traffic changes in the tidal effect. A prediction model was established for each key link, with an average prediction lead time of 18 minutes and an accuracy of 83.7%. An example of the key link load prediction results is shown in Table 4: Table 4: Example of key link load prediction results (opening day of the technology exhibition)

[0084] Topology elasticity scoring and resource optimization allocation: Implementation of topology resilience scoring system: Establish a topology resilience scoring system, with scoring indicators including link utilization balance (weight 0.4), redundant path availability (weight 0.3) and key node connectivity (weight 0.3). The current network topology is scored, with a normal score of 0.72. If no adjustment is made during large-scale events, the score will drop to 0.48, which is lower than the safety threshold of 0.65 set by the system.

[0085] Calculation of key links and resource adjustment plans: For the 12 key links, calculate the resource adjustment priorities and ratios. Some examples of priority sorting results are shown in Table 5: Table 5: Priority ranking of key link adjustments (before the opening of the technology exhibition)

[0086] Resource allocation optimization calculation: Calculate the optimal resource allocation plan based on link adjustment priority and network resource constraints. The total resource constraint is 120 units, which need to be allocated among 32 base stations. The optimization goal is to maximize the network resilience score while ensuring that all predicted loads can be met. The final calculation results show that by dynamically adjusting 30% of the resources in the low-load area to the high-load area, the network resilience score can be maintained above 0.78, and good performance can be maintained even during the peak period of the exhibition.

[0087] Adaptive network topology reconstruction execution: Deployment of network resource redistribution control system: Deploy a resource redistribution control system on the core network management platform. The system can receive the optimized resource allocation plan and convert it into device-level configuration instructions. The system supports three types of operations: bandwidth adjustment, route optimization, and load balancing, with the minimum adjustment granularity of 5% resource units.

[0088] Dynamic bandwidth adjustment implementation: Dynamic bandwidth adjustment is implemented for key links. The adjustment adopts a smooth transition strategy and is completed gradually in a 5-minute cycle to avoid interruption of current services. In the test on the opening day of the science and technology exhibition, the system successfully increased the bandwidth of the three key links at the southern entrance by 40%, and reduced the bandwidth of the five links in the low-load area by 15% for resource recovery.

[0089] Routing strategy optimization execution: Based on the predicted traffic distribution, the network routing strategy is optimized. The system calculates 20 backup paths and sets the trigger threshold to 85% link load. In actual operation, when the L3 link load reaches 83%, the system triggers the routing adjustment in advance, diverting about 25% of the traffic to the backup path, effectively avoiding congestion.

[0090] Feedback mechanism and adaptive adjustment: A real-time monitoring system was established to collect actual network performance data every 5 minutes and compare it with expected performance. When the deviation exceeds 15%, the adaptive adjustment process is triggered. In the first month of the system operation, the number of times the adaptive adjustment was triggered was 12 times, and in the third month it dropped to 3 times, indicating that the system's predictive ability is continuously improving.

[0091] Application effect verification: After this implementation was actually deployed in the convention and exhibition center for 6 months, its application effect was evaluated through multiple indicators, verifying the technical advantages of the 5G intelligent optimization digital platform and method based on the tidal effect.

[0092] Network performance improvement effect: Before and after the system deployment, the core network performance indicators were compared for exhibition activities of the same scale and type. The results are shown in Table 6: Table 6: Comparison of network performance before and after system deployment (opening day of a large technology exhibition)

[0093] As can be seen from Table 6, after the system was deployed, the network performance was significantly improved. In particular, the number of network congestion occurrences decreased by 88.9%, and the video service freeze rate decreased by 77.3%. These indicators directly reflect the improvement in user experience. At the same time, the average link utilization rate increased by 47.7%, indicating that the efficiency of network resource utilization has been significantly improved.

[0094] Verification of prediction accuracy and warning time: In order to verify the prediction accuracy and early warning capability of the system, during the 6-month operation, the prediction results and actual traffic data of different types of activities were collected, and the prediction accuracy and average early warning lead time were statistically analyzed, as shown in Table 7: Table 7: Statistics of prediction accuracy and warning time for different activity types

[0095] As can be seen from Table 7, the system has the highest prediction accuracy for regular exhibitions, reaching 92.3%, and also has a high prediction accuracy for the opening of large exhibitions and concerts. Even for emergencies, the prediction accuracy reached 76.8%, and the average warning lead time was 13.2 minutes, which is sufficient for network adjustment. The system's comprehensive average prediction accuracy was 88.5%, the topology adjustment hit rate was 86.3%, and the average warning lead time was 20.2 minutes, which verified the system's prediction ability and warning effect.

[0096] Resource utilization efficiency analysis: The changes in resource utilization efficiency before and after system deployment are shown in Table 8: Table 8: Comparison of network resource utilization distribution (unit: number of base stations)

[0097] As can be seen from Table 8, before the system was deployed, resource utilization was polarized, with some base stations having utilization rates below 20%, and three base stations being overloaded on peak days. After the system was deployed, resource utilization distribution became more reasonable, with most base stations having utilization rates between 40% and 80%, which not only avoided resource waste but also eliminated overload. This result shows that the system can effectively realize dynamic allocation of resources and improve overall resource utilization efficiency.

[0098] User experience improvement verification: Through user satisfaction surveys and network operation data analysis, the system's improvement effect on user experience was evaluated. 1,000 user satisfaction surveys were conducted before and after deployment, and the results are shown in Table 9: Table 9: User satisfaction survey results before and after system deployment

[0099] As can be seen from Table 9, user satisfaction has significantly improved after the system was deployed, especially in terms of network congestion perception, which has increased by 57.7%. Overall satisfaction has increased from 3.0 points to 4.3 points, an increase of 43.3%, which shows that the system has significantly improved user experience.

[0100] In addition, during the six months of the system's operation, the number of network-related complaints decreased from an average of 68 per month to 12 per month, a decrease of 82.4%. Especially during large exhibitions, the number of related complaints decreased from an average of 35 per exhibition to 5 per exhibition, a decrease of 85.7%.

[0101] System long-term stability verification: In order to verify the long-term stability and adaptive ability of the system, the monthly change data of the key indicators of the system were collected during the 6-month operation period, as shown in Table 10: Table 10: Monthly changes in key system indicators

[0102] As can be seen from Table 10, as the system operation time increases, the prediction accuracy increases from 83.5% to 92.3%, the system response time decreases from 8.5 seconds to 5.0 seconds, and the number of adaptive adjustments decreases from 32 times / month to 8 times / month, which shows that the system's self-learning ability continues to improve and its adaptability to the network environment is getting stronger and stronger. At the same time, the number of system failures decreases month by month, and zero failure operation has been achieved in the 5th and 6th months, proving the long-term stability of the system.

[0103] Through the above multi-dimensional effect verification, it is fully proved that the technical feasibility and superiority of the 5G intelligent optimization digital platform and method based on the tidal effect provided by this embodiment in practical applications. This method can effectively deal with the problem of rapid changes in network traffic distribution caused by the tidal effect, significantly improve network performance and user experience, and has important application value.

[0104] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are protected by this embodiment.

Claims

1. A 5G intelligent optimization method based on tidal effect, characterized in that: The following steps are involved: Fine-grained crowd movement trajectory analysis: Analyze the movement patterns of people in the area based on anonymized user location data and generate a movement trajectory prediction model; Regional crowd activity-network traffic correlation modeling: predict regional crowd activity changes based on multi-source data and generate network traffic prediction results; Time-varying network topology feature extraction: Extract topology features based on the time-varying network graph model and graph embedding algorithm to identify the temporal and spatial variation patterns of traffic distribution; Topology elasticity scoring and resource optimization allocation: Based on the topology scoring system, the priority and resource allocation ratio are calculated and adjusted to achieve forward-looking network resource layout; Adaptive network topology reconstruction execution: Based on the prediction results and scoring system, network topology adjustments are performed in real time to ensure that the network structure is pre-adapted to traffic changes.

2. According to a 5G intelligent optimization method based on tidal effect according to claim 1, it is characterized in that: The fine-grained crowd movement trajectory analysis step specifically includes: Collect anonymized user location data at multiple time points in the region, perform data cleaning, remove outliers, and extract spatiotemporal features; Divide the area into grids, count the crowd density of each grid at different times, and construct a regional mobility heat map; Apply spatiotemporal trajectory clustering algorithm to analyze user movement trajectories and identify typical movement patterns and key paths; Based on the trajectory clustering results, a path flow prediction function is established, and the flow of each key path at a specific time point is predicted by combining the time factor.

3. According to a 5G intelligent optimization method based on tidal effect according to claim 2, it is characterized in that: The regional crowd activity-network traffic association modeling steps specifically include: Obtain POI information, traffic data, and event calendar data in the area, and perform data processing and integration; Based on multi-source data, a regional activity feature vector is constructed, which contains information on activity intensity, activity type, and number of participants at a specific time point; Build a dual-mode topology pre-adaptive system including normal mode and event mode, and define event impact factors to trigger mode switching; Based on historical traffic data and regional activity characteristics, the traffic-activity association model is trained to output the predicted network traffic distribution.

4. According to claim 3, a 5G intelligent optimization method based on tidal effect is characterized in that: The time-varying network topology feature extraction step specifically includes: Based on historical traffic data, a time-varying network graph model is constructed to represent the dynamic characteristics of network topology changing over time; Apply the time-series graph embedding algorithm to convert the time-varying network graph into a low-dimensional feature representation, optimizing the reconstruction error and temporal consistency; Based on the graph embedding results, the typical spatiotemporal patterns of traffic distribution are analyzed to extract periodic patterns and abnormal patterns. Identify the critical link set in the network, establish a link load timing prediction model, and generate a future network topology load distribution prediction map.

5. According to claim 4, a 5G intelligent optimization method based on tidal effect is characterized in that: The topology elasticity scoring and resource optimization allocation steps specifically include: Establish a network topology resilience scoring system to evaluate the adaptability of network topology to traffic changes; Based on the network topology and traffic prediction results, identify the critical link set and redundant path set; Calculate the adjustment priority score of each link in the topology based on the predicted traffic changes; Based on the link adjustment priority score, the optimal allocation ratio of network resources is calculated to solve the constrained optimization problem.

6. The 5G intelligent optimization method based on tidal effect according to claim 5 is characterized in that: The steps of executing the adaptive network topology reconstruction specifically include: Build a network resource redistribution control system to convert resource allocation plans into specific resource allocation instructions; Based on resource allocation instructions, the link bandwidth is dynamically adjusted, and a smooth transition strategy is adopted to avoid service interruption caused by sudden changes; Based on the link bandwidth adjustment results, the network routing strategy is optimized to find a set of routing paths to optimize the overall network performance; After implementing topology reconstruction, a real-time monitoring and feedback mechanism is established to trigger the adaptive adjustment process based on performance deviations.

7. The 5G intelligent optimization method based on tidal effect according to claim 6 is characterized in that: The traffic-activity association model adopts a multi-layer perceptron structure, including: The input layer receives the regional activity feature vector; The hidden layer contains two sub-layers and introduces a temporal attention mechanism to assign different weights to activity features at different time points; The output layer outputs the predicted traffic distribution, and the number of neurons is equal to the number of network area divisions.

8. The 5G intelligent optimization method based on tidal effect according to claim 7 is characterized in that: The network topology resilience score is composed of the following indicators: Link utilization balance, which evaluates the balance of load distribution on each link in the network; The number of redundant paths, which evaluates the availability of alternative paths in the network in the event of a link failure; Key node connectivity, evaluates the connection status of key nodes in the network.

9. The 5G intelligent optimization method based on tidal effect according to claim 8, characterized in that: The resource allocation instructions include the following types: A bandwidth adjustment instruction, used to adjust the bandwidth allocation of the key link identified by the time-varying network topology feature extraction step; Routing update instructions, used to update network routing tables and forwarding rules; Load balancing instructions, used to distribute traffic among multiple paths; Resource reclaim instructions are used to reclaim resources from low-load areas.

10. A 5G intelligent optimization digital platform based on tidal effect, which is used to execute a 5G intelligent optimization method based on tidal effect according to any one of claims 1 to 9, comprising: Data collection module, used to obtain anonymized user location data and regional activity data; Trajectory analysis module, used to analyze crowd movement patterns and generate movement trajectory prediction models; Traffic prediction module, used to establish a correlation model between regional crowd activities and network traffic and predict traffic distribution; Topology analysis module, used to extract network topology features and identify the temporal and spatial variation patterns of traffic distribution; Resource optimization module, used to calculate the topology elasticity score and the optimal allocation of network resources; Topology reconstruction module, used to perform network topology adjustment and conduct real-time monitoring and feedback; The central control module is used to coordinate the work of each functional module and provide a unified management interface.

Images