Hotspot service flow identification, distribution method and device
By combining a hierarchical identification model with a global cache pool, the problem of limited transmission bandwidth in 5G networks is solved, enabling cross-base station collaborative identification and caching, thereby improving transmission efficiency and bandwidth utilization.
Patent Information
- Application Number
- CN202611331759.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-28
- Publication Date
- 2026-11-17
AI Technical Summary
In 5G networks, when transmission bandwidth is limited or has reached its physical limit, existing technologies cannot effectively utilize air interface capabilities and lack intelligent cross-base station collaborative identification and caching capabilities, resulting in serious waste of transmission ring bandwidth and making it difficult to achieve cross-base station collaborative optimization.
A hierarchical identification model is used to identify the multidimensional features of the service flow. The model parameters are updated by combining machine learning models and feedback data. Cross-base station collaborative identification and caching are achieved through a global cache pool and a two-level distribution mechanism, reducing backhaul bandwidth requirements.
It enables intelligent identification of hotspot service flows on the access network side, breaks through the physical bandwidth limit of the transmission ring, and compresses the backhaul bandwidth requirement from the user level to the base station level through a global cache pool and a two-level distribution mechanism, thereby optimizing transmission efficiency.
Smart Images

Figure CN123124136A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and more specifically, to a method and apparatus for identifying and distributing hotspot service flows. Background Technology
[0002] In 5G networks, to support multiple users simultaneously receiving the same data (such as live streaming, group communication, and batch file downloads), two main methods are currently used: unicast and 5G multicast / broadcast (5MBS). In unicast, each user independently establishes a PDU (Protocol Data Unit) session. The core network's UPF (User Plane Function) forwards a data stream independently for each user. The base station receives multiple streams with the same content and sends them to each user separately. This results in a linear increase in transmission bandwidth requirements with the number of users. In 5MBS, the core network establishes a multicast session and sends a single data stream to the base station through a shared tunnel. The base station then sends the data to multiple users via PTM (Point-to-Multipoint) over the air interface. This requires the core network, transmission network, and terminals to support IP multicast across the entire link.
[0003] However, neither the aforementioned unicast nor 5MBS methods can effectively utilize air interface capabilities and achieve cross-base station collaborative optimization in scenarios where transmission bandwidth is limited or has reached its physical limit.
[0004] First, transmission bandwidth becomes a bottleneck with a physical limit. With the continuous growth of air interface capabilities (such as millimeter wave and multi-cell aggregation), the downlink capability of a single base station can reach tens of Gbps or even higher. However, the transmission ring bandwidth between the base station and upstream has a physical limit; upgrades require replacing optical fibers and equipment, which is extremely costly and time-consuming. For example, in scenarios such as OTA (Over-The-Air) upgrades in smart car factories or large concerts, the massive traffic aggregated by multiple base stations far exceeds the physical capacity of the transmission ring, preventing the full utilization of air interface capabilities. Furthermore, traditional multicast solutions are difficult to implement due to complex deployment, high industry chain collaboration costs, poor terminal support, and inability to handle asynchronous request scenarios.
[0005] Secondly, the relevant technical solutions lack intelligent dynamic recognition capabilities. Whether unicast or multicast, service identification and distribution decisions rely on static configurations (such as pre-allocated multicast addresses and fixed service templates) or simple rules (such as traffic statistics based on IP 5-tuples). These methods cannot perceive the service content itself, cannot predict hotspot trends, and can only perform post-event statistics. They cannot adapt to different scenarios (such as concerts, factories, and online classes, where service models differ greatly but rules are fixed); furthermore, each base station processes independently, lacking a global perspective and failing to achieve cross-base station collaborative perception and cache sharing. This results in significant bandwidth waste in the transmission ring, making it difficult to break through the physical limits of backhaul bandwidth and achieve intelligent cross-base station collaborative caching and secondary distribution without relying on core network and transmission network modifications.
[0006] There is currently no effective solution to the above problems. Summary of the Invention
[0007] This application provides a method and apparatus for identifying and distributing hotspot service flows, which at least solves the technical problem that the backhaul bandwidth cannot exceed the physical limit due to the lack of cross-base station collaborative identification and caching capabilities on the access network side in related technologies.
[0008] According to one aspect of the embodiments of this application, a hotspot service flow identification method is provided, applied to a smart board, comprising: acquiring multi-dimensional features of the service flow; identifying the multi-dimensional features using a hierarchical identification model, and outputting a hotspot confidence level corresponding to the service flow, wherein the hierarchical identification model includes: a first identification path based on preset rules for matching, a second identification path based on a machine learning model for multi-dimensional feature analysis, and a third identification path based on feedback data for updating model parameters; aggregating request patterns from multiple base stations, determining hotspot service flows based on hotspot confidence levels and request patterns, and storing the hotspot service flows in a global cache pool, wherein the request patterns are used to reflect the degree of aggregation of group requests of the service flow within the scope of multiple base stations.
[0009] Optionally, the method further includes: performing multi-level distribution operations on hot service flows in the global cache pool according to the request pattern, wherein the multi-level distribution operations include: first-level distribution, distributing the hot service flows to multiple base stations; and second-level distribution, copying the received hot service flows through multiple base stations and distributing them to multiple user equipment.
[0010] Optionally, the multidimensional features include at least two or more of the following: spatiotemporal features of the service flow, content features, user behavior features, network status features, historical pattern features, and service layer features.
[0011] Optionally, a hierarchical recognition model is used to identify multidimensional features, including: in the first recognition path, matching multidimensional features based on preset rules or business labels in the rule base; if the match is successful, outputting the hotspot confidence score corresponding to the business flow; if the match fails or the hotspot confidence score is lower than a first threshold, proceeding to the second recognition path; in the second recognition path, using a machine learning model to identify multidimensional features and obtaining an initial hotspot confidence score; if the initial hotspot confidence score is greater than or equal to the second threshold, determining the initial hotspot confidence score as the hotspot confidence score corresponding to the business flow; if the initial hotspot confidence score is less than the second threshold, proceeding to the third recognition path; in the third recognition path, marking the business flow as a sample to be learned and triggering model parameter updates in online or offline mode.
[0012] Optionally, the method further includes: acquiring actual heat data of business flows, comparing the actual heat data with the initial hotspot confidence level, and generating feedback samples, wherein the feedback samples are used to characterize the deviation between the model recognition results and actual business needs; determining the performance indicators of the hierarchical recognition model based on the feedback samples; and incrementally updating the model parameters in the hierarchical recognition model when the performance indicators are lower than a preset performance threshold or the number of feedback samples reaches a preset number threshold.
[0013] Optionally, aggregating request patterns from multiple base stations and determining hotspot service flows based on hotspot confidence and request patterns includes: when the hotspot confidence is greater than a first preset heat threshold, obtaining request records from multiple base stations for the target service flow, and forming a request pattern for the corresponding base station based on the request records, wherein the target service flow is any one of the service flows, and the request records include at least one of the following: the total number of requests for the service flow within the ring network, the number of requesting base stations, and the request growth rate; determining an aggregation heat index for multiple base stations based on the request patterns, wherein the aggregation heat index is used to reflect the degree of aggregation of group requests for the service flow within the range of multiple base stations and the heat diffusion trend; when the aggregation heat index is greater than a second preset heat threshold, determining the target service flow as a hotspot service flow, and generating a corresponding hotspot identifier.
[0014] Optionally, the method further includes: predicting the target base station for hot service flows to be distributed and which has not yet generated a request pattern based on the request growth rate or historical pattern characteristics of the hot service flows; sending a pre-caching instruction to the target base station, wherein the pre-caching instruction is used to instruct the target base station to prepare to receive hot service flows from the smart board, or to instruct the smart board to preload hot service flows in the global cache pool in advance.
[0015] According to another aspect of the embodiments of this application, a hotspot service flow distribution method is also provided, applied to a base station, comprising: receiving a hotspot service flow sent by a smart board, wherein the smart board is used to acquire multi-dimensional features of the service flow; identifying the multi-dimensional features using a hierarchical identification model and outputting a hotspot confidence level corresponding to the service flow, wherein the hierarchical identification model includes: a first identification path for matching based on preset rules, a second identification path for multi-dimensional feature analysis based on a machine learning model, and a third identification path for updating model parameters based on feedback data of the identification results; aggregating request patterns of multiple base stations and determining the hotspot service flow based on the hotspot confidence level and the request pattern, wherein the request pattern is used to represent the degree of aggregation of group requests of the service flow within the range of multiple base stations; copying the hotspot service flow and distributing it to multiple user equipments.
[0016] According to another aspect of the embodiments of this application, a hotspot service flow identification device is also provided, comprising: an acquisition module for acquiring multi-dimensional features of a service flow; an identification module for identifying the multi-dimensional features using a hierarchical identification model and outputting a hotspot confidence level corresponding to the service flow, wherein the hierarchical identification model includes: a first identification path based on preset rules for matching, a second identification path based on a machine learning model for multi-dimensional feature analysis, and a third identification path based on feedback data for updating model parameters; and a determination module for aggregating request patterns from multiple base stations, determining hotspot service flows based on hotspot confidence levels and request patterns, and storing the hotspot service flows in a global cache pool. The request patterns reflect the degree of aggregation of group requests for the service flow across multiple base stations.
[0017] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and used to execute the above-described hotspot service flow identification or distribution method.
[0018] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned hot spot service flow identification or distribution method by running the computer program.
[0019] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions, which, when executed by a processor, implement the above-described hot spot service flow identification or distribution method.
[0020] In this embodiment, multi-dimensional features of the service flow are acquired; a hierarchical identification model is used to identify the multi-dimensional features and output the hotspot confidence level corresponding to the service flow. The hierarchical identification model includes: a first identification path based on preset rules for matching, a second identification path based on machine learning model for multi-dimensional feature analysis, and a third identification path based on feedback data for model parameter updating. The request patterns of multiple base stations are aggregated, and hotspot service flows are determined based on the hotspot confidence level and request patterns. The hotspot service flows are stored in a global cache pool. The request patterns are used to reflect the degree of aggregation of group requests of the service flow in the range of multiple base stations. This achieves the purpose of intelligently identifying hotspot service flows on the access network side and realizing cross-base station shared caching. This achieves the technical effect of compressing the backhaul bandwidth requirement from the user level to the base station level, breaking through the physical bandwidth limit of the transmission ring. This solves the technical problem that the backhaul bandwidth cannot break through the physical limit due to the lack of cross-base station collaborative identification and caching capabilities on the access network side in related technologies. Attached Figure Description
[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0022] Figure 1 This is a hardware structure diagram of a computer terminal for implementing a hotspot service flow identification or distribution method according to an embodiment of this application;
[0023] Figure 2 This is a flowchart of a hotspot service flow identification method according to an embodiment of this application;
[0024] Figure 3 This is a schematic diagram of a network element architecture according to an embodiment of this application;
[0025] Figure 4 This is a schematic diagram of a three-layer progressive recognition architecture of a hierarchical recognition model according to an embodiment of this application;
[0026] Figure 5 This is a schematic diagram of a cross-base station collaborative pre-caching process according to an embodiment of this application;
[0027] Figure 6 This is a flowchart of a hotspot service flow distribution method according to an embodiment of this application;
[0028] Figure 7 This is a structural diagram of a hotspot service flow identification device according to an embodiment of this application;
[0029] Figure 8 This is a structural diagram of a hotspot service flow distribution device according to an embodiment of this application. Detailed Implementation
[0030] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.
[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0032] First, some nouns or terms that appear in the explanation of the embodiments of this application shall be interpreted as follows:
[0033] Smart board: A smart hardware unit deployed on the base station side or access network side, possessing independent AI computing power, storage, and inference capabilities. It typically serves as the data entry and exit point for one or more base stations in a transmission ring network. The smart board is not only a traffic forwarding point but also an executor of intelligent decisions, responsible for managing the data interaction of subordinate base stations and achieving ring-network-level data aggregation and distribution.
[0034] Hotspot traffic refers to data streams that are simultaneously or asynchronously requested by multiple users within a specific time window, often of the same or highly similar nature. Common examples include live streams, firmware download streams (such as over-the-air (OTA) updates for automobiles), popular videos, and online course videos. Unlike regular traffic, hotspot traffic is highly shareable, meaning different users are requesting the same data content, thus offering the potential to save bandwidth through caching and replication.
[0035] Global Cache Pool: A shared cache established on the smart board, accessible to all base stations it manages. The AI engine determines the cached content and eviction policy. The global cache pool enables resource sharing across base stations, avoiding storage waste and duplicate data transmission caused by multiple base stations independently caching the same content.
[0036] Two-level distribution: The core execution mechanism for bandwidth optimization in this application consists of two levels: Level 1 distribution: executed by the smart board, sending a single hotspot data stream to each managed base station (instead of one per user); Level 2 distribution: executed by the base station, replicating the received single data stream on the base station side and distributing it to multiple users under its coverage. Through this mechanism, the backhaul link (from the smart board to the base station) only needs to transmit one data, greatly saving transmission bandwidth.
[0037] This application provides a method for identifying hotspot service flows and a method for distributing hotspot service flows. The hotspot service flow identification or distribution method can run on... Figure 1 The computer terminal shown is described below.
[0038] The hotspot service flow identification or distribution method embodiments provided in this application can be executed on mobile terminals, computer terminals or similar computing devices. Figure 1 A hardware block diagram of a computer terminal for implementing a method for identifying or distributing hotspot service flows is shown. Figure 1 As shown, the computer terminal 10 may include a processor 102, a memory 104 for storing data, and a transmission module 106 for communication functions. The processor 102 may include one or more processors. For ease of explanation, Figure 1 The diagram illustrates multiple processors, including a first processor 102a, a second processor 102b, ..., an nth processor 102n. The processor 102 may be a processing device including, but not limited to, a microcontroller unit (MCU) or a field programmable gate array (FPGA).
[0039] In addition, a computer terminal may also include: a display, an input / output interface (I / O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.
[0040] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).
[0041] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the hotspot service flow identification or distribution method in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned hotspot service flow identification or distribution method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
[0042] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a radio frequency (RF) module, used for data interaction with external networks or devices via wired and / or wireless network connections.
[0043] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10. In addition, the computer terminal 10 can also connect to a cursor control device and a keyboard via an input / output interface. The cursor control device (e.g., a mouse, trackball, or touchpad) is mainly used to transmit directional information and command selection to the terminal and control the movement of the cursor on the display; the keyboard is mainly used to receive letters, numbers, and other control commands input by the user, thereby enabling interaction between the user and the computer terminal 10.
[0044] It should be noted here that, in some optional embodiments, the above... Figure 1The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.
[0045] In the above operating environment, this application provides an embodiment of a method for identifying or distributing hot service flows. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Also, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than that shown here.
[0046] Figure 2 This is a flowchart of a hotspot service flow identification method according to an embodiment of this application, such as... Figure 2 As shown, this method is applied to smart boards and includes the following steps:
[0047] Step S202: Obtain the multidimensional features of the business flow.
[0048] Step S204: A hierarchical recognition model is used to identify multidimensional features and output the hotspot confidence level corresponding to the business flow. The hierarchical recognition model includes: a first recognition path based on preset rules for matching, a second recognition path based on machine learning model for multidimensional feature analysis, and a third recognition path based on feedback data for model parameter updating.
[0049] Step S206: Aggregate the request patterns of multiple base stations, determine the hot service flow based on the hotspot confidence and request pattern, and store the hot service flow in the global cache pool. The request pattern is used to reflect the degree of aggregation of group requests of the service flow in the range of multiple base stations.
[0050] Through the above steps S202 to S206, the goal of intelligently identifying hotspot service flows on the access network side and realizing cross-base station shared caching is achieved. This reduces the backhaul bandwidth requirement from the user level to the base station level, breaking through the physical bandwidth limit of the transmission ring. This solves the technical problem that the backhaul bandwidth cannot break through the physical limit due to the lack of cross-base station collaborative identification and caching capabilities on the access network side.
[0051] In this embodiment, the core idea is to bring AI technology down to the edge of the access network, utilize the local computing power of the built-in smart board to perform real-time multi-dimensional feature analysis and online learning inference on the passing service flow, and automatically discover shareable hot service flows; through a global cache pool and a two-level distribution mechanism (smart board → base station → user), a transmission mode of "one data flow, serving multiple base stations, and covering a massive number of users" is realized, fundamentally compressing the backhaul bandwidth requirement from the user level to the base station level.
[0052] The embodiments of this application are particularly applicable to high-density user scenarios, large-scale concurrent download scenarios, and mobile communication network scenarios where transmission bandwidth is limited or has reached its physical limit, such as large-scale asynchronous concurrent firmware download in smart car factories, large-scale concerts and sports events, uplink video optimization of multiple cameras in smart factories, multi-terminal video distribution in remote education classrooms, and millimeter-wave high-speed transmission optimization in various large-scale events. Specifically, they can be applied to access devices such as 5G / 5G-A / 6G base stations, 4GeNB, Wi-FiAP, and edge gateways in various scenarios.
[0053] Figure 3 This is a schematic diagram of a network element architecture according to an embodiment of this application, such as... Figure 3 As shown, it uses a "smart board" as the brain of the ring network, employing a hierarchical (AI recognition engine - global cache pool - base station scheduling management) and distributed (smart board - multiple base stations - multiple terminal devices) design philosophy. Through the collaborative work of primary distribution (single data stream) and secondary distribution (replicated distribution), it achieves the identification and distribution scheduling of hotspot service flows from central intelligent decision-making to base stations and then to terminal devices (such as vehicles). Its core advantage lies in combining powerful AI capabilities with distributed, real-time communication capabilities, and ensuring the continuous evolution and maintenance of the system through OTA technology, providing strong technical support for large-scale intelligent transportation applications (such as autonomous driving and vehicle-to-everything).
[0054] The following combination Figure 3 The hot spot service flow identification and scheduling distribution logic in the above steps S202 to S206 are explained in detail.
[0055] In step S202 above, the main responsibility is to extract multi-dimensional features such as time, content, behavior, and network status from business traffic to provide comprehensive data input for the AI model, so as to accurately judge the potential popularity of business traffic in the future.
[0056] In the embodiments of this application, the multidimensional features include at least two or more of the following: spatiotemporal features of the service flow, content features, user behavior features, network status features, historical pattern features, and service layer features.
[0057] Specifically, unlike the identification methods based on IP quintuples or simple frequency statistics in related technologies, this application's embodiments introduce for the first time a multi-dimensional feature fusion identification mechanism based on machine learning / deep learning models on the access network side. The hierarchical identification model integrates at least two or more feature dimensions from the service flow during deep inference, as shown in Table 1.
[0058] Table 1. Examples of Multidimensional Features
[0059]
[0060] By integrating multi-dimensional features, the hierarchical identification model can overcome the limitations of misjudgment caused by a single indicator (such as request frequency), comprehensively capture the true popularity attributes of business flows, thereby significantly improving the accuracy and adaptability of hot business flow identification and ensuring that the global cache pool can accurately store business data with high reusability.
[0061] In step S204 above, the aim is to combine AI technology to calculate hotspot confidence through a three-layer architecture of rule matching, model inference and online learning, so as to dynamically and adaptively identify hotspot business flows while balancing low latency and high accuracy.
[0062] Optionally, a hierarchical recognition model is used to identify multidimensional features, including: in the first recognition path, matching multidimensional features based on preset rules or business labels in the rule base; if the match is successful, outputting the hotspot confidence score corresponding to the business flow; if the match fails or the hotspot confidence score is lower than a first threshold, proceeding to the second recognition path; in the second recognition path, using a machine learning model to identify multidimensional features and obtaining an initial hotspot confidence score; if the initial hotspot confidence score is greater than or equal to the second threshold, determining the initial hotspot confidence score as the hotspot confidence score corresponding to the business flow; if the initial hotspot confidence score is less than the second threshold, proceeding to the third recognition path; in the third recognition path, marking the business flow as a sample to be learned and triggering model parameter updates in online or offline mode.
[0063] In this embodiment, the hierarchical recognition model in the smart board adopts a three-layer progressive recognition architecture, such as... Figure 4 As shown, it specifically includes:
[0064] First identification path L1: Fast path (rule / tag matching)
[0065] The L1 path relies on preset rules for automated matching, such as whitelist matching (for known hotspots or security entities) or business tag matching (by business tags marked by partners). The processing latency is usually less than 1ms, but its hit rate is relatively low, only 20%-30%.
[0066] If the business flow is matched, the processing is complete, the hotspot confidence score after successful matching is output, and the process ends; if it is not matched (rule / tag matching fails or the output hotspot confidence score is lower than the first threshold), it proceeds to the next level (L2 path).
[0067] Second recognition path L2: Deep path (model inference)
[0068] The L2 path employs a model-based inference mechanism, analyzing the complex attributes of the business flow through multi-dimensional feature extraction. It utilizes machine learning models such as lightweight neural networks or random forest inference for judgment and outputs initial hotspot confidence scores. Compared to the L1 path, the L2 path has a significantly increased processing latency, typically 5-20ms, but its recognition capability is also greatly improved, with a hit rate of 60%-70%.
[0069] If the business flow is hit (e.g., the initial hotspot confidence is greater than or equal to the second threshold), the processing is complete, and the output initial hotspot confidence is used as the final hotspot confidence of the business flow; if it is not hit (e.g., the initial hotspot confidence is less than the second threshold), it proceeds to the next level (L3 path).
[0070] Third recognition path L3: Learning path (online feedback)
[0071] The L3 path is the final level of the processing flow, designed to optimize the entire system through continuous learning. When the confidence level of the L2 decision is low, the business flow is sent to the L3 path. At this point, the system marks it as a new pattern, uses this unknown or abnormal business flow as a new training sample (i.e., the aforementioned learning sample), and triggers an online learning mechanism (or offline learning mechanism) to directly update and optimize the model using the learning sample. The processing latency of the L3 path is within an acceptable range, and its core value lies in enabling generational evolution of the model, thereby improving the overall accuracy and adaptability of the system in the long term.
[0072] Through the aforementioned three-layer progressive recognition architecture, a balance is achieved between low-latency fast matching and high-precision deep inference. In uncertain scenarios, an online / offline learning mechanism is introduced, enabling the AI model to automatically adapt to the dynamic changes in business scenarios without human intervention. This significantly improves the accuracy, real-time performance, and long-term adaptability of hotspot recognition.
[0073] In this embodiment, to achieve self-optimization and adaptive capability of the hierarchical recognition model in a dynamic network environment, a closed-loop feedback mechanism (online learning mechanism) is introduced. This mechanism drives incremental updates of the model by continuously monitoring the deviation between the recognition results and actual business performance. The online learning mechanism includes: acquiring actual heat data of business flows and comparing the actual heat data with the initial heat confidence to generate feedback samples, wherein the feedback samples are used to characterize the deviation between the model recognition results and actual business needs; determining the performance index of the hierarchical recognition model based on the feedback samples; and incrementally updating the model parameters in the hierarchical recognition model when the performance index is lower than a preset performance threshold or the number of feedback samples reaches a preset number threshold.
[0074] Specifically, after completing the hotspot confidence determination, the smart board will continuously track the actual heat data of the service flow in the subsequent time period. This actual heat data can be quantified in multiple dimensions, such as the total number of actual user requests of the service flow within a preset time window, the number of base stations making cross-base station collaborative requests, or the change in transmission bandwidth utilization caused by the service flow.
[0075] Subsequently, the actual popularity data is compared with the initial hotspot confidence score (or final hotspot confidence score) output by the model to generate feedback samples. These feedback samples are used to characterize the deviation between the model's identification results and actual business needs, specifically including two typical scenarios: one is a "false positive" sample, where the model determines a certain business flow as a hotspot (high confidence), but the actual popularity data shows that its user requests are sparse or the popularity is rapidly declining, indicating that the model is too sensitive; the other is a "false negative" sample, where the model does not determine a certain business flow as a hotspot (low confidence), but the actual popularity data shows that its request volume has surged, indicating that the model failed to capture potential hotspots in time.
[0076] Furthermore, based on the accumulated feedback samples, the performance metrics of the hierarchical recognition model are calculated periodically or quantitatively, such as precision, recall, F1 score, or mean absolute error. When the performance metrics are detected to be below a preset performance threshold (e.g., precision drops below 90%), or the number of accumulated feedback samples reaches a preset quantity threshold (e.g., every 1000 new samples collected or every hour), an incremental update of the model parameters is triggered.
[0077] During incremental updates, feedback samples are used to fine-tune the model, such as adjusting the weights of each feature dimension or the parameters of neural network layers, to make the model more closely reflect the current network environment and business model. The updated model parameters are saved and take effect immediately for subsequent business flow recognition, thereby ensuring that the model can maintain high-accuracy recognition capabilities over time and with the emergence of new scenarios, without the need for manual retraining.
[0078] In the above process, by establishing a feedback sample mechanism based on actual heat deviation, online incremental updates of the AI model were realized, enabling the system to automatically adapt to dynamic changes in network environment, business type and user behavior. This solved the problem that traditional offline training models could not cope with scene drift, and significantly improved the long-term accuracy of hotspot identification and the system's adaptive capability.
[0079] In step S206 above, the aim is to aggregate request data from multiple base stations to determine ring network-level hotspots and store the data in a global cache pool, thereby achieving cross-base station collaborative identification and centralized caching, and providing data support for subsequent secondary distribution.
[0080] Optionally, aggregating request patterns from multiple base stations and determining hotspot service flows based on hotspot confidence and request patterns includes: when the hotspot confidence is greater than a first preset heat threshold, obtaining request records from multiple base stations for the target service flow, and forming a request pattern for the corresponding base station based on the request records, wherein the target service flow is any one of the service flows, and the request records include at least one of the following: the total number of requests for the service flow within the ring network, the number of requesting base stations, and the request growth rate; determining an aggregation heat index for multiple base stations based on the request patterns, wherein the aggregation heat index is used to reflect the degree of aggregation of group requests for the service flow within the range of multiple base stations and the heat diffusion trend; when the aggregation heat index is greater than a second preset heat threshold, determining the target service flow as a hotspot service flow, and generating a corresponding hotspot identifier.
[0081] In this embodiment, the smart board serves as the data entry and exit point of the ring network, possessing the ability to manage multiple base stations (e.g., 5-10 base stations on a transmission ring). It achieves cross-base station hotspot identification and coordination by aggregating the request patterns of each base station. The specific execution steps are as follows:
[0082] First, when the AI recognition engine within the smart board detects a target service flow (i.e., any data flow being requested within the ring network, such as a firmware version or video stream) with a hotspot confidence level greater than a first preset hotspot threshold in the L2 path, if three users request the same firmware, the service flow is determined to have potential hotspot characteristics, and then enters the cross-base station collaborative analysis phase. At this time, the smart board sends query commands to all base stations it manages, or obtains the request records of the target service flow within the ring network from the traffic statistics modules reported by each base station. These request records include at least the following dimensions of information:
[0083] (1) Total number of requests: The total number of user requests received by all base stations in the ring network within the target time window for the target service flow;
[0084] (2) Number of requesting base stations: How many different base stations in the ring network have detected the request for the target service flow (used to determine the coverage of the hotspot, whether it is a single-site hotspot or a multi-site hotspot).
[0085] (3) Request growth rate: The increase in the number of requests per unit time (such as the last second or minute), used to predict whether the popularity is in the rising period.
[0086] Secondly, based on the obtained request records of each base station, a request pattern for each base station is constructed. For example, base station A may exhibit "high number of requests, low growth rate," while base stations B and C may also have similar requests. The smart board aggregates these local request patterns and calculates an aggregation heat index to reflect the degree of aggregation and heat diffusion trend of group requests throughout the entire ring network.
[0087] Finally, the aggregated heat index is compared with a second preset heat threshold. If the aggregated heat index is greater than the second preset heat threshold, it is confirmed that the target service flow not only has high local heat but also forms a group hotspot within the ring network, possessing value for cross-base station sharing. At this point, the target service flow is officially identified as a hotspot service flow, and a unique hotspot identifier (e.g., an ID generated based on a file hash value or URL) is generated.
[0088] Through the above process, a leap from independent judgment by a single base station to collaborative perception at the ring network level has been achieved, effectively avoiding missed or false identifications caused by the limited perspective of a single base station, and maximizing the utilization of the transmission ring bandwidth, compressing the original N times user-level bandwidth requirement to base station-level transmission overhead.
[0089] In this embodiment, after identifying the hot service flows, a global caching, pre-caching, and proactive push mechanism are also included. A detailed analysis follows:
[0090] (1) Global cache entry: When the hot service flow has not yet been cached, the smart board pulls a single data flow from the core network UPF and stores it in the global cache pool, and establishes a mapping with the hot spot identifier as the index.
[0091] For global cache pool management, the caching strategy adopts LRU (Least Recently Used) or LFU (Least Frequently Used) algorithms, and is dynamically adjusted based on the prediction results of hot topic trends by the AI engine. When a base station user requests data, the global cache pool is queried first; if a match is found, the cached data is directly retrieved; otherwise, the retrieval process is triggered. Simultaneously, when a significant drop in business activity or content version update is detected, a cache invalidation and release mechanism is automatically triggered to ensure efficient utilization of cache space and data freshness.
[0092] (2) Pre-caching and proactive push, including: predicting the target base station that hot service flows are to be distributed and have not yet generated a request pattern based on the request growth rate or historical pattern characteristics of hot service flows; sending a pre-caching instruction to the target base station, wherein the pre-caching instruction is used to instruct the target base station to prepare to receive hot service flows from the smart board, or to instruct the smart board to preload hot service flows in the global cache pool in advance.
[0093] Specifically, if the request growth rate indicates that the heat is spreading to the target base station that has not yet generated a request, the smart board can also proactively send a pre-caching instruction to the target base station to give advance instructions, or prepare the relevant content of the hot service flow in the global cache pool in advance, so as to respond in time when the request arrives and cope with the upcoming request surge.
[0094] By completing data preparation in advance before the hot spot traffic actually spreads to other base stations, the initial transmission delay caused by the sudden spread of hot spots is effectively eliminated, significantly improving the first packet acquisition speed and overall download experience for users in the early stage of hot spot formation, and realizing the optimal scheduling of ring network-level resources and seamless connection of service awareness.
[0095] Figure 5 This is a schematic diagram of a cross-base station collaborative pre-caching process according to an embodiment of this application, such as... Figure 5 As shown, this process describes a highly efficient intelligent caching system based on group awareness. The system uses intelligent cards to perceive group behavior and monitor the traffic trends of base station 1 in real time. For example, if the number of requests for flow A reported by base station 1 increases rapidly (3→15), it analyzes the rising popularity of flow A on base station 1. Subsequently, using cross-base station prediction technology, it determines that this traffic will also become popular on base station 2, which has similar user behavior patterns. Based on this prediction, an active pre-caching mechanism is triggered, instructing the UPF to retrieve flow A data from the network in advance, store it in the global cache pool, and send a pre-caching instruction to base station 2. When a user on base station 2 actually requests this flow, the system has already pre-loaded the data, thus achieving a cache hit and ultimately providing users with a seamless experience with zero waiting time and zero extra transmission. The entire process realizes a shift from passive response to proactive prediction, greatly optimizing user experience and network resource utilization efficiency.
[0096] In this embodiment of the application, step S208 is further included, which performs a multi-level distribution operation on the hot service flow in the global cache pool according to the request mode. The multi-level distribution operation includes: first-level distribution, which distributes the hot service flow to multiple base stations; and second-level distribution, which distributes the received hot service flow to multiple user equipment after copying it through multiple base stations.
[0097] In step S208 above, based on the request pattern, a multi-level distribution (second-level distribution) operation is performed on the hot business flows in the global cache pool, as shown in Table 2.
[0098] Table 2 Two-Level Distribution Mechanism Table
[0099]
[0100] During the primary distribution phase, the smart board, acting as the data entry and exit point of the ring network, identifies hot service flows and retrieves only one complete data stream from the core network (such as UPF) to its maintained global cache pool. Subsequently, based on the actual requests from users under each managed base station, it sends a single copy of the data to the base stations that require it. This mechanism significantly reduces the bandwidth usage on the transmission ring from "single stream bit rate × total number of users" to "single stream bit rate × number of base stations requiring transmission," achieving cross-base station traffic aggregation and intensive transmission.
[0101] In the secondary distribution phase, the base station, acting as the execution node on the access side, receives a single data stream from the smart board and replicates the data at the PDCP layer. Subsequently, based on the specific number of users requesting the hotspot within the coverage area, the replicated data is sent to each user terminal via unicast or multicast. This process is completely transparent to the user side; the user terminal still receives data using the conventional unicast method, requiring no modification to the terminal hardware or software. This ensures a seamless user experience while efficiently distributing data from the base station to multiple users.
[0102] Taking 10 millimeter-wave base stations as an example, where 8 of them have users requesting the same firmware:
[0103] In related technologies, transmission bandwidth = 500Mbps × 300 users = 150Gbps;
[0104] By adopting the embodiment of this application, the transmission bandwidth is 500Mbps × 8 base stations = 4Gbps, the bandwidth requirement is reduced by 97.3%, effectively breaking through the physical bandwidth limit of the transmission ring and achieving the ultimate optimization of transmission resource utilization.
[0105] Overall, this application's embodiments are the first to integrate AI technology into smart cards on the access network side, constructing a closed-loop intelligent hotspot flow identification and backhaul bandwidth optimization architecture. Its core lies in abandoning traditional solutions relying on static configuration or full-link multicast. Through an AI recognition engine deployed on the smart card, it integrates multi-dimensional features such as spatiotemporal, content, and behavior, employing a layered recognition architecture (L1 fast matching, L2 deep inference, L3 online learning) to achieve dynamic and accurate identification of hotspot service flows. It also utilizes the smart card's global perspective for cross-base station collaborative perception and pre-caching decisions. Ultimately, through a two-level distribution mechanism of "smart card → base station → user" and a global cache pool, the backhaul bandwidth requirement is compressed from the "user level" to the "base station level." This effectively breaks through the physical bandwidth limit of the transmission ring without relying on core network / transmission network modifications or terminal upgrades, solving network bottleneck problems in high-density concurrent scenarios.
[0106] Figure 6 This is a flowchart of a hotspot service flow distribution method according to an embodiment of this application, such as... Figure 6 As shown, this method is applied to a base station and includes the following steps:
[0107] Step S602: Receive hotspot service flows sent by the smart board, wherein the smart board is used to acquire multi-dimensional features of the service flows; use a hierarchical recognition model to identify the multi-dimensional features and output the hotspot confidence level corresponding to the service flows; wherein the hierarchical recognition model includes: a first recognition path based on preset rules for matching, a second recognition path based on machine learning model for multi-dimensional feature analysis, and a third recognition path for updating model parameters based on feedback data of the recognition results; aggregate request patterns from multiple base stations and determine hotspot service flows based on hotspot confidence level and request patterns; wherein the request pattern is used to represent the degree of aggregation of group requests of the service flows within the range of multiple base stations.
[0108] Step S604: Copy the hot service flow and distribute it to multiple user devices.
[0109] Through the above steps S602 to S604, the local replication and multi-level distribution of data streams are performed on the base station side. Combined with the centralized identification and cache scheduling of the smart board, the transmission path is optimized. This simplifies the multiple transmissions of duplicate data in the backhaul network into the retrieval and local distribution of a single data set, significantly reducing backhaul bandwidth usage and improving the transmission efficiency of multi-user concurrent services. In turn, it solves the technical problem that the backhaul bandwidth cannot exceed the physical limit due to the lack of cross-base station collaborative identification and caching capabilities on the access network side.
[0110] According to embodiments of this application, a hotspot service flow identification device is provided. It should be noted that the hotspot service flow identification device of this application embodiment can be used to execute the hotspot service flow identification method provided in this application embodiment. The hotspot service flow identification device provided in this application embodiment is described below.
[0111] Figure 7 This is a structural diagram of a hotspot service flow identification device provided according to an embodiment of this application. Figure 7 As shown, the device includes:
[0112] Acquisition module 70 is used to acquire multidimensional features of the business flow;
[0113] The identification module 72 is used to identify multi-dimensional features using a hierarchical identification model and output the hotspot confidence level corresponding to the business flow. The hierarchical identification model includes: a first identification path based on preset rules for matching, a second identification path based on machine learning model for multi-dimensional feature analysis, and a third identification path based on feedback data for model parameter updating.
[0114] The determination module 74 is used to aggregate request patterns from multiple base stations, identify hotspot service flows based on hotspot confidence and request patterns, and store the hotspot service flows in a global cache pool. The request pattern reflects the degree of aggregation of group requests across multiple base stations.
[0115] Through the acquisition module, identification module, and determination module in the aforementioned hotspot service flow identification device, the goal of intelligently identifying hotspot service flows on the access network side and realizing cross-base station shared caching is achieved. This enables the technical effect of compressing backhaul bandwidth requirements from the user level to the base station level, breaking through the physical bandwidth limit of the transmission ring. Furthermore, it solves the technical problem that the backhaul bandwidth cannot break through the physical limit due to the lack of cross-base station collaborative identification and caching capabilities on the access network side in related technologies.
[0116] According to embodiments of this application, a hotspot service flow distribution device is also provided. It should be noted that the hotspot service flow distribution device of this application embodiment can be used to execute the hotspot service flow distribution method provided in this application embodiment. The hotspot service flow distribution device provided in this application embodiment is described below.
[0117] Figure 8 This is a structural diagram of a hotspot service flow distribution device provided according to an embodiment of this application. Figure 8 As shown, the device includes:
[0118] The receiving module 80 is used to receive hotspot service flows sent by the smart board, wherein the smart board is used to acquire multi-dimensional features of the service flows; a hierarchical recognition model is used to identify the multi-dimensional features and output the hotspot confidence level corresponding to the service flows; wherein the hierarchical recognition model includes: a first recognition path based on preset rules for matching, a second recognition path based on machine learning model for multi-dimensional feature analysis, and a third recognition path based on feedback data of the recognition results for updating model parameters; the request patterns of multiple base stations are aggregated, and hotspot service flows are determined based on hotspot confidence level and request patterns; wherein the request patterns are used to represent the degree of aggregation of group requests of the service flows within the range of multiple base stations.
[0119] Distribution module 82 is used to replicate hot service flows and distribute them to multiple user devices.
[0120] Through the receiving and distribution modules in the aforementioned hotspot service flow distribution device, local replication and multi-level distribution of data streams are achieved at the base station side. Combined with the centralized identification and cache scheduling of the smart board, the transmission path is optimized. This simplifies the multiple transmissions of duplicate data in the backhaul network into the retrieval and local distribution of a single data set, significantly reducing backhaul bandwidth usage and improving the transmission efficiency of multi-user concurrent services. Furthermore, it solves the technical problem that the backhaul bandwidth cannot exceed the physical limit due to the lack of cross-base station collaborative identification and caching capabilities on the access network side.
[0121] This application also provides an electronic device, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the above-described hotspot service flow identification or distribution method.
[0122] It should be noted that the aforementioned electronic equipment is used to perform Figure 2 The hotspot service flow identification method shown or Figure 6 The hotspot service flow distribution method shown above also applies to this electronic device, and will not be repeated here.
[0123] This application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device containing the non-volatile storage medium executes the above-mentioned hot spot service flow identification or distribution method by running the computer program.
[0124] It should be noted that the aforementioned non-volatile storage media is used for execution. Figure 2 The hotspot service flow identification method shown or Figure 6 The hot spot service flow distribution method shown above also applies to this non-volatile storage medium, and will not be repeated here.
[0125] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-described hotspot service flow identification or distribution method.
[0126] It should be noted that the above-mentioned computer program product is used to execute... Figure 2 The hotspot service flow identification method shown or Figure 6 The hot spot service flow distribution method shown above also applies to this computer program product, and will not be repeated here.
[0127] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0128] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0129] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.
[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0131] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0132] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.
[0133] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method of hotspot traffic identification, the method comprising: Applications in smart boards, including: Obtain multidimensional features of the business flow; A hierarchical identification model is used to identify the multidimensional features and output the hotspot confidence score corresponding to the business flow. The hierarchical identification model includes: a first identification path based on preset rules for matching, a second identification path based on machine learning model for multidimensional feature analysis, and a third identification path based on feedback data for model parameter updating. The request patterns of multiple base stations are aggregated, and hot service flows are determined based on the hotspot confidence and the request patterns. The hot service flows are then stored in a global cache pool. The request patterns are used to reflect the degree of aggregation of group requests of the service flows across multiple base stations.
2. The method of claim 1, wherein, The method further includes: Based on the request pattern, a multi-level distribution operation is performed on the hot service flows in the global cache pool, wherein the multi-level distribution operation includes: First-level distribution: Distributing the hotspot service streams to the multiple base stations; Secondary distribution involves replicating the received hotspot service streams from the multiple base stations and distributing them to multiple user devices.
3. The method of claim 1, wherein, The multidimensional features include at least two or more of the following: spatiotemporal features of the service flow, content features, user behavior features, network status features, historical pattern features, and service layer features.
4. The method of claim 1, wherein, The multidimensional features are identified using a hierarchical recognition model, including: In the first identification path, the multidimensional features are matched based on preset rules or business tags in the rule base. If the match is successful, the hotspot confidence score corresponding to the business flow is output. If the match fails or the hotspot confidence score is lower than the first threshold, the second identification path is entered. In the second identification path, a machine learning model is used to identify the multidimensional features to obtain an initial hotspot confidence score; if the initial hotspot confidence score is greater than or equal to a second threshold, the initial hotspot confidence score is determined as the hotspot confidence score corresponding to the business flow; if the initial hotspot confidence score is less than the second threshold, the process proceeds to the third identification path. In the third identification path, the business flow is marked as a sample to be learned, and the model parameters are updated in online or offline mode.
5. The method of claim 4, wherein, The method further includes: The actual popularity data of the business flow is obtained, and the actual popularity data is compared with the initial hot spot confidence score to generate a feedback sample, wherein the feedback sample is used to characterize the deviation between the model recognition result and the actual business needs; The performance metrics of the hierarchical recognition model are determined based on the feedback samples. If the performance index is lower than a preset performance threshold or the number of feedback samples reaches a preset number threshold, the model parameters in the hierarchical recognition model are incrementally updated.
6. The method of claim 1, wherein, Aggregating request patterns from multiple base stations and determining hotspot service flows based on the hotspot confidence level and the request patterns, including: When the hotspot confidence level is greater than a first preset hotspot threshold, multiple base stations' request records for the target service flow are obtained, and a request pattern for the corresponding base station is formed based on the request records. The target service flow is any one of the service flows, and the request records include at least one of the following: the total number of requests for the service flow within the ring network, the number of requesting base stations, and the request growth rate. The aggregation heat index of the multiple base stations is determined based on the request pattern, wherein the aggregation heat index is used to reflect the degree of aggregation of group requests of the service flow within the range of multiple base stations and the heat diffusion trend; If the aggregated popularity index is greater than the second preset popularity threshold, the target service flow is identified as a hot service flow, and a corresponding hot spot identifier is generated.
7. The method of claim 1, wherein, The method further includes: Based on the request growth rate or historical pattern characteristics of the hot service flow, predict the target base stations of the hot service flow that are to be distributed but have not yet generated a request pattern. Send a pre-caching instruction to the target base station, wherein the pre-caching instruction is used to instruct the target base station to prepare to receive hotspot service streams from the smart board, or to instruct the smart board to preload the hotspot service streams in the global cache pool in advance.
8. A method for hotspot traffic flow distribution, characterized by, Applied to base stations, including: The system receives hotspot service flows sent by a smart board, wherein the smart board is used to acquire multi-dimensional features of the service flows; it uses a hierarchical identification model to identify the multi-dimensional features and outputs a hotspot confidence score corresponding to the service flows, wherein the hierarchical identification model includes: a first identification path for matching based on preset rules, a second identification path for multi-dimensional feature analysis based on a machine learning model, and a third identification path for updating model parameters based on feedback data of the identification results; it aggregates request patterns from multiple base stations and determines hotspot service flows based on the hotspot confidence score and the request patterns, wherein the request patterns represent the degree of aggregation of group requests of the service flows within the range of multiple base stations; The hot service flow is copied and distributed to multiple user devices.
9. A hotspot traffic identification apparatus, characterized by comprising: include: The acquisition module is used to acquire multi-dimensional features of the business flow; The identification module is used to identify the multidimensional features using a hierarchical identification model and output the hotspot confidence level corresponding to the business flow. The hierarchical identification model includes: a first identification path based on preset rules for matching, a second identification path based on machine learning model for multidimensional feature analysis, and a third identification path based on feedback data for model parameter updating. The determination module is used to aggregate request patterns from multiple base stations, determine hotspot service flows based on the hotspot confidence level and the request patterns, and store the hotspot service flows in a global cache pool. The request patterns are used to reflect the degree of aggregation of group requests of the service flows within the scope of multiple base stations.
10. An electronic device, comprising: include: A memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the hotspot service flow identification method according to any one of claims 1 to 7 or the hotspot service flow distribution method according to claim 8.
