Network use behavior analysis and bandwidth allocation method and system based on portable WIFI device

Abstract

The invention provides a network use behavior analysis and bandwidth allocation method and system based on a portable WIFI device, and the method comprises the steps: carrying out the grouping of user behaviors through employing a clustering method in machine learning according to a preliminary behavior feature set, carrying out the discrimination of use modes for different purposes, such as entertainment, work, learning, and the like, and carrying out the recognition of the user behaviors. Determining core category distribution behind the user behavior diversity; according to the resource pre-allocation instruction, obtaining an available state of a current network resource, dynamically adjusting a bandwidth allocation proportion in combination with predicted demand peak data, and determining specific execution parameters of an intelligent allocation mechanism; and obtaining real-time feedback data of the network equipment in the execution process through the adjusted resource allocation scheme, and judging whether the dynamic response speed meets the requirement of improving the service quality or not according to the fluctuation condition of the resource allocation efficiency to obtain a final optimization record.

Description

Technical Field

[0001] The present invention relates to the technical field of network resource scheduling, and in particular to a method and system for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device. Background Art

[0002] In today's information age, mobile network devices have become an essential component of people's daily lives and work. As a bridge connecting users to the digital world, performance optimization and resource management of portable wireless network devices are crucial. Given the increasing diversity of user needs and complexity of scenarios, efficiently allocating network resources to meet the demands of diverse scenarios has become a crucial area of focus for industry development. However, most current solutions still have significant shortcomings in network resource management. Many devices use a uniform bandwidth allocation method, lacking accurate identification of specific user behaviors and needs. This results in wasted network resources or insufficient support for user needs during peak hours or in specific scenarios.

[0003] This extensive management approach is unable to adapt to the dynamic changes in user behavior and fails to provide a personalized service experience. A deeper analysis of the challenges in this area reveals that the core issues lie in the accurate classification of user behavior and the dynamic prediction of network resources. First, due to the diverse nature of user network behaviors, such as entertainment, work, or study, their demand patterns for network resources vary greatly. Existing technologies struggle to accurately distinguish the characteristics underlying these behaviors, making it impossible to optimize resource allocation in a targeted manner. This lack of identification directly leads to a deeper challenge: devices are unable to detect and prepare resources in advance, such as pre-allocating bandwidth or pre-establishing connections, before peak user demand arrives, significantly impacting the user experience. These two interrelated issues: the former is the foundation of identification, and the latter is the dynamic response based on identification. Together, they constitute a core obstacle to optimizing resource management.

[0004] Therefore, how to achieve intelligent allocation and advance preparation of network resources through refined classification of user network usage behavior combined with accurate prediction of future demand peaks has become a key issue in improving the service quality of portable wireless network devices. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device, which mainly includes: By collecting historical user usage data from portable wireless network devices, we construct a multi-dimensional feature dataset of user behavior based on information such as network traffic, usage duration, and application type in different time periods and scenarios. This allows us to capture the specific manifestations of user behavior diversity and obtain a preliminary set of behavioral features. Based on a preliminary set of behavioral features, we use clustering methods from machine learning to group user behaviors, differentiate usage patterns for different purposes, such as entertainment, work, and study, and identify the core category distribution behind the diversity of user behaviors. Based on the core category distribution results, we obtain the network resource demand characteristics corresponding to each type of user behavior. Combined with the records of traffic peaks and troughs in historical data, we determine the dynamic changes in demand and derive resource demand models for each type of behavior in different scenarios. By using the resource demand model, we establish a demand peak forecasting analysis framework based on the traffic change trend of each user type within a specific time period. If the traffic demand of a certain user type within a certain time period is predicted to exceed the preset threshold, the corresponding resource pre-allocation instructions are generated. According to the resource pre-allocation instructions, the current network resource availability status is obtained, and combined with the predicted peak demand data, the bandwidth allocation ratio is dynamically adjusted to determine the specific execution parameters of the intelligent allocation mechanism; Based on the execution parameters of the intelligent allocation mechanism, real-time connection load data from network devices is obtained. If a deviation is detected between the actual demand of a certain type of user behavior and the predicted value, the advance preparation strategy correction process is triggered to obtain an adjusted resource allocation plan; Through the adjusted resource allocation plan, we can obtain real-time feedback data from network devices during the execution process. Based on the fluctuation of resource allocation efficiency, we can determine whether the dynamic response speed meets the requirements for improving service quality and obtain the final optimization record. Based on the final optimization records, we obtain specific scenarios where network resource waste or resource shortage occurs during the resource allocation process. Based on the behavioral feature extraction results in these scenarios, we update the multi-dimensional feature dataset of user behavior and determine the input basis for the next round of optimization.

[0006] The present invention provides a network usage behavior analysis and bandwidth allocation system based on a portable WIFI device, which mainly includes: The data collection module is used to collect historical usage data of users on portable wireless network devices, and build a multi-dimensional feature data set of user behavior based on information such as network traffic, usage duration, and application type in different time periods and scenarios, thereby obtaining the specific manifestations of user behavior diversity and obtaining a preliminary set of behavioral features; The behavior clustering module is used to group user behaviors based on a preliminary set of behavioral features using clustering methods in machine learning. This module differentiates usage patterns for different purposes, such as entertainment, work, and learning, and identifies the core category distribution behind the diversity of user behaviors. The demand modeling module is used to obtain the network resource demand characteristics corresponding to each type of user behavior based on the core category distribution results. Combined with the records of traffic peaks and troughs in historical data, it determines the dynamic changes in demand and obtains the resource demand model for each type of behavior in different scenarios. The prediction and analysis module is used to establish a demand peak forecast analysis framework in advance based on the traffic change trend of each type of user behavior within a specific time period through the resource demand model. If the traffic demand of a certain type of user within a certain time period is predicted to exceed the preset threshold, the corresponding resource pre-allocation instruction is generated; The resource pre-allocation module is used to obtain the current available status of network resources based on the resource pre-allocation instructions, dynamically adjust the bandwidth allocation ratio based on the predicted peak demand data, and determine the specific execution parameters of the intelligent allocation mechanism; The dynamic adjustment module is used to obtain real-time connection load data from network devices based on the execution parameters of the intelligent allocation mechanism. If a deviation is detected between the actual demand of a certain type of user behavior and the predicted value, the advance preparation strategy correction process is triggered to obtain an adjusted resource allocation plan; The performance evaluation module is used to obtain real-time feedback data from network devices during the execution process based on the adjusted resource allocation plan. Based on the fluctuation of resource allocation efficiency, it determines whether the dynamic response speed meets the requirements for improving service quality and obtains the final optimization record. The dataset update module is used to obtain specific scenarios of network resource waste or resource shortage during the resource allocation process based on the final optimization records, extract behavioral features in these scenarios, update the multi-dimensional feature dataset of user behavior, and determine the input basis for the next round of optimization.

[0007] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention collects historical user usage data, constructs a multi-dimensional feature data set, uses machine learning clustering methods to classify usage patterns for different purposes, obtains network resource demand characteristics for various behaviors, predicts peak traffic based on the demand model, generates resource pre-allocation instructions, dynamically adjusts bandwidth allocation, triggers a correction process through real-time monitoring of the deviation between actual demand and predicted values, optimizes resource allocation plans, analyzes resource allocation efficiency, determines whether the dynamic response speed meets service quality requirements, and updates the feature data set based on the optimization records to provide a basis for the next round of optimization. The present invention realizes the intelligent allocation of wireless network resources, improves resource utilization efficiency, effectively avoids resource waste or shortage, and thus significantly improves the user's network service experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0009] Figure 1 The present invention is a flowchart of a method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device.

[0010] Figure 2 Schematic diagram of a network usage behavior analysis and bandwidth allocation method based on a portable WIFI device according to the present invention.

[0011] Figure 3 The figure is a structural diagram of a network usage behavior analysis and bandwidth allocation system based on a portable WIFI device according to the present invention. DETAILED DESCRIPTION

[0012] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present application. Rather, they are merely examples of methods and systems consistent with certain aspects of the present application, as detailed in the appended claims.

[0013] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0014] The following describes in detail the specific implementation methods, features and effects of the present invention in conjunction with the accompanying drawings and preferred embodiments.

[0015] Example 1 See also Figure 1-Figure 2 This embodiment provides a method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device, which may specifically include: Step S101 collects historical usage data of users in portable wireless network devices, constructs a multi-dimensional feature dataset of user behavior based on information such as network traffic, usage duration, and application type in different time periods and scenarios, obtains specific manifestations of user behavior diversity, and obtains a preliminary behavioral feature set.

[0016] By collecting historical usage data from portable wireless network devices, we can obtain users' network traffic, usage duration, and application types in different time periods and scenarios, and build an original data set. Data preprocessing technology is used to clean the original data set, remove missing values and outliers, and obtain a standardized data set; Based on the standardized data set, the statistical characteristics of network traffic, usage duration and application type are extracted to construct a multi-dimensional user behavior feature data set; If the feature dimensions in the multi-dimensional user behavior feature dataset meet the preset threshold, the feature is reduced in dimension using the principal component analysis algorithm to obtain a reduced-dimensional behavior feature set. Based on the reduced dimensionality behavioral feature set, the K-means clustering algorithm is used to classify user behaviors and obtain subsets of different behavioral patterns. By analyzing subsets of different behavior patterns, we extract the behavioral diversity of each subset in different time periods and scenarios, and obtain the user behavior diversity characteristics. If the user behavior diversity features match the preset behavior pattern template, a user behavior feature set is generated based on the matching results.

[0017] For example, by collecting historical usage data of users in portable wireless network devices, building a multi-dimensional feature data set and analyzing the diversity of user behaviors, the following technical implementation methods can be adopted; Suppose we extract the network usage data of 1,000 users over a 30-day period from the logs of a portable Wi-Fi device, including timestamp, traffic (MB), usage duration (minutes), and application type (video, social, game, etc.); First, use Python scripts combined with SQL queries to extract data from the device database and filter out hourly traffic data. For example, user A's average traffic between 8:00 AM and 9:00 AM on weekdays is 50 MB, with social applications accounting for 60% of the traffic. During data cleaning, we use the Z-score algorithm to remove outliers. We set the threshold to 3 and remove records where the traffic or duration deviates from the mean by 3 standard deviations. For example, a user with a single-hour traffic of 5000MB will be removed. Next, we constructed a multi-dimensional feature dataset and used feature engineering methods to extract features such as time (weekdays / weekends, morning / evening), scene (home / public places, inferred by Wi-Fi SSID), and application type ratio. This generated a matrix containing 20 features. For example, user B's video traffic at home on weekend evenings accounts for 80%, with a duration of 120 minutes. We used the K-means clustering algorithm (K=5) to classify user behavior. We calculated the similarity between feature vectors based on Euclidean distance and obtained five types of behavior patterns. For example, a "heavy video user" has an average daily traffic of 200MB, with video accounting for 70% of the total traffic. To analyze behavioral diversity, we calculated the characteristic entropy of each user using the formula H = -Σ(p_i * log(p_i)), where p_i represents the proportion of application types. The entropy value ranges from 0 to 2.5. A higher entropy indicates greater behavioral diversity. For example, user C has an entropy value of 2.1, indicating balanced application usage. Finally, through principal component analysis (PCA) dimensionality reduction, three principal components with 80% variance were retained to generate a preliminary set of behavioral features, including traffic peaks, duration distribution, and application preferences, forming a user behavior profile. The above process is implemented through automated scripts, and data processing and analysis are completed on cloud servers to ensure efficiency and scalability.

[0018] In step S102 , based on the preliminary set of behavioral features, a clustering method in machine learning is used to group user behaviors, differentiate usage patterns for different purposes such as entertainment, work, and study, and determine the core category distribution behind the diversity of user behaviors.

[0019] By analyzing the behavioral characteristics in the preliminary set, the user behaviors are grouped using clustering methods to obtain the initial category distribution; Based on the initial category distribution, corresponding usage pattern features are extracted for entertainment mode, work mode, and learning mode, and the behavioral feature subsets under each mode are determined; If the distribution density of the behavioral feature subset in a certain mode is lower than the preset threshold, the user behavior in that mode is clustered again to obtain more refined core categories; By dividing core categories into time periods and analyzing the behavior behind each category in different time periods, we can obtain the time-related usage pattern distribution; Based on the time-related usage pattern distribution and the specific scenarios of entertainment mode, work mode, and learning mode, we can determine the changing trends of the behavioral characteristics of each category in different scenarios. If the trend of behavioral characteristics changes significantly in a certain scenario, the data in that scenario will be weighted and adjusted through the information processing link to determine the final category distribution result; By performing a multi-dimensional comparison of the final category distribution results, we can obtain the comprehensive performance of user behavior in each category and determine the deep pattern associations behind the behavior.

[0020] For example, we extracted network usage logs of 2,000 users from portable Wi-Fi devices over a period of 60 days, including timestamps, uplink / downlink traffic (MB), connection device type (mobile phone / tablet / computer), and network environment (4G / 5G / Wi-Fi); Use Python combined with NoSQL databases (such as MongoDB) to query and extract daily traffic and device usage data by time period (every 4 hours). For example, the average downlink traffic of user D using his mobile phone on the 5G network from 12:00 PM to 4:00 PM on weekdays is 80 MB. During data preprocessing, the IQR (interquartile range) method was used to eliminate outliers. The upper and lower limits were set to Q1-1.5*IQR and Q3+1.5*IQR. Records with traffic exceeding 3000MB in a single period were removed. For example, user E's traffic of 3500MB in a certain period was removed. Next, we constructed a multidimensional dataset through feature engineering. We extracted features including daily average traffic (uplink / downlink), device type preference (mobile phone share, e.g., 60%), network type distribution (5G share, e.g., 45%), and usage time concentration (based on temporal entropy, using the formula H_t=-Σ(q_i*log(q_i)), where q_i is the time period and the entropy ranges from 0 to 1.8, with higher entropy indicating more dispersed time periods). Taking user F as an example, his average daily downlink traffic is 150MB, of which mobile phones account for 70% and 5G networks account for 50%. The time entropy is 1.5; We used the DBSCAN clustering algorithm (eps=0.5, min_samples=10) to group user behaviors and calculated feature vector distances based on cosine similarity, identifying four patterns. For example, "high-efficiency users" primarily use computers, have a high proportion of 5G, and have an average daily traffic of 100MB. To differentiate between usage purposes such as entertainment, work, and study, we combined application layer protocol analysis (such as the proportion of HTTP / HTTPS traffic) to calculate the traffic distribution of each user type in specific scenarios. For example, HTTPS traffic from work users between 8:00 AM and 6:00 PM on weekdays accounts for 65%. Using the t-SNE algorithm to reduce dimensionality, we retained two-dimensional features to visualize the distribution of user groups and generate a core set of categories including traffic patterns, device preferences, and network types. All processes are automatically executed in the cloud through distributed computing frameworks (such as Apache Spark), ensuring efficient and scalable data processing.

[0021] In step S103, based on the results of the core category distribution, the network resource demand characteristics corresponding to each type of user behavior are obtained. Combined with the records of traffic peaks and troughs in the historical data, the law of dynamic changes in demand is determined to obtain the resource demand model for each type of behavior in different scenarios.

[0022] For core categories and distribution results, we use data integration to obtain network resource usage records for each type of user behavior from historical traffic. By combining peak and trough records, we determine the resource consumption range for each type of behavior in different time periods. Based on the resource consumption range, the information processing link is used to analyze demand characteristics and demand changes, and the resource demand fluctuations of each type of user behavior under the time pattern are determined. If the fluctuation of a category exceeds the preset threshold, it is marked with a priority to obtain a set of marked categories; For the labeled category set, we divide the scenarios by differentiating them to obtain the network resource allocation for each type of user behavior in different scenarios. Combined with the data distribution in historical traffic, we determine the resource demand priority for each type of behavior in a specific scenario. Based on resource demand priorities, a data mapping method is used to analyze the correlation between demand changes and scenario differences. If the demand change trend in a certain scenario is inconsistent with historical traffic records, the data for that scenario is corrected through a weighted adjustment process to obtain an adjusted demand distribution. Based on the adjusted demand distribution, we analyze temporal patterns to obtain the foundation for resource model construction for each type of user behavior in different time periods. We then combine peak and trough records to determine the applicability of resource models in different time periods. Based on the applicability of the resource model, the K-means clustering method is used to group network resources. The matching degree between each group of resources and user behavior is evaluated to obtain the final resource allocation plan. For the final resource allocation plan, through a comprehensive analysis of scenario differences and time patterns, we obtain the resource model optimization direction for each type of user behavior in different scenarios and time periods, and determine the optimized resource allocation strategy.

[0023] For example, based on the results of the core category distribution, the system first automatically extracts key indicators by analyzing the network resource demand characteristics of each type of user behavior and combining them with the traffic peak and trough records in the historical data; For example, historical data for a certain type of "nighttime entertainment user" shows that traffic peaks at an average of 200 MB / hour between 10:00 PM and 2:00 AM, while traffic peaks at a low of only 20 MB / hour between 6:00 AM and 10:00 AM. The system uses time series analysis algorithms (such as the ARIMA model with parameters p=2, d=1, and q=1) to predict traffic fluctuations and calculates a peak-to-valley ratio of 10:1, indicating that bandwidth demand for this type of user surges at night. Next, the system automatically builds demand models based on the dynamic changes in different scenarios. It analyzes the differences between "night entertainment users" on weekends and weekdays and finds that the peak traffic volume on weekend nights increases to 250MB per hour, a 25% increase. Through regression analysis (linear regression, R 2 =0.85) confirmed that the positive correlation between increased bandwidth demand and online video traffic during weekends reached 0.78; To further refine the model, the system incorporates environmental variables such as the impact of holidays. It automatically calculates that peak traffic for this type of user during holidays increases to 280MB per hour. It also adjusts resource allocation priorities based on historical data on network latency sensitivity (60% of users have a latency tolerance below 50ms). Finally, the system incorporates all categories (such as "morning study users" with peak traffic of 80MB per hour and trough traffic of 10MB per hour) into a unified resource demand model, uses a weighted average algorithm (weights are based on user proportions, such as entertainment users accounting for 40%) to generate a comprehensive forecast, and automatically outputs bandwidth allocation plans for each scenario to ensure that the dynamic resource adjustment logic is rigorous and adaptable.

[0024] In step S104, a resource demand model is used to pre-establish an analysis framework for demand peak prediction based on the traffic change trend of each type of user behavior within a specific time period. If it is predicted that the traffic demand of a certain type of user within a certain time period exceeds a preset threshold, a corresponding resource pre-allocation instruction is generated.

[0025] For each type of user behavior, we obtain traffic change records within a specific time period from historical data. Combined with the characteristics of the change trend, we build an initial traffic demand analysis basis and obtain the traffic change pattern for each type of behavior. Based on traffic change patterns, a pre-built peak forecast analysis framework is used to estimate traffic demand within a specific time period, determine traffic change trends for each type of user behavior in future time periods, and identify potential peak time distributions. For peak time distribution, if the traffic demand forecast value of a certain type of user behavior exceeds the preset threshold, the resource demand of this type of behavior is prioritized through the information processing link to obtain a sorted resource demand list; Obtain a preliminary resource pre-allocation plan based on the sorted resource demand list. Analyze the feasibility of resource pre-allocation based on traffic trends and peak forecast results, and determine the adjusted allocation strategy framework. Based on the adjusted allocation strategy framework, information integration tools are used to evaluate the matching degree between traffic demand and resource pre-allocation. If the matching degree does not meet the preset standard, the allocation strategy is optimized through the data correction process to obtain the optimized resource allocation plan; Generate corresponding resource pre-allocation instructions based on the optimized resource allocation plan. Combined with the traffic change pattern within a specific time period, the allocation instructions are sent to the relevant resource management modules through the instruction distribution mechanism to complete the resource pre-configuration process. For the resource pre-allocation process, obtain comparison data between actual traffic demand and predicted results, analyze the dynamic adaptability of resource allocation through continuous monitoring tools, judge the execution status of resource pre-allocation instructions, and determine the direction of subsequent adjustments.

[0026] For example, the system constructs a resource demand model, conducts in-depth analysis of traffic change trends of various user behaviors within a specific time period, automatically generates an analysis framework for demand peak forecasting, and formulates resource pre-allocation instructions based on this.

[0027] For example, for the "lunchtime office users" category, the system first collected traffic data from 11:00 AM to 1:00 PM daily over the past three months and found that the average traffic peak was 150 MB per hour, with some days surging to 180 MB per hour due to online meeting demands. The system uses a time series decomposition algorithm to break down traffic data into three components: trend, seasonality, and residuals. The calculated standard deviation of traffic fluctuation during the lunch period is 15.5MB per hour, and the predicted peak value for the next week is 195MB per hour. The system then sets a preset threshold of 170MB per hour, and automatically triggers the resource pre-allocation mechanism when the predicted traffic exceeds this threshold; Next, the system used a cluster analysis algorithm (K-means, K=3) to categorize user behavior into three subgroups: high, medium, and low demand. Analysis of historical data for the high-demand subgroup (approximately 25%) showed a correlation coefficient of 0.72 between peak traffic and cloud storage access frequency. Based on this, the system calculated that an additional 30% of bandwidth reserve would be required, equivalent to approximately 45MB of incremental resources per hour. At the same time, the system automatically analyzes the reliance of this type of user on network stability based on business relevance and finds that their tolerance for interruptions is low (the complaint rate for interruptions lasting more than 2 minutes reaches 35%). Therefore, the system binds resource pre-allocation instructions to the backup server activation mechanism to ensure that at least 10% of the redundant capacity in the resource pool is immediately available. Finally, the system integrates the prediction results with resource pre-allocation instructions to generate an automated scheduling script. If it is predicted that traffic demand will exceed the threshold within the next 24 hours, a bandwidth increase instruction will be sent to the core node four hours in advance to ensure the foresight and stability of resource allocation.

[0028] Step S105 , according to the resource pre-allocation instruction, obtain the current available status of network resources, combine the predicted demand peak data, dynamically adjust the bandwidth allocation ratio, and determine the specific execution parameters of the intelligent allocation mechanism.

[0029] By combining the available bandwidth status data obtained from the network resource management module with the peak demand distribution information, the current network resource load is analyzed to obtain a preliminary assessment result of bandwidth allocation; Based on the preliminary assessment results of bandwidth allocation, a pre-established allocation optimization model is used to perform matching calculations based on peak demand distribution information to determine the adjustment direction of bandwidth allocation ratios. If the adjustment direction of the bandwidth allocation ratio does not match the current network resource load, the status data will be checked again through the information processing link to determine whether the allocation ratio needs to be corrected; The configuration requirements of the intelligent allocation mechanism are obtained by adjusting the allocation ratio, and the initial value range of the execution parameters is determined based on the peak demand distribution information. Based on the initial value range of the execution parameters, the logistic regression model is used to optimize and adjust the parameter values to obtain the final execution parameters of the intelligent allocation mechanism; If the final execution parameters do not reach the preset thresholds during the verification of the information processing phase, the parameter values are fine-tuned through the data integration tool to determine the adjusted execution parameters; According to the adjusted execution parameters, the corresponding bandwidth allocation instructions are generated by the instruction management module and sent to the network resource management module to complete the dynamic allocation process of bandwidth resources.

[0030] For example, the system first obtains the current available status of network resources through a real-time monitoring tool, scans the bandwidth capacity of the core nodes, and finds that the current total available bandwidth is 2.5 GB per hour, of which 1.2 GB per hour is occupied, leaving 1.3 GB per hour of available resources; Next, the system calls the prediction module to obtain peak demand data for the "evening entertainment users" category over the next 12 hours. Based on historical data analysis, the average traffic for this category from 20:00 to 22:00 is 320 MB / hour. Using an ARIMA model (with parameters p=2, d=1, q=1) and combining it with autoregressive analysis of traffic fluctuations over the past month, the system concludes that the peak traffic over the next 12 hours could reach 380 MB / hour, with a standard error of 22.3 MB / hour. The system then dynamically adjusts the bandwidth allocation ratio, initially allocating 29% of the remaining bandwidth, or approximately 377MB per hour, based on the ratio of the predicted peak to the currently available bandwidth (380MB / 1300MB ≈ 0.29). To optimize allocation, the system introduces a linear programming algorithm, setting the objective function to minimize bandwidth waste. The constraints include: allocated bandwidth must not fall below 90% of the predicted peak (342MB per hour) and must not exceed 80% of the remaining bandwidth (1040MB per hour). The calculation results show that the optimal bandwidth allocation is 360MB per hour, with the remaining 17MB per hour as a buffer; The system then analyzed the reliance of "evening entertainment users" on video streaming and found a correlation coefficient of 0.65 between their traffic and the frequency of video definition switching. This inferred that high-definition demand accounted for approximately 40% of the users' usage. Therefore, an additional 10% of bandwidth (36MB per hour) was allocated to support 4K streaming. To ensure stability, the system has solidified the parameters of the intelligent allocation mechanism into an automated script. When the predicted traffic approaches the threshold (350MB per hour), a bandwidth adjustment instruction is sent to the edge node two hours in advance, and the backup bandwidth pool is activated to maintain 5% redundant capacity (18MB per hour). Finally, the system generates execution parameters that include 360MB per hour of primary allocation, 36MB per hour of incremental allocation, and 18MB per hour of redundancy to ensure maximum resource utilization.

[0031] In step S106, based on the execution parameters of the intelligent allocation mechanism, real-time connection load data in the network device is obtained. If a deviation is detected between the actual demand of a certain type of user behavior and the predicted value, the correction process of the advance preparation strategy is triggered to obtain an adjusted resource allocation plan.

[0032] Collect real-time data of connected loads through network devices, and use data processing modules to clean and format the collected data to obtain structured load information; If the actual demand for a certain type of user behavior in the structured load information deviates from the predicted value by more than a preset threshold, a dynamic correction process is activated through a trigger mechanism to determine the correction requirement; Based on the correction requirements, the support vector machine model is used to classify and analyze the actual needs and predicted deviations of user behavior to obtain the direction of resource adjustment; Based on the resource adjustment direction, obtain the pre-established configuration template and generate a preliminary resource configuration plan based on the load information; If the preliminary resource allocation plan does not meet the preset threshold in the verification of the data processing module, the data integration tool is used to fine-tune the plan to obtain an optimized allocation plan; Generate corresponding resource allocation instructions based on the optimized configuration plan and send them to network devices to complete dynamic adjustment of resource configuration; Based on the adjusted resource configuration, new connection load data is collected in real time, and the applicability of the allocation plan is cyclically verified to obtain continuously optimized configuration results.

[0033] For example, the system collects real-time connection load data of core routers and switches through the built-in network monitoring module and finds that the actual traffic of the current "online education user" category is 180MB per hour, while the predicted value is 250MB per hour, with a deviation rate of up to 28%; The system then automatically triggers the deviation detection algorithm. Based on the weighted moving average method of historical traffic data (weight parameters are 0.6, 0.3, and 0.1), it recalculates the average traffic volume over the past three hours to 200MB per hour. Combined with the current load data, it adjusts the predicted deviation value to 10%. Next, the system initiated a revision process for the pre-prepared strategy, invoking the dynamic resource scheduling engine to reallocate bandwidth resources based on the deviation. The initial allocated bandwidth was reduced from the predicted value of 250MB / hour to 210MB / hour, and the released 40MB / hour of resources was transferred to the backup pool. At the same time, through related business analysis, the system found that the correlation between the traffic of "online education users" and the frequency of use of real-time interactive functions is 0.72. It is estimated that interactive demand during peak hours accounts for approximately 30%. Therefore, an additional 15MB of bandwidth per hour is reserved to support low-latency transmission; Finally, the system generated an adjusted resource allocation plan, setting the total allocated bandwidth at 225MB per hour. An automated monitoring script was also set up. If traffic fluctuations exceeded 15% (i.e., exceeded 258.75MB per hour) within the next hour, resources would be automatically transferred from the backup pool to ensure network stability. This entire process is automatically executed by the system, forming a complete closed-loop logic from data collection, deviation analysis to resource adjustment.

[0034] Step S107, through the adjusted resource allocation plan, obtain the real-time feedback data of the network equipment during the execution process, and judge whether the dynamic response speed meets the requirements of service quality improvement based on the fluctuation of resource allocation efficiency to obtain the final optimization record.

[0035] Collect real-time feedback data through network devices, and use data processing modules to clean and format the collected data to obtain structured efficiency information; Based on the structured efficiency information, the time series analysis method is used to extract the fluctuation characteristics of allocation efficiency and determine the periodic pattern of the fluctuation characteristics. If the periodic pattern of the fluctuation characteristics exceeds the preset threshold, the corresponding response strategy is matched through the pre-established rule base to obtain the dynamic response adjustment direction; Based on the dynamic response adjustment direction, a decision tree model is used to classify and analyze the potential impact on service quality, generating optimization requirements for service quality; Based on service quality optimization requirements, we obtain pre-established configuration templates and generate preliminary resource adjustment plans based on efficiency information. If the initial resource adjustment plan does not meet the service quality requirements in the verification module, the data integration tool is used to fine-tune the plan to obtain an optimized configuration plan; Through the optimized configuration plan, the corresponding resource allocation instructions are generated and sent to the network equipment to generate the final optimization record.

[0036] For example, the system collects resource utilization data of edge servers and core switches in real time through the performance monitoring module of the network device, and records that the current bandwidth usage of the "remote office user" category is 320MB per hour, the CPU usage is 65%, and the memory usage is 72%; The system uses a time series analysis algorithm based on the resource utilization data of the past six hours (using the autoregressive moving average (ARIMA) model with parameters p=2, d=1, and q=1) to predict that the bandwidth demand for the next hour will be 350MB per hour, the CPU utilization will be 70%, and the memory utilization will be 75%. By comparing real-time data with predicted values, we calculated that the bandwidth deviation rate was 8.57%, the CPU deviation rate was 7.14%, and the memory deviation rate was 4%. The system then starts the dynamic response speed evaluation process, calling the resource efficiency analysis engine to calculate resource allocation efficiency based on real-time feedback data. The formula is: efficiency = (actual throughput / allocated resources) × 100%, resulting in a current bandwidth efficiency of 91.4%; To determine whether the response speed meets the service quality requirements, the system sets thresholds: response time must be less than 50 milliseconds and bandwidth efficiency must be higher than 90%; Analysis shows that the current response time is 45 milliseconds, which meets the requirement. However, the bandwidth efficiency is close to the lower threshold, triggering the optimization adjustment mechanism. The system, combined with business correlation analysis, found that the bandwidth demand of "remote office users" and the frequency of video conferencing usage were correlated at 0.68. It was estimated that video conferencing accounted for approximately 25% of peak hours. Therefore, an additional 20MB of bandwidth per hour was allocated to support high-concurrency video streams. At the same time, the system uses a dynamic scheduling algorithm (based on a greedy algorithm that prioritizes low-latency resources) to adjust CPU resources from 70% to 68%, freeing up 2% of CPU resources to the backup pool, and reduces memory allocation from 75% to 73%, freeing up 2% of memory resources. Finally, the system generates an optimization record, updates the bandwidth allocation to 340MB per hour, sets the CPU usage rate to 68%, and the memory usage rate to 73%. It also sets an automated monitoring script. If the response time exceeds 50 milliseconds or the bandwidth efficiency is less than 90% within the next hour, resources will be automatically transferred from the backup pool to ensure stable service quality.

[0037] Step S108: Based on the final optimization record, obtain the specific scenarios of network resource waste or resource shortage in the resource allocation process, extract the behavioral features in these scenarios, update the multi-dimensional feature data set of user behavior, and determine the input basis for the next round of optimization.

[0038] Extract scenario data of network resource waste and resource shortage from optimization records, use data cleaning tools to denoise and format the scenario data, and obtain a structured scenario feature set; Through the structured scene feature set, the K-means clustering algorithm is applied to divide the scene categories and determine the behavior patterns of each scene; If the behavior pattern deviates from the preset threshold, the corresponding behavior feature adjustment strategy is obtained through the pre-established rule matching module to generate the feature update direction; According to the feature update direction, use data integration tools to update the multi-dimensional feature dataset of user behavior to obtain an updated feature dataset; Using the updated feature dataset, the decision tree model is applied to classify and predict resource allocation scenarios, generating optimized requirements for resource allocation. According to the optimization requirements, the corresponding resource adjustment plan is matched from the pre-established configuration template library to generate a preliminary resource configuration plan; If the preliminary resource allocation plan does not meet the preset performance threshold, the data optimization tool is used to iteratively adjust the plan to obtain the final resource allocation plan.

[0039] For example, the system first scanned the network resource usage records for the past 24 hours through the resource allocation log analysis module and found that the bandwidth allocation for the "online education user" category during peak hours was 280MB per hour, but the actual usage was only 210MB per hour, with a waste rate of up to 25%. At the same time, during the off-peak period in the evening, the bandwidth demand suddenly increased to 300MB per hour, resulting in insufficient resources and causing the delay to increase to 80 milliseconds. Next, the system extracts behavioral features from these scenarios and calls the user behavior analysis engine. Based on the K-means clustering algorithm (with the number of cluster centers set to k=3), it categorizes user behavior into three categories: peak resource oversupply, off-peak resource shortage, and normal usage. The analysis shows a correlation of 0.75 between resource waste during peak hours and online course suspension rates, and a correlation of 0.82 between resource shortages during off-peak hours and temporary live streaming needs. Subsequently, the system updated the multi-dimensional feature dataset of user behavior, entering feature values such as a bandwidth waste rate of 25%, a latency peak of 80 milliseconds, a course pause rate of 0.75, and a live broadcast demand correlation of 0.82 into the database. Combined with historical data, the system used a decision tree algorithm (based on C4.5, with an information gain ratio threshold set to 0.1) to classify user behavior patterns and generate a feature weight distribution, where the live broadcast demand weight was 0.45 and the course pause weight was 0.3. Finally, based on the feature weights and classification results, the system determines the input basis for the next round of optimization, sets the dynamic bandwidth adjustment range to 220MB to 310MB per hour, and associates it with business scenarios, automatically taking the evening live broadcast demand as the priority factor (weight 0.5). Through the resource prediction model (based on linear regression, with an error range controlled within 5%), the system calculates the resource allocation plan for the next cycle to ensure improved resource matching.

[0040] Example 2 like Figure 3 As shown, this embodiment provides a network usage behavior analysis and bandwidth allocation system based on a portable WIFI device, which mainly includes: The data collection module is used to collect historical usage data of users on portable wireless network devices, and build a multi-dimensional feature data set of user behavior based on information such as network traffic, usage duration, and application type in different time periods and scenarios, thereby obtaining the specific manifestations of user behavior diversity and obtaining a preliminary set of behavioral features; The behavior clustering module is used to group user behaviors based on a preliminary set of behavioral features using clustering methods in machine learning. This module differentiates usage patterns for different purposes, such as entertainment, work, and learning, and identifies the core category distribution behind the diversity of user behaviors. The demand modeling module is used to obtain the network resource demand characteristics corresponding to each type of user behavior based on the core category distribution results. Combined with the records of traffic peaks and troughs in historical data, it determines the dynamic changes in demand and obtains the resource demand model for each type of behavior in different scenarios. The prediction and analysis module is used to establish a demand peak forecast analysis framework in advance based on the traffic change trend of each type of user behavior within a specific time period through the resource demand model. If the traffic demand of a certain type of user within a certain time period is predicted to exceed the preset threshold, the corresponding resource pre-allocation instruction is generated; The resource pre-allocation module is used to obtain the current available status of network resources based on the resource pre-allocation instructions, dynamically adjust the bandwidth allocation ratio based on the predicted peak demand data, and determine the specific execution parameters of the intelligent allocation mechanism; The dynamic adjustment module is used to obtain real-time connection load data from network devices based on the execution parameters of the intelligent allocation mechanism. If a deviation is detected between the actual demand of a certain type of user behavior and the predicted value, the advance preparation strategy correction process is triggered to obtain an adjusted resource allocation plan; The performance evaluation module is used to obtain real-time feedback data from network devices during the execution process based on the adjusted resource allocation plan. Based on the fluctuation of resource allocation efficiency, it determines whether the dynamic response speed meets the requirements for improving service quality and obtains the final optimization record. The dataset update module is used to obtain specific scenarios of network resource waste or resource shortage during the resource allocation process based on the final optimization records, extract behavioral features in these scenarios, update the multi-dimensional feature dataset of user behavior, and determine the input basis for the next round of optimization.

[0041] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A network usage behavior analysis and bandwidth allocation method based on a portable WIFI device, characterized in that: The method comprises: By collecting historical user usage data from portable wireless network devices, we construct a multi-dimensional feature dataset of user behavior based on information such as network traffic, usage duration, and application type in different time periods and scenarios. This allows us to capture the specific manifestations of user behavior diversity and obtain a preliminary set of behavioral features. Based on a preliminary set of behavioral characteristics, user behaviors are grouped, usage patterns for different purposes such as entertainment, work, and study are differentiated, and the core category distribution behind the diversity of user behaviors is determined; Based on the core category distribution results, we obtain the network resource demand characteristics corresponding to each type of user behavior. Combined with the records of traffic peaks and troughs in historical data, we determine the dynamic changes in demand and derive resource demand models for each type of behavior in different scenarios. Through the resource demand model, an analysis framework for demand peak prediction is established in advance based on the traffic change trend of each type of user behavior within a specific time period. If it is predicted that the traffic demand of a certain type of user within a certain time period exceeds the preset threshold, the corresponding resource pre-allocation instructions are generated.

2. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 1, characterized in that: The preliminary behavioral feature set obtained includes: By collecting historical usage data from portable wireless network devices, we can obtain users' network traffic, usage duration, and application types in different time periods and scenarios, and build an original data set. Data preprocessing technology is used to clean the original data set, remove missing values and outliers, and obtain a standardized data set; Based on the standardized data set, the statistical characteristics of network traffic, usage duration and application type are extracted to construct a multi-dimensional user behavior feature data set; If the feature dimensions in the multi-dimensional user behavior feature dataset meet the preset threshold, the feature is reduced in dimension using the principal component analysis algorithm to obtain a reduced-dimensional behavior feature set. Based on the reduced dimensionality behavioral feature set, the K-means clustering algorithm is used to classify user behaviors and obtain subsets of different behavioral patterns. By analyzing subsets of different behavior patterns, we extract the behavioral diversity of each subset in different time periods and scenarios, and obtain the user behavior diversity characteristics. If the user behavior diversity features match the preset behavior pattern template, a user behavior feature set is generated based on the matching results.

3. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 1, characterized in that: The core category distribution behind determining user behavior diversity includes: By analyzing the behavioral characteristics in the preliminary set, the user behaviors are grouped using clustering methods to obtain the initial category distribution; Based on the initial category distribution, corresponding usage pattern features are extracted for entertainment mode, work mode, and learning mode, and the behavioral feature subsets under each mode are determined; If the distribution density of the behavioral feature subset in a certain mode is lower than the preset threshold, the user behavior in that mode is clustered again to obtain more refined core categories; By dividing core categories into time periods and analyzing the behavior behind each category in different time periods, we can obtain the time-related usage pattern distribution; Based on the time-related usage pattern distribution and the specific scenarios of entertainment mode, work mode, and learning mode, we can determine the changing trends of the behavioral characteristics of each category in different scenarios. If the trend of behavioral characteristics changes significantly in a certain scenario, the data in that scenario will be weighted and adjusted through the information processing link to determine the final category distribution result; By performing a multi-dimensional comparison of the final category distribution results, we can obtain the comprehensive performance of user behavior in each category and determine the deep pattern associations behind the behavior.

4. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 1, characterized in that: Obtaining the resource requirement model for each type of behavior in different scenarios includes: For core categories and distribution results, we use data integration to obtain network resource usage records for each type of user behavior from historical traffic. By combining peak and trough records, we determine the resource consumption range for each type of behavior in different time periods. Based on the resource consumption range, the information processing link is used to analyze demand characteristics and demand changes, and the resource demand fluctuations of each type of user behavior under the time pattern are determined. If the fluctuation of a category exceeds the preset threshold, it is marked with a priority to obtain a set of marked categories; For the labeled category set, we divide the scenarios by differentiating them to obtain the network resource allocation for each type of user behavior in different scenarios. Combined with the data distribution in historical traffic, we determine the resource demand priority for each type of behavior in a specific scenario. Based on resource demand priorities, a data mapping method is used to analyze the correlation between demand changes and scenario differences. If the demand change trend in a certain scenario is inconsistent with historical traffic records, the data for that scenario is corrected through a weighted adjustment process to obtain an adjusted demand distribution. Based on the adjusted demand distribution, we analyze temporal patterns to obtain the foundation for resource model construction for each type of user behavior in different time periods. We then combine peak and trough records to determine the applicability of resource models in different time periods. Based on the applicability of the resource model, the K-means clustering method is used to group network resources. The matching degree between each group of resources and user behavior is evaluated to obtain the final resource allocation plan. For the final resource allocation plan, through a comprehensive analysis of scenario differences and time patterns, we obtain the resource model optimization direction for each type of user behavior in different scenarios and time periods, and determine the optimized resource allocation strategy.

5. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 1, characterized in that: Generating the corresponding resource pre-allocation instruction includes: For each type of user behavior, we obtain traffic change records within a specific time period from historical data. Combined with the characteristics of the change trend, we build an initial traffic demand analysis basis and obtain the traffic change pattern for each type of behavior. Based on traffic change patterns, a pre-built peak forecast analysis framework is used to estimate traffic demand within a specific time period, determine traffic change trends for each type of user behavior in future time periods, and identify potential peak time distributions. For peak time distribution, if the traffic demand forecast value of a certain type of user behavior exceeds the preset threshold, the resource demand of this type of behavior is prioritized through the information processing link to obtain a sorted resource demand list; Obtain a preliminary resource pre-allocation plan based on the sorted resource demand list. Analyze the feasibility of resource pre-allocation based on traffic trends and peak forecast results, and determine the adjusted allocation strategy framework. Based on the adjusted allocation strategy framework, information integration tools are used to evaluate the matching degree between traffic demand and resource pre-allocation. If the matching degree does not meet the preset standard, the allocation strategy is optimized through the data correction process to obtain the optimized resource allocation plan; Generate corresponding resource pre-allocation instructions based on the optimized resource allocation plan. Combined with the traffic change pattern within a specific time period, the allocation instructions are sent to the relevant resource management modules through the instruction distribution mechanism to complete the resource pre-configuration process. For the resource pre-allocation process, obtain comparison data between actual traffic demand and predicted results, analyze the dynamic adaptability of resource allocation through continuous monitoring tools, judge the execution status of resource pre-allocation instructions, and determine the direction of subsequent adjustments.

6. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 1, characterized in that: Also includes: According to the resource pre-allocation instructions, the current available status of network resources is obtained. Combined with the predicted peak demand data, the bandwidth allocation ratio is dynamically adjusted to determine the specific execution parameters of the intelligent allocation mechanism. Based on the execution parameters of the intelligent allocation mechanism, real-time connection load data from network devices is obtained. If a deviation is detected between the actual demand of a certain type of user behavior and the predicted value, the advance preparation strategy correction process is triggered to obtain an adjusted resource allocation plan; Through the adjusted resource allocation plan, we can obtain real-time feedback data from network devices during the execution process. Based on the fluctuation of resource allocation efficiency, we can determine whether the dynamic response speed meets the requirements for improving service quality and obtain the final optimization record. Based on the final optimization records, we obtain specific scenarios where network resource waste or resource shortage occurs during the resource allocation process. Based on the behavioral feature extraction results in these scenarios, we update the multi-dimensional feature dataset of user behavior and determine the input basis for the next round of optimization.

7. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 6, characterized in that: The specific execution parameters of the intelligent allocation mechanism include: By combining the available bandwidth status data obtained from the network resource management module with the peak demand distribution information, the current network resource load is analyzed to obtain a preliminary assessment result of bandwidth allocation; Based on the preliminary assessment results of bandwidth allocation, a pre-established allocation optimization model is used to perform matching calculations based on peak demand distribution information to determine the adjustment direction of bandwidth allocation ratios. If the adjustment direction of the bandwidth allocation ratio does not match the current network resource load, the status data will be checked again through the information processing link to determine whether the allocation ratio needs to be corrected; The configuration requirements of the intelligent allocation mechanism are obtained by adjusting the allocation ratio, and the initial value range of the execution parameters is determined based on the peak demand distribution information. Based on the initial value range of the execution parameters, the logistic regression model is used to optimize and adjust the parameter values to obtain the final execution parameters of the intelligent allocation mechanism; If the final execution parameters do not reach the preset thresholds during the verification of the information processing phase, the parameter values are fine-tuned through the data integration tool to determine the adjusted execution parameters; According to the adjusted execution parameters, the corresponding bandwidth allocation instructions are generated by the instruction management module and sent to the network resource management module to complete the dynamic allocation process of bandwidth resources.

8. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 6, characterized in that: The adjusted resource allocation plan includes: Collect real-time data of connected loads through network devices, and use data processing modules to clean and format the collected data to obtain structured load information; If the actual demand for a certain type of user behavior in the structured load information deviates from the predicted value by more than a preset threshold, a dynamic correction process is activated through a trigger mechanism to determine the correction requirement; Based on the correction requirements, the support vector machine model is used to classify and analyze the actual needs and predicted deviations of user behavior to obtain the direction of resource adjustment; Based on the resource adjustment direction, obtain the pre-established configuration template and generate a preliminary resource configuration plan based on the load information; If the preliminary resource allocation plan does not meet the preset threshold in the verification of the data processing module, the data integration tool is used to fine-tune the plan to obtain an optimized allocation plan; Generate corresponding resource allocation instructions based on the optimized configuration plan and send them to network devices to complete dynamic adjustment of resource configuration; Based on the adjusted resource configuration, new connection load data is collected in real time, and the applicability of the allocation plan is cyclically verified to obtain continuously optimized configuration results.

9. The method for analyzing network usage behavior and allocating bandwidth based on a portable WIFI device according to claim 6, characterized in that: The final optimization record includes: Collect real-time feedback data through network devices, and use data processing modules to clean and format the collected data to obtain structured efficiency information; Based on the structured efficiency information, the time series analysis method is used to extract the fluctuation characteristics of allocation efficiency and determine the periodic pattern of the fluctuation characteristics. If the periodic pattern of the fluctuation characteristics exceeds the preset threshold, the corresponding response strategy is matched through the pre-established rule base to obtain the dynamic response adjustment direction; Based on the dynamic response adjustment direction, a decision tree model is used to classify and analyze the potential impact on service quality, generating optimization requirements for service quality; Based on service quality optimization requirements, we obtain pre-established configuration templates and generate preliminary resource adjustment plans based on efficiency information. If the initial resource adjustment plan does not meet the service quality requirements in the verification module, the data integration tool is used to fine-tune the plan to obtain an optimized configuration plan; Through the optimized configuration plan, the corresponding resource allocation instructions are generated and sent to the network equipment to generate the final optimization record.

10. A network usage behavior analysis and bandwidth allocation system based on portable WIFI devices, characterized in that: The system comprises: The data collection module is used to collect historical usage data of users on portable wireless network devices, and build a multi-dimensional feature data set of user behavior based on information such as network traffic, usage duration, and application type in different time periods and scenarios, thereby obtaining the specific manifestations of user behavior diversity and obtaining a preliminary set of behavioral features; The behavior clustering module is used to group user behaviors based on a preliminary set of behavioral features using clustering methods in machine learning. This module differentiates usage patterns for different purposes, such as entertainment, work, and learning, and identifies the core category distribution behind the diversity of user behaviors. The demand modeling module is used to obtain the network resource demand characteristics corresponding to each type of user behavior based on the core category distribution results. Combined with the records of traffic peaks and troughs in historical data, it determines the dynamic changes in demand and obtains the resource demand model for each type of behavior in different scenarios. The prediction and analysis module is used to establish a demand peak forecast analysis framework in advance based on the traffic change trend of each type of user behavior within a specific time period through the resource demand model. If the traffic demand of a certain type of user within a certain time period is predicted to exceed the preset threshold, the corresponding resource pre-allocation instruction is generated; The resource pre-allocation module is used to obtain the current available status of network resources based on the resource pre-allocation instructions, dynamically adjust the bandwidth allocation ratio based on the predicted peak demand data, and determine the specific execution parameters of the intelligent allocation mechanism; The dynamic adjustment module is used to obtain real-time connection load data from network devices based on the execution parameters of the intelligent allocation mechanism. If a deviation is detected between the actual demand of a certain type of user behavior and the predicted value, the advance preparation strategy correction process is triggered to obtain an adjusted resource allocation plan; The performance evaluation module is used to obtain real-time feedback data from network devices during the execution process based on the adjusted resource allocation plan. Based on the fluctuation of resource allocation efficiency, it determines whether the dynamic response speed meets the requirements for improving service quality and obtains the final optimization record. The dataset update module is used to obtain specific scenarios of network resource waste or resource shortage during the resource allocation process based on the final optimization records, extract behavioral features in these scenarios, update the multi-dimensional feature dataset of user behavior, and determine the input basis for the next round of optimization.

Images