Intelligent data return method, apparatus and device, and medium

Abstract

The invention relates to the technical field of computer network communication, and discloses an intelligent data return method, which comprises the following steps of: acquiring an original index and local context information of a network state; the original indexes and the local context information are combined to construct a time sequence vector data set; obtaining a prediction result according to the time sequence vector data set; determining a return scheduling decision according to the predicted congestion mode result and the predicted transmission window result; and processing and transmitting the to-be-transmitted data according to the prediction result and the return scheduling decision. The method can actively predict the network trend, avoid the congestion time period, select the optimal window for data transmission, improve the transmission efficiency and success rate, have the adaptive learning ability, and adapt to the dynamic change of the network environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer network communication, and in particular to a method and device for intelligent data backhaul, equipment and storage medium. BACKGROUND

[0002] In the existing strategy of client backhaul data to central management equipment (such as internal network security hardware), the common methods include continuous streaming and fixed period batch transmission. However, these traditional strategies often have difficulty in achieving an ideal balance between optimizing network resource occupation and ensuring data timeliness. In order to improve this problem, some schemes attempt to make opportunistic transmission based on simple network probing (such as Ping delay), but these improved schemes still have the following deep limitations.

[0003] Firstly, these schemes lack multi-dimensional state awareness. They usually only rely on simple single-point indicators (such as delay) to evaluate network state, and cannot comprehensively depict the real state of the complex link between the client and a specific management device. In fact, the evaluation of network state needs to consider multiple factors such as jitter, packet loss rate, available bandwidth, etc., as well as the mutual influence between these factors. Relying only on the delay indicator cannot accurately reflect the comprehensive performance of the network, which may lead to inaccurate transmission decisions.

[0004] Secondly, the decision-making of these schemes is reactive rather than predictive. They usually make transmission decisions based on the current, instantaneous network state, and cannot foresee the upcoming network congestion or idle window. This reactive decision-making method may miss the best transmission opportunity. For example, starting transmission during a short "pseudo-idle" period of the network may encounter congestion during transmission, thereby affecting transmission efficiency and success rate.

[0005] Thirdly, these schemes fail to learn and utilize the patterns of network load. In the internal network environment, network load often presents periodic or event-based patterns (such as collective work, lunch break, work, scheduled backup, etc.). However, existing methods usually cannot learn these patterns and use them to guide transmission decisions. If these patterns can be learned and utilized, the network state can be better predicted, and more reasonable transmission decisions can be made.

[0006] Finally, the optimization granularity of these schemes is relatively rough. They usually use fixed threshold-based decision logic, which is too rigid and cannot adapt to the dynamic changes of the network environment. Under different network congestion levels, the optimal transmission strategy may be different. However, the fixed threshold-based decision-making method cannot learn and adapt to these changes, and thus cannot achieve fine-grained transmission optimization.

[0007] In summary, the existing client data back to the central management device strategy has many limitations. These limitations lead to a difficult balance between optimizing network resource occupation and ensuring data timeliness. Therefore, a new method is needed to overcome these limitations and achieve more intelligent and efficient data back. SUMMARY

[0008] The main purpose of the present application is to provide a method, device, equipment and storage medium for intelligent data back, aiming to solve the problem of balancing between optimizing network resource occupation and ensuring data timeliness in the prior art.

[0009] To achieve the above-mentioned purpose, the present application provides a method for intelligent data back, comprising: The client periodically probes the management hardware device to obtain the original index of the network state; According to the preset time interval, the local context information of the client is obtained; The original index and local context information are combined to construct a multi-dimensional feature vector, and the multi-dimensional feature vector is arranged in time sequence to form a time series vector dataset; The time series vector dataset is input into a prediction model for analysis, and the prediction result is output; the prediction result includes a predicted congestion mode result and a predicted transmission window result; According to the predicted congestion mode result and the predicted transmission window result, the back scheduling decision is determined; According to the prediction result and the back scheduling decision, the to-be-transmitted data is processed and transmitted.

[0010] In one embodiment, the client periodically probes the management hardware device to obtain the original index of the network state, comprising: The client sends a probe signal to the management hardware device according to a preset time interval; The management hardware device accepts the probe signal and responds, and feeds back the response signal to the client; The client obtains the original index of the network state according to the received feedback response signal.

[0011] In one embodiment, the time series vector dataset is input into a prediction model for analysis, and the prediction result is output; the prediction result includes a predicted congestion mode result and a predicted transmission window result, comprising: The time series vector dataset is input into a prediction model, and the prediction model analyzes the time series vector dataset; The prediction model predicts the network congestion mode within a preset time according to the original index of the time series vector dataset, and outputs a predicted congestion mode result; inputting the time series vector dataset into the prediction model, the prediction model identifying the time series vector dataset; The prediction model predicts the transmission window within a preset time according to the original index and local context information of the time series vector dataset, and outputs a predicted transmission window result.

[0012] In one embodiment, the backhaul scheduling decision is determined according to the predicted congestion pattern result and the predicted transmission window result, comprising: Obtain the to-be-transmitted data, prioritize and mark the to-be-transmitted data, and output the buffer state of the to-be-transmitted data; According to the predicted congestion pattern result, the predicted transmission window result and the buffer state, the backhaul scheduling decision of the to-be-transmitted data is determined.

[0013] In one embodiment, the backhaul scheduling decision of the to-be-transmitted data is determined according to the predicted congestion pattern result, the predicted transmission window result and the buffer state, comprising: According to the predicted congestion pattern result and the predicted transmission window, the transmission window and the transmission time are determined; According to the transmission window, the transmission time and the buffer state, the backhaul scheduling decision is determined; When the backhaul scheduling decision is immediate transmission, the data transmission task is executed immediately; When the backhaul scheduling decision is scheduling transmission, the transmission task is scheduled to the start time of the transmission window, and the state of waiting is maintained; When the backhaul scheduling decision is re-prediction, the prediction model is re-run to obtain the prediction result.

[0014] In one embodiment, before the time series vector dataset is input into the prediction model for analysis and the prediction result is output, the prediction result includes the predicted congestion pattern result and the predicted transmission window result, comprising: The client collects historical data, and constructs the historical data into a historical multi-dimensional feature vector; The historical multi-dimensional feature vector is arranged in time sequence to generate a historical time series vector dataset; The historical time series vector dataset is used to train the prediction model; After training, the prediction model parameters and the historical time series vector dataset are saved to the local storage.

[0015] In one embodiment, the to-be-transmitted data is processed and transmitted according to the prediction result and the backhaul scheduling decision, comprising: According to the prediction result, the to-be-transmitted data is preprocessed; The preprocessed to-be-transmitted data is encrypted; The encrypted to-be-transmitted data is transmitted according to the backhaul scheduling decision.

[0016] In one embodiment, to achieve the above object, the present application provides an intelligent data backhaul device, comprising: An original index data acquisition module is configured to periodically detect a management hardware device through a client to obtain original indexes of network states; A local context information acquisition module is configured to obtain local context information of the client according to a preset time interval; A vector construction module is configured to combine and construct the original indexes and the local context information into a multi-dimensional feature vector, and arrange the multi-dimensional feature vector in time sequence to form a time series vector dataset; A prediction module is configured to input the time series vector dataset into a prediction model for analysis, and output a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result; A decision and scheduling module is configured to determine a backhaul scheduling decision according to the predicted congestion mode result and the predicted transmission window result; A data transmission module is configured to process and transmit the to-be-transmitted data according to the backhaul scheduling decision.

[0017] In one embodiment, to achieve the above object, the present application further provides a computer device, which comprises a memory, a processor, and a program of intelligent data backhaul stored in the memory and executable on the processor; when the program of intelligent data backhaul is executed by the processor, the steps of the method of intelligent data backhaul are implemented.

[0018] In one embodiment, to achieve the above object, the present application further provides a computer readable storage medium, which stores a program of intelligent data backhaul; when the program of intelligent data backhaul is executed by a processor, the steps of the method of intelligent data backhaul are implemented.

[0019] Beneficial effects: The application relates to the technical field of computer network communication, and discloses a method for intelligent data backhaul, which comprises the following steps: periodically detecting a management hardware device through a client to obtain original indexes of network states; acquiring local context information of the client according to a preset time interval; combining the original indexes and the local context information to construct a multi-dimensional feature vector, and arranging the multi-dimensional feature vector in time sequence to form a time series vector dataset; inputting the time series vector dataset into a prediction model for analysis and outputting a prediction result; the prediction result comprises a predicted congestion mode result and a predicted transmission window result; determining a backhaul scheduling decision according to the predicted congestion mode result and the predicted transmission window result; and processing and transmitting the to-be-transmitted data according to the prediction result and the backhaul scheduling decision. The original indexes of network states and the local context information are periodically collected by the client agent, a multi-dimensional feature vector is constructed, and a time series vector dataset is formed. The future network congestion mode and the optimal transmission window are predicted through a prediction model. The data transmission strategy is determined according to the prediction result and the buffer state, the data backhaul task is intelligently scheduled. The method can actively predict the network trend, avoid the congestion period, select the optimal window for data transmission, improve the transmission efficiency and success rate, and has self-adaptive learning ability and can adapt to the dynamic change of the network environment. BRIEF DESCRIPTION OF DRAWINGS

[0020] The application will be further described below in combination with the drawings and embodiments, and the drawings show: Figure 1 An application environment schematic diagram of a method for intelligent data backhaul in an embodiment of the application; Figure 2 A flowchart of the method for intelligent data backhaul in an embodiment of the application; Figure 3 A functional module schematic diagram of a preferred embodiment of the device for intelligent data backhaul; Figure 4 A structure schematic diagram of a computer device in an embodiment of the application; Figure 5 Another structure schematic diagram of a computer device in an embodiment of the application. DETAILED DESCRIPTION

[0021] It should be understood that the specific embodiments described herein are merely intended to explain the application, and are not intended to limit the application.

[0022] The method for intelligent data backhaul provided in the embodiments of the application can be applied to, for example, Figure 1In an application environment, the user end communicates with the server end through the network. The server end can periodically detect the management hardware device through the client end to obtain the original indicators of the network status; obtain the local context information of the client end according to a preset time interval; combine the original indicators and local context information to construct a multidimensional feature vector, and arrange the multidimensional feature vector in chronological order to form a time series vector data set; input the time series vector data set into a prediction model for analysis, and output a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result; based on the predicted congestion mode result and the predicted transmission window result, a return scheduling decision is determined; and the data to be transmitted is processed and transmitted based on the prediction result and the return scheduling decision. The present invention periodically collects the original indicators and local context information of the network status through the client agent, constructs a multidimensional feature vector, and forms a time series vector data set. The future network congestion mode and the optimal transmission window are predicted by the prediction model. The data transmission strategy is determined based on the prediction result and the buffer status, and the data return task is intelligently scheduled. This method can proactively predict network trends, avoid congested periods, select the optimal window for data transmission, and improve transmission efficiency and success rate. It also possesses adaptive learning capabilities to adapt to dynamic changes in the network environment. The user end can include, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server end can be implemented as a standalone server or a server cluster consisting of multiple servers. The present invention is described in detail below using specific embodiments.

[0023] See also Figure 2 , Figure 2 This is a flow chart of an embodiment of a method for intelligent data transmission provided by the present invention. It should be noted that although a logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0024] like Figure 2 As shown, the method of intelligent data return proposed by the present invention includes the following steps: S100: The client periodically detects the management hardware device to obtain original indicators of the network status.

[0025] In this embodiment, the client agent proactively and periodically sends a probe signal to a specific management hardware device to obtain raw network status indicators. These indicators include RTT, jitter, packet loss rate, and estimated available bandwidth. This allows the system to understand the network link status between the client and the management device in real time.

[0026] S200: Obtain local context information of the client according to a preset time interval.

[0027] In this embodiment, in the network communication system, the client Agent plays a crucial role, not only collecting network status indicators, but also collecting local context information that may affect network status, thereby providing more comprehensive and accurate data support for intelligent scheduling and prediction of the network. The client Agent obtains the local context information of the client device at preset time intervals (e.g., every minute, every 5 minutes, etc.). These information mainly include the current time, user active state and the running state of specific high-bandwidth applications. These factors have a significant impact on network status, and by collecting these information, the client Agent can more comprehensively assess the actual use of the network.

[0028] Specifically, the current time is the system time of the client device, including the specific hour, minute and day of the week, etc. Network status often shows periodic changes with time. For example, during working hours (such as 9 am to 5 pm on Monday to Friday), network load is usually high because most users are using the network for work-related activities such as file transfer, video conference, etc. During non-working hours (such as weekends or late at night), network load may be low because users' demand for network use decreases. In addition, in some specific time periods, such as lunch break time (12 pm to 1 pm) or pre-off work peak (17 pm to 18 pm), network congestion may occur. By recording the current time, the client Agent can identify these time periods and predict changes in network status based on historical data, thereby providing a basis for the rational allocation of network resources.

[0029] The user active state is the current activity state of the user on the client device, indicating whether the user is currently using the client device and the type of activity the user is performing. User behavior can significantly affect the occupation of network resources. For example, if the user is watching high-definition video, conducting video conference or downloading large files, these activities will occupy a large amount of network bandwidth, causing network congestion. Conversely, if the user is only browsing the web or using lightweight applications, the network load may be low. If the user is not using the client device, the network resources may be relatively idle, suitable for data transfer. The client Agent can obtain the user active state in several ways. One way is through system event listening, listening to system events (such as screen lock, screen unlock, application startup and shutdown, etc.) to determine whether the user is using the device. For example, when the screen is unlocked, it can be considered that the user starts using the device; when the screen is locked, it can be considered that the user temporarily leaves the device. Another way is through application monitoring, monitoring the currently running applications and their network usage to identify the running state of high-bandwidth applications. For example, if a video conference software is detected to be running, it can be judged that the user is performing high-bandwidth activities.

[0030] Specific high-bandwidth applications refer to the status of high-bandwidth applications running on the client device, i.e. those applications that may occupy a large amount of network bandwidth, such as video conferencing software (such as Zoom, Teams), streaming applications (such as Netflix, YouTube), file download tools, etc. The running status of these applications directly affects the available bandwidth of the network. For example, if the user is using video conferencing software, the network bandwidth will be heavily occupied, causing the network performance of other applications to decline. If the user is downloading a large file, the network bandwidth will also be occupied, affecting other data transmission tasks. The client Agent can obtain the running status of specific high-bandwidth applications in multiple ways. One way is through process monitoring, monitoring the process list in the system to identify whether high-bandwidth applications are running. For example, by checking whether there is a process of Zoom or Teams in the system process list, it can be determined whether the video conferencing software is running. Another way is through network traffic analysis, by analyzing network traffic to identify the data transmission characteristics of high-bandwidth applications to determine whether they are occupying network resources. For example, by analyzing the packet characteristics in network traffic, the data transmission mode of streaming applications can be identified to determine whether streaming applications are running.

[0031] The client Agent can more comprehensively evaluate the network status by collecting local context information (such as current time, user activity status, running status of specific high-bandwidth applications, etc.). These context information together with network status indicators provide richer data support for intelligent scheduling and prediction. By integrating these information, the system can better predict network congestion patterns, identify the optimal transmission window, and thus improve the efficiency and success rate of data transmission. For example, during working hours, if the client Agent detects that the user is running high-bandwidth video conferencing software and the current time is in a period of high network load, the system can predict a high probability of network congestion. At this time, the system can adjust the data transmission strategy to prioritize the smoothness of video conferencing, while postponing other non-critical data transmission tasks to a period of low network load. During non-working hours, if the client Agent detects that the user is not using the device and the current time is in a period of low network load, the system can consider that the network resources are relatively idle and suitable for large-scale data transmission tasks. In this way, the client Agent can dynamically adjust the allocation of network resources and data transmission strategies according to local context information and network status indicators, thereby optimizing network performance and improving user experience.

[0032] S300, combine the original indicators and local context information to construct a multi-dimensional feature vector, and arrange the multi-dimensional feature vector in time sequence to form a time series vector dataset.

[0033] In this embodiment, the network probe indicators and context information collected at the same time point are combined into a multi-dimensional feature vector. For example, V(t) = [rtt(t), jitter(t), loss(t), bandwidth_est(t), time_feature(t), activity_state(t),...]. These vectors are arranged in chronological order to form a time series vector dataset, providing a data basis for subsequent analysis and prediction. At the same time point, the client agent combines the network probe indicators and context information collected into a multi-dimensional feature vector. The client agent periodically collects network probe indicators and context information in chronological order and combines these information into multi-dimensional feature vectors. These vectors are arranged in chronological order to form a time series vector dataset. The time series vector dataset provides a data basis for subsequent analysis and prediction. The client agent collects network probe indicators and context information at the same time point and combines these information into a multi-dimensional feature vector. These vectors are arranged in chronological order to form a time series vector dataset. The time series vector dataset provides rich data support for subsequent analysis and prediction, helping the system to realize intelligent data transmission scheduling and network congestion avoidance.

[0034] S400, inputting the time series vector dataset into a prediction model for analysis, and outputting a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result.

[0035] In this embodiment, in network state prediction and intelligent scheduling, the combination of time series prediction model and pattern recognition and clustering model provides strong support for data transmission decision. The main task of the time series prediction model is to analyze the historical network state indicators in the time series vector dataset and predict the network congestion pattern in the future period. Specifically, the model receives a time series vector dataset containing network state indicators (such as bandwidth utilization, delay, packet loss rate, etc.) as input. Through the analysis of these historical data, the model can learn the law of network state change over time, and then predict the congestion degree that the network may appear in the future period, such as "deep congestion" or "mild congestion". This prediction ability enables the system to understand the trend of network congestion in advance, providing an important basis for subsequent data transmission decisions. The main task of the pattern recognition and clustering model is to analyze the historical network state indicators and context information (such as current time, user activity state, high-bandwidth application running state, etc.) in the time series vector dataset, and identify the patterns and state transitions of network state. The model divides the network state into different patterns through clustering algorithm and analyzes the transition rules between these patterns. Based on these analysis results, the model can predict the optimal transmission window in the future period, i.e. identify the period when the network state is good and suitable for data transmission. For example, the model can find that the network load is low and user activity is less in some period, and thus recommend these periods as the preferred choice for data transmission.

[0036] By combining the time series prediction model and the pattern recognition and clustering model, the system can more comprehensively predict the future network state. The time series prediction model focuses on the prediction of network congestion patterns, helping the system to avoid potential network congestion risks in advance; while the pattern recognition and clustering model focuses on identifying the optimal transmission window, providing the best opportunity for data transmission. This combination overcomes the limitations of traditional methods that only make decisions based on the current instantaneous network state, and uses machine learning techniques such as time series prediction, pattern recognition or reinforcement learning to analyze the historical sequence of network state vectors in depth, thus more accurately predicting future network conditions or identifying the optimal transmission pattern.

[0037] In summary, the synergy of these two models enables the system to understand the trend of network state change in advance, accurately identify the optimal transmission window, achieve intelligent data transmission scheduling and network congestion avoidance, and significantly improve the utilization efficiency of network resources and the reliability of data transmission.

[0038] S500, determining a backhaul scheduling decision according to the predicted congestion pattern result and the predicted transmission window result.

[0039] In this embodiment, the optimal timing and strategy for data transmission are determined based on the predicted congestion pattern results and the predicted transmission window results. For example, data return is selected to be performed within the predicted optimal transmission window, avoiding transmission during the predicted congestion period. If the decision result is transmission, there can be two cases. One is that the current time is the start time of the predicted optimal window (or very close), in which case the transmission will be prepared and executed immediately; the other is that the predicted optimal window starts at a certain time point in the future, in which case the transmission task will be scheduled to be executed at that time point, and the current waiting will be maintained. If there is no suitable window in the near future, or there is no data to be transmitted in the buffer, the system will continue to wait and re-predict. The client agent dynamically adjusts the decision logic according to the real-time network state and buffer state. For example, if the network state changes significantly during the waiting period, the client agent can re-evaluate the prediction results and adjust the transmission strategy. If the amount of data in the buffer or the priority of the data changes during the waiting period, the client agent can re-evaluate the transmission strategy and adjust the scheduling plan. The optimal transmission window decision and scheduling module is one of the core innovations of the client agent. It integrates the output results of the prediction module and the state of the local data buffer to dynamically make the final data return scheduling decision. The module can intelligently select the optimal transmission timing according to the results output by different prediction models (time series prediction, clustering / state model, reinforcement learning model) and the buffer state, ensuring the efficiency and reliability of data transmission.

[0040] S600, processing and transmitting the data to be transmitted according to the prediction results and the return scheduling decision.

[0041] In this embodiment, the system analyzes and predicts the historical sequence of network state vectors through a machine learning model, and predicts the future network congestion pattern and the optimal transmission window in advance. Based on these prediction results, the system can actively and intelligently schedule the return task of local encrypted cache data, and select the best time window for data transmission without passive waiting for the current network state to meet the conditions.

[0042] The system dynamically adjusts the data processing strategy according to the predicted network quality. Specifically, it optimizes data selection, prioritizes high-priority data transmission; uses appropriate compression algorithms to reduce data volume; strengthens data encryption to ensure information security; adjusts transmission rate according to network conditions to avoid congestion; and develops flexible retry strategies to deal with transmission failures. Through the comprehensive use of these strategies, the system can ensure the efficiency, security and reliability of data transmission, and improve overall performance and user experience.

[0043] During data transmission, the client Agent dynamically adjusts the transmission strategy based on real-time network status and buffer status. If there is a significant change in network status during the waiting period (such as sudden network congestion or improvement), the Agent can re-evaluate the prediction results and adjust the transmission strategy. If the amount of data in the buffer increases or the data priority changes, the Agent can re-evaluate the transmission strategy and adjust the scheduling plan. Through this dynamic adjustment mechanism, the client Agent can flexibly respond to changes in network status, ensuring the efficiency and reliability of data transmission.

[0044] In one embodiment, the step S100 comprises: S101, sending a probe signal to the management hardware device according to a preset time interval by the client; S102, the management hardware device accepts the probe signal and responds, and feeds back the response signal to the client; S103, the client acquires the original indicators of network status according to the received feedback response signal.

[0045] In this embodiment, the client Agent is a program running on the client device, and its core function is to monitor and evaluate the network link status between the client and a specific management hardware device (such as an internal network security hardware). By actively sending probe signals and receiving responses, the client Agent can obtain real-time data of network status, providing basic support for subsequent intelligent data backhaul and network congestion avoidance.

[0046] The client Agent actively sends probe signals to the management hardware device according to a preset time interval (for example, every minute, every 5 minutes, etc.). These signals are usually lightweight network requests, such as ICMP (Internet Control Message Protocol) Ping requests, TCP probe packets or UDP probe packets. The purpose of the probe signal is to trigger the response of the management hardware device, so as to obtain the relevant data of network status.

[0047] When the management hardware device receives the probe signal, it will return a response signal. The client Agent acquires the original indicators of network status by receiving this response signal. These indicators include RTT (Round Trip Time), Jitter, Packet Loss Rate and Estimated Available Bandwidth, etc. Key parameters, which can fully reflect the real-time status of network link.

[0048] Specifically, RTT (Round Trip Time) refers to the time interval from the client sending a probe signal to receiving the management hardware device's response. The client Agent records the timestamp of sending the probe signal and the timestamp of receiving the response. RTT reflects the delay of the network link. A lower RTT indicates a smaller network delay and faster data transmission speed; a higher RTT indicates a larger network delay, which may affect the real-time performance of data transmission.

[0049] Jitter refers to the degree of variation between consecutive RTT measurements. It reflects the stability of network delay. The client Agent sends probe signals at multiple time points and measures RTT, calculating the standard deviation or variance of these RTT values. Lower jitter indicates that the network delay is more stable, suitable for applications with high real-time requirements (such as video conferencing, voice calls, etc.); higher jitter may cause unstable data transmission, affecting user experience.

[0050] Packet Loss Rate refers to the proportion of probe signals sent within a certain period of time that fail to receive a response. The client Agent sends N probe signals within a period of time and records the number of responses M. A lower packet loss rate indicates that the network link is more reliable, and the success rate of data transmission is higher; a higher packet loss rate may indicate that the network is congested or faulty, and measures (such as retransmission mechanism) need to be taken to ensure data integrity.

[0051] Estimated Available Bandwidth refers to the maximum data transmission rate that can be used between the client and the management hardware device under current network conditions. The client Agent can estimate the available bandwidth by time-based methods, sending a certain amount of data within a short period of time, measuring the time required to transmit these data, and calculating the transmission rate. The client Agent can estimate the available bandwidth by packet loss rate-based methods, gradually increasing the data transmission rate until the packet loss rate significantly increases, and the transmission rate at this time is the estimated value of the available bandwidth. Higher available bandwidth indicates that the network link can support more data transmission, suitable for large file transfer or high-bandwidth applications; lower available bandwidth may need to compress or batch data transmission to avoid congestion.

[0052] By periodically sending probe signals and obtaining the above network status indicators, the client Agent can understand the real-time network link status between the client and the management hardware device. These indicators provide real-time network link status information for the system and are the basis for intelligent data backhaul and network congestion avoidance.

[0053] In summary, the client agent can obtain raw indicators of network status (such as RTT, jitter, packet loss rate, available bandwidth, etc.) by actively and periodically sending probe signals and receiving responses. These indicators not only provide real-time status information of network links for the system, but also provide important decision-making basis for intelligent data backhaul and network congestion avoidance. In this way, the client agent can optimize the data transmission process to ensure efficient, secure and reliable transmission of data.

[0054] In one embodiment, the step S400 comprises: S401, inputting the time series vector dataset into a prediction model, the prediction model analyzing the time series vector dataset; S401, the prediction model predicting network congestion patterns within a preset time according to the raw indicators of the time series vector dataset, and outputting a predicted congestion pattern result; S401, inputting the time series vector dataset into the prediction model, the prediction model identifying the time series vector dataset; S401, the prediction model predicting a transmission window within a preset time according to the raw indicators of the time series vector dataset and local context information, and outputting a predicted transmission window result.

[0055] In this embodiment, the system analyzes and predicts network status through three main machine learning models: time series prediction model, pattern recognition and clustering model, and reinforcement learning model, thereby realizing intelligent data transmission scheduling and network congestion avoidance. These models analyze historical data of network status, predict future network status, and optimize data transmission strategies accordingly, significantly improving the efficiency and reliability of data transmission.

[0056] The core task of the time series prediction model is to analyze the trend of network status indicators over time and predict future network status. In this embodiment, a variant of recurrent neural network, LSTM (Long Short-Term Memory), is used to achieve this goal. LSTM network can capture long-term dependencies of network status, thereby accurately predicting future network congestion patterns. Specifically, the system inputs the historical sequence of multi-dimensional feature vectors into the time series prediction model. These feature vectors contain key indicators of network status. Based on these historical data, the model learns the dynamic change law of network status and predicts the network congestion pattern in the future period. For example, if the model detects that RTT and packet loss rate show an upward trend in certain time periods, it can predict that network congestion may occur in these time periods.

[0057] The main task of the pattern recognition and clustering model is to identify the patterns of network states and state transitions, so as to predict the optimal transmission window. In this embodiment, the K-Means clustering algorithm is used to cluster the network state vectors into different states, such as "heavy congestion", "light congestion", "stable smooth", and "high instability". The model learns the transition probabilities and typical durations between these states by analyzing historical data. Specifically, the system inputs the historical sequence of multi-dimensional feature vectors into the pattern recognition and clustering model. The model analyzes these historical data to identify the patterns and changes of network states, especially those "low tide periods" (i.e. periods of good network state and low congestion level) that are suitable for data transmission. For example, the model can identify that during certain time periods, the network delay is low, the packet loss rate is low, and the available bandwidth is high, which are considered as optimal transmission windows. By identifying the optimal transmission windows, the system can efficiently complete the data transmission task during the periods of good network state, thereby improving the success rate and efficiency of data transmission.

[0058] The main task of the reinforcement learning model is to learn the optimal transmission strategy, which directly outputs the optimal action (transmit or wait) in the current state. In this embodiment, an RL Agent is constructed, whose state space includes the current network state vector, buffer state, etc., and the action space is "transmit (optional parameters)" or "wait". The RL Agent learns a policy function by interacting with the environment (actual network probing and transmission results). This function directly outputs the optimal action according to the current state vector, with the goal of maximizing the long-term cumulative reward. The reward function is designed to reward successful, efficient, and low-interference transmission. For example, if transmission is chosen in a certain state and the task is successfully completed, the Agent will receive a positive reward; if transmission is chosen during a congested period and results in failure, the Agent will be punished. Through the reinforcement learning model, the system can dynamically adjust the transmission strategy according to the real-time network state and buffer state, ensuring data transmission under optimal conditions.

[0059] The time series vector dataset is input into the prediction model for analysis, and the model can output prediction results, including the predicted network congestion pattern within a predetermined time and the predicted optimal transmission window within a predetermined time. These prediction results provide decision-making basis for intelligent data backhaul and network congestion avoidance, helping the system to choose the best transmission opportunity and improve the efficiency and success rate of data transmission.

[0060] In one embodiment, the step S500 comprises: S501, acquiring the data to be transmitted, prioritizing and marking the data to be transmitted, and outputting the buffer state of the data to be transmitted; S502, determining the backhaul scheduling decision of the data to be transmitted according to the predicted congestion pattern result, the predicted transmission window result, and the buffer state.

[0061] In this embodiment, before data transmission, the client agent first classifies and prioritizes the data in the buffer. These data may be marked and sorted according to their priority, type, size, etc. For example, high-priority real-time data (such as data in video conferencing) may be given priority, while low-priority log files may be scheduled for transmission at a later time. Prioritize data in the buffer according to the importance and timeliness of the data.

[0062] Data priority: (1) High priority: data with high real-time requirements (such as video conferencing, voice calls, etc.).

[0063] (2) Medium priority: important but less time-sensitive data (such as file transfer tasks).

[0064] (3) Low priority: data with low real-time requirements (such as log files, backup data, etc.).

[0065] High-priority data needs to be transmitted as soon as possible to ensure real-time performance; low-priority data can be transmitted during poor network conditions or off-peak hours to avoid interfering with the transmission of high-priority data.

[0066] Data size: according to the predicted duration of the transmission window and network quality, select the appropriate data size for transmission.

[0067] (1) If the predicted window is short or the network quality is poor, select a smaller data block for transmission.

[0068] (2) If the predicted window is long and the network quality is good, select a larger data block for transmission.

[0069] By reasonably selecting the data size, network resources can be maximized while avoiding transmission failures or network congestion caused by excessive data size.

[0070] Further, the client agent determines the final data transmission strategy based on the predicted network congestion pattern and the predicted transmission window, combined with the buffer state (such as whether there is data to be transmitted, the priority of the data, the size of the data, etc.).

[0071] The specific prediction congestion pattern is: (1) Deep congestion period: the predicted result shows that the network delay is high, the packet loss rate is high, and the available bandwidth is low.

[0072] (2) Mild congestion period: the predicted result shows that the network delay is moderate, the packet loss rate is low, and the available bandwidth is moderate.

[0073] (3) Non-congestion period: a period where the network delay is low, the packet loss rate is low, and the available bandwidth is high.

[0074] By identifying these congestion periods, the client Agent can avoid data transmission during congestion periods, thereby reducing the risk of transmission failure and transmission delay.

[0075] The client Agent first determines which data in the buffer needs to be transmitted, and classifies and sorts the data according to its priority, type, size, etc. According to the output results of the time series prediction model, pattern recognition and clustering model and reinforcement learning model, the client Agent predicts the future network state, identifies the congestion period and the optimal transmission window. Combined with the buffer state and the prediction result, the client Agent determines the final data transmission strategy, including selecting the appropriate data volume, data priority sorting and transmission opportunity. The client Agent calls the data processing and reliable transmission module to process the data and execute the transmission task in the predicted optimal transmission window.

[0076] This embodiment realizes comprehensive analysis and prediction of network state through the intelligent scheduling mechanism of the client Agent, optimizes the timing and method of data transmission. Through dynamic adjustment of transmission strategy, the client Agent can efficiently complete data transmission tasks at the right time, avoid network congestion, and ensure the success rate and efficiency of data transmission. This intelligent scheduling method based on prediction not only improves the reliability of data transmission, but also optimizes the utilization of network resources, and improves user experience.

[0077] In one embodiment, the step S501 specifically comprises: S5021, determining the transmission window and transmission time according to the predicted congestion pattern result and predicted transmission window; S5022, determining the backhaul scheduling decision according to the transmission window, transmission time and buffer state; S5023, when the backhaul scheduling decision is to execute transmission immediately, executing data transmission task immediately; S5024, when the backhaul scheduling decision is to schedule transmission, scheduling the transmission task to the start time of the transmission window and keeping the waiting state; S5025, when the backhaul scheduling decision is to re-predict, re-running the prediction model to obtain the prediction result.

[0078] In this embodiment, the client Agent determines the appropriate data transmission time based on the predicted network congestion pattern and the predicted transmission window. The predicted congestion pattern result provides detailed information about the future network status, such as which time periods are likely to experience deep congestion (high latency, high packet loss rate, low available bandwidth), which time periods are likely to experience light congestion (moderate latency, low packet loss rate, moderate available bandwidth), and which time periods have good network status (low latency, low packet loss rate, high available bandwidth). Based on this information, the Agent can identify the time window in the future that is suitable for data transmission, i.e., the predicted transmission window.

[0079] The transmission time, on the other hand, refers to the specific time point within these predicted transmission windows that the Agent chooses to start executing the data transmission task. For example, if the predicted transmission window is from 10 PM to 2 AM, the Agent may choose to start transmission at 10:30 PM, depending on the buffer status and other factors.

[0080] The client Agent combines the transmission window, transmission time, and buffer status to determine the final backhaul scheduling decision. The buffer status includes information such as whether there is data to be transmitted in the buffer, the priority of the data, the size of the data, etc. The Agent makes one of the following three decisions based on these comprehensive factors: If the current time is close to the start time of the predicted optimal transmission window (or is already within the window), and there is data to be transmitted in the buffer, the Agent will make a "perform transmission immediately" decision. In this case, the Agent will immediately invoke the data processing and reliable transmission module to start executing the data transmission task. For example, if the predicted optimal transmission window is from 10 PM to 2 AM, and the current time is 10:15 PM, the Agent will immediately start transmitting high-priority data.

[0081] If the predicted optimal transmission window starts at a future time point, the Agent will make a "schedule transmission" decision. In this case, the Agent will schedule the transmission task to the start time of the predicted transmission window and remain in a waiting state until the scheduled transmission time arrives. For example, if the predicted optimal transmission window is from 10 PM to 2 AM, and the current time is 8 PM, the Agent will schedule the transmission task to 10 PM and remain in a waiting state at the current time.

[0082] If there is no suitable transmission window in the near future or no data to be transmitted in the buffer, the Agent will make a "re-prediction" decision. In this case, the Agent will re-run the prediction model to obtain new prediction results. For example, if the current prediction result shows that the network status is not ideal in the next few hours, the Agent may wait for a period of time and then re-run the prediction model to obtain more accurate future network status information.

[0083] During the execution of the above decision-making process, the client Agent dynamically adjusts the decision-making logic according to the real-time network status and buffer status. For example, if the network status changes significantly during the waiting period (such as sudden network congestion or improvement), the Agent can re-evaluate the prediction results and adjust the transmission strategy. If the amount of data in the buffer increases or the data priority changes, the Agent can also re-evaluate the transmission strategy and adjust the scheduling plan.

[0084] Through this dynamic adjustment mechanism, the client Agent can flexibly respond to changes in network status, ensuring the efficiency and reliability of data transmission. At the same time, the Agent can adjust the transmission strategy in real time through the real-time feedback mechanism to adapt to the changing network environment.

[0085] Through the combination of congestion pattern prediction results and predicted transmission windows, the client Agent can intelligently determine transmission windows and transmission times. Based on this information and buffer status, the Agent can make "immediate execution of transmission", "scheduled transmission" or "re-prediction" return scheduling decisions. This intelligent scheduling based on prediction not only improves the efficiency and reliability of data transmission, but also optimizes the utilization of network resources and improves user experience.

[0086] In one embodiment, before the step S400, the following steps are included: S401, collecting historical data by the client, and constructing the historical data into a historical multi-dimensional feature vector; S402, arranging the historical multi-dimensional feature vector in chronological order to generate a historical time series vector dataset; S403, training the prediction model using the historical time series vector dataset; S404, saving the prediction model parameters and the historical time series vector dataset to the local storage after training is completed.

[0087] In this embodiment, the client agent is first responsible for collecting various historical data related to network status. The client agent continuously collects these data at preset time intervals (e.g., every minute, every 5 minutes, etc.) to ensure the completeness and timeliness of the data. The collected historical data is organized by the client agent into the form of a multi-dimensional feature vector. Each feature vector contains network status indicators and contextual information collected at a specific time point. The client agent arranges the constructed historical multi-dimensional feature vectors in chronological order to generate a historical time series vector dataset. This dataset is an ordered sequence, where each feature vector corresponds to a specific time point. The generated historical time series vector dataset is used to train the prediction model. The prediction model can be a time series prediction model (such as an LSTM network), a pattern recognition and clustering model (such as a K-Means clustering algorithm), or a reinforcement learning model, etc. During the training process, the model learns the patterns and rules in the historical data, such as the periodic changes in network status, the characteristics of congestion periods, etc. Taking the LSTM network as an example, the model learns the long-term dependencies in the historical time series vector dataset to predict the future trend of network status changes. During the training process, the model continuously adjusts its parameters to minimize prediction errors and improve prediction accuracy.

[0088] Further, after training is completed, the client agent saves the parameters of the prediction model and the historical time series vector dataset to the local storage. These saved contents will be used for subsequent network status prediction and data transmission scheduling. The prediction model parameters define the structure and weights of the model, which are the key to the model's ability to make predictions. After saving these parameters, the client agent can directly load the model in subsequent runs without retraining, thereby improving the efficiency of the system. The saved historical dataset can be used for subsequent model evaluation, updating, or further training. For example, if the system collects new data, the new data can be added to the historical dataset, and the model can be retrained to improve its prediction performance.

[0089] Through the above steps, the client agent realizes the complete process from historical data collection to prediction model training and saving. This process provides the system with powerful prediction capabilities, enabling the client agent to intelligently predict future network status based on historical data and optimize data transmission strategies accordingly. By saving the prediction model parameters and historical dataset, the system can efficiently perform subsequent prediction and scheduling tasks, ensuring the efficiency and reliability of data transmission.

[0090] In one embodiment, the step S600 comprises: S601, preprocessing the to-be-transmitted data according to the prediction result; S602, performing encryption processing on the preprocessed data to be transmitted; S603, performing encrypted transmission on the encrypted data to be transmitted according to the backhaul scheduling decision.

[0091] In this embodiment, before data transmission, the client agent first preprocesses the data to be transmitted in the buffer according to the predicted network state and transmission window. The purpose of preprocessing is to optimize the efficiency and success rate of data transmission. Preprocessing includes data compression, which reduces the bandwidth and time required for data transmission and improves transmission efficiency. Data blocking divides large files or large amounts of data into smaller data blocks, facilitating transmission within the predicted transmission window and reducing the risk of transmission failure due to network problems. According to the predicted transmission window and network quality, the data is divided into data blocks of appropriate size. After completing the preprocessing, the client agent encrypts the data to ensure the security of data transmission. After completing the preprocessing and encryption of the data, the client agent transmits the encrypted data according to the backhaul scheduling decision.

[0092] In an embodiment, an intelligent data backhaul device is provided, which corresponds one-to-one to the intelligent data backhaul method in the above embodiment. Referring to Figure 3 , Figure 3 The functional module diagram of a preferred embodiment of the intelligent data backhaul device of the present application. The original index data acquisition module 10, the local context information acquisition module 20, the vector construction module 30, the prediction module 40, the decision and scheduling module 50 and the data transmission module 60. The detailed description of each functional module is as follows: The original index data acquisition module 10 is used to periodically detect the management hardware through the client to obtain the original index of the network state; The local context information acquisition module 20 is used to obtain the local context information of the client according to the preset time interval; The vector construction module 30 is used to combine and construct the original index and local context information into a multi-dimensional feature vector, and arrange the multi-dimensional feature vector in time sequence to form a time series vector data set; The prediction module 40 is used to input the time series vector data set into the prediction model for analysis, and output the prediction result; the prediction result includes the predicted congestion mode result and the predicted transmission window result; The decision and scheduling module 50 is used to determine the backhaul scheduling decision according to the predicted congestion mode result and the predicted transmission window result; The data transmission module 60 is used to process and transmit the data to be transmitted according to the backhaul scheduling decision.

[0093] In one embodiment, the original index data collection module 10 comprises: a probe signal unit for sending a probe signal to the management hardware device according to a preset time interval by the client; a signal response unit for the management hardware device accepting the probe signal and responding, and feeding back a response signal to the client; an original index unit for the client obtaining the original index of the network state according to the received feedback response signal.

[0094] In one embodiment, the prediction module 40 comprises: a data analysis unit for inputting the time series vector dataset into the prediction model, and the prediction model analyzing the time series vector dataset; a congestion prediction unit for the prediction model predicting the network congestion mode within a preset time according to the original index of the time series vector dataset, and outputting a predicted congestion mode result; a data recognition unit for inputting the time series vector dataset into the prediction model, and the prediction model recognizing the time series vector dataset; a window prediction unit for the prediction model predicting the transmission window within a preset time according to the original index of the time series vector dataset and the local context information, and outputting a predicted transmission window result.

[0095] In one embodiment, the decision and scheduling module 50 comprises: a priority sorting unit for obtaining the to-be-transmitted data, sorting and marking the to-be-transmitted data according to priority, and outputting the buffer state of the to-be-transmitted data; a decision determination unit for determining the backhaul scheduling decision of the to-be-transmitted data according to the predicted congestion mode result, the predicted transmission window result, and the buffer state.

[0096] In one embodiment, the backhaul scheduling decision module specifically comprises: a window and time determination unit for determining the transmission window and the transmission time according to the predicted congestion mode result and the predicted transmission window; a backhaul scheduling determination unit for determining the backhaul scheduling decision according to the transmission window, the transmission time, and the buffer state; an immediate execution unit for executing the data transmission task immediately when the backhaul scheduling decision is to execute the transmission immediately; a scheduling transmission unit for scheduling the transmission task to the start time of the transmission window and keeping the state of waiting when the backhaul scheduling decision is to schedule the transmission; a re-prediction unit for re-running the prediction model to obtain the prediction result when the backhaul scheduling decision is to re-predict.

[0097] In one embodiment, the model training module specifically comprises: a data collection unit configured to collect historical data by the client, and construct the historical data into a historical multi-dimensional feature vector; a historical data arrangement unit configured to arrange the historical multi-dimensional feature vector in time sequence to generate a historical time series vector dataset; a model training unit configured to train the prediction model using the historical time series vector dataset; a data saving unit configured to save the prediction model parameters and the historical time series vector dataset to a local storage after the training is completed.

[0098] In one embodiment, the data transmission module 60 comprises: a data processing unit configured to pre-process the to-be-transmitted data according to the prediction result; a data encryption unit configured to encrypt the pre-processed to-be-transmitted data; an encrypted transmission unit configured to perform encrypted transmission on the encrypted to-be-transmitted data according to the backhaul scheduling decision.

[0099] In one embodiment, a computer device is provided, which can be a server, and an internal structure diagram thereof can be as shown in Figure 4 The computer device comprises a processor, a memory, a network interface and a database connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile and / or volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is configured to communicate with an external client through a network connection. The computer program is executed by the processor to implement the functions or steps of the server side of the intelligent data backhaul method.

[0100] In one embodiment, a computer device is provided, which can be a client, and an internal structure diagram thereof can be as shown in Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with the external server through the network connection. The computer program is executed by the processor to realize the functions or steps of the method for intelligent data backhaul on the user side In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the computer program to implement the following steps: Periodically probe the management hardware device through the client to obtain original indicators of network status; Obtain local context information of the client according to a preset time interval; Combine and construct the original indicators and the local context information into a multi-dimensional feature vector, and arrange the multi-dimensional feature vector in time sequence to form a time series vector dataset; Input the time series vector dataset into a prediction model for analysis, and output a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result; Determine a backhaul scheduling decision according to the predicted congestion mode result and the predicted transmission window result; Process and transmit the to-be-transmitted data according to the prediction result and the backhaul scheduling decision.

[0101] In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, the computer program being executed by a processor to implement the following steps: Periodically probe the management hardware device through the client to obtain original indicators of network status; Obtain local context information of the client according to a preset time interval; Combine and construct the original indicators and the local context information into a multi-dimensional feature vector, and arrange the multi-dimensional feature vector in time sequence to form a time series vector dataset; Input the time series vector dataset into a prediction model for analysis, and output a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result; Determine a backhaul scheduling decision according to the predicted congestion mode result and the predicted transmission window result; Process and transmit the to-be-transmitted data according to the prediction result and the backhaul scheduling decision.

[0102] It should be noted that the above functions or steps that can be achieved by the computer readable storage medium or the computer device can correspond to the related description of the server side and the user side in the foregoing method embodiments, and will not be described here to avoid repetition.

[0103] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM) and the like.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is exemplified, and in actual application, the above-mentioned functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.

[0105] It should be noted that if a non-company software tool or component appears in the embodiments of the present application, it is only used for example introduction and does not represent actual use. The above-described embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for intelligent data transmission, characterized in that: The following steps are involved: The client periodically detects the management hardware device to obtain the original indicators of the network status; Obtaining local context information of the client at a preset time interval; Combining the original indicators and local context information into a multidimensional feature vector, and arranging the multidimensional feature vector in chronological order to form a time series vector dataset; Inputting the time series vector data set into a prediction model for analysis and outputting a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result; Determining a backhaul scheduling decision based on the predicted congestion mode result and the predicted transmission window result; The data to be transmitted is processed and transmitted according to the prediction result and the return scheduling decision.

2. The method for intelligent data transmission according to claim 1, wherein: The client periodically detects the management hardware device to obtain the original indicators of the network status, including: The client sends a detection signal to the management hardware device according to a preset time interval; The management hardware device receives the detection signal and responds, and feeds the response signal back to the client; The client obtains the original indicator of the network status according to the response signal received as feedback.

3. The method for intelligent data transmission according to claim 1, wherein: The time series vector data set is input into the prediction model for analysis, and a prediction result is output; the prediction result includes a predicted congestion mode result and a predicted transmission window result, including: Inputting the time series vector dataset into a prediction model, and the prediction model analyzing the time series vector dataset; The prediction model predicts the network congestion pattern within a preset time based on the original indicators of the time series vector data set, and outputs the predicted congestion pattern result; Inputting the time series vector dataset into the prediction model, and the prediction model identifying the time series vector dataset; The prediction model predicts a transmission window within a preset time based on the original indicators of the time series vector data set and local context information, and outputs a predicted transmission window result.

4. The method for intelligent data transmission according to claim 1, wherein: The determining of a backhaul scheduling decision according to the predicted congestion mode result and the predicted transmission window result includes: Acquire data to be transmitted, prioritize and mark the data to be transmitted, and output a buffer status of the data to be transmitted; A backhaul scheduling decision for the data to be transmitted is determined based on the predicted congestion mode result, the predicted transmission window result, and the buffer state.

5. The method for intelligent data transmission according to claim 4, wherein: The determining of a backhaul scheduling decision for the data to be transmitted according to the predicted congestion mode result, the predicted transmission window result, and the buffer state includes: Determining a transmission window and a transmission time according to the predicted congestion mode result and the predicted transmission window; Determining a backhaul scheduling decision based on the transmission window, transmission time, and buffer status; When the backhaul scheduling decision is to execute the transmission immediately, the data transmission task is executed immediately; When the return scheduling decision is to schedule transmission, the transmission task is scheduled to the start time of the transmission window and remains in a waiting state; When the feedback scheduling decision is a re-prediction, the prediction model is re-run to obtain a prediction result.

6. The method for intelligent data transmission according to claim 1, wherein: The step of inputting the time series vector data set into a prediction model for analysis and outputting a prediction result, wherein the prediction result includes a prediction congestion mode result and a prediction transmission window result, including: The client collects historical data and constructs the historical data into a historical multi-dimensional feature vector; Arranging the historical multidimensional feature vectors in chronological order to generate a historical time series vector dataset; Using the historical time series vector dataset to train the prediction model; After training is completed, the prediction model parameters and historical time series vector dataset are saved to local storage.

7. The method for intelligent data transmission according to claim 1, wherein: The processing and transmitting the data to be transmitted according to the prediction result and the backhaul scheduling decision includes: Preprocessing the data to be transmitted according to the prediction result; Encrypting the pre-processed data to be transmitted; The encrypted data to be transmitted is encrypted and transmitted according to the return scheduling decision.

8. An intelligent data return device, characterized in that: The intelligent data transmission device includes: The original indicator data collection module is used to periodically detect the management hardware device through the client to obtain the original indicators of the network status; The context information collection module is used to obtain the local context information of the client according to the preset time interval; A vector construction module, configured to combine the original indicators and local context information into a multidimensional feature vector, and arrange the multidimensional feature vectors in chronological order to form a time series vector dataset; A prediction module, configured to input the time series vector data set into a prediction model for analysis and output a prediction result; the prediction result includes a predicted congestion mode result and a predicted transmission window result; A decision and scheduling module, configured to determine a backhaul scheduling decision based on the predicted congestion mode result and the predicted transmission window result; The data transmission module is used to process and transmit the data to be transmitted according to the backhaul scheduling decision.

9. A computer device, characterized in that: The computer device includes a memory, a processor, and an intelligent data return program stored in the memory and capable of running on the processor. When the intelligent data return program is executed by the processor, the steps of the intelligent data return method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The storage medium stores a program for intelligent data transmission, and when the program for intelligent data transmission is executed by the processor, the steps of the method for intelligent data transmission as described in any one of claims 1 to 7 are implemented.

Images