IPv6 traffic prediction system based on machine learning

Abstract

The invention relates to the technical field of traffic prediction, and discloses an IPv6 traffic prediction system based on machine learning, and the system comprises a traffic collection module which is used for capturing the original traffic data of an IPv6 network interface and carrying out the protocol analysis and traffic classification preprocessing, and obtaining the preprocessed IPv6 traffic data; the multi-dimensional feature extraction module is used for performing multi-modal feature extraction on the preprocessed IPv6 flow data to generate a fusion feature vector; the real-time prediction module is used for inputting the fusion feature vector into a dynamic prediction model to obtain an IPv6 traffic prediction value of a future preset duration; and the dynamic scheduling module is used for pushing the IPv6 traffic prediction value of the future preset duration to a traffic scheduling controller so as to realize dynamic scheduling of network resources. According to the method, the network traffic prediction accuracy and real-time performance are improved, the adaptive capacity to IPv6 traffic mode change is enhanced, and reasonable distribution and efficient utilization of network resources are effectively promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic prediction, and in particular to an IPv6 traffic prediction system based on machine learning. Background Art

[0002] In today's digital age, accurate network traffic prediction is crucial for the rational allocation and efficient utilization of network resources. IPv6 (Internet Protocol Version 6), the sixth version of the Internet Protocol, features complex and volatile traffic. Traditional traffic prediction methods struggle to meet the precise prediction requirements in an IPv6 environment. Existing network traffic prediction technologies often suffer from insufficient prediction accuracy, poor real-time performance, and limited adaptability to changing traffic patterns when processing large-scale, high-speed IPv6 data flows.

[0003] Therefore, it is urgent to provide a technical solution to solve the above problems. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides an IPv6 traffic prediction system based on machine learning.

[0005] In a first aspect, the present invention provides an IPv6 traffic prediction system based on machine learning, the technical solution of the system is as follows: The traffic collection module is used to capture the original traffic data of the IPv6 network interface in real time through distributed probe nodes, and perform protocol analysis and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data; A multi-dimensional feature extraction module is used to extract multi-modal features from the pre-processed IPv6 traffic data to generate a fused feature vector; The real-time prediction module is used to input the fused feature vector into a dynamic prediction model to obtain the IPv6 traffic prediction value for a preset time in the future; the expression of the dynamic prediction model is: ;in, Indicates the preset duration in the future. represents the predicted IPv6 traffic value for the preset future time period, Represents the dynamic prediction model; the dynamic prediction model is: a machine learning model based on a lightweight inference engine deployment, combining a gated recurrent unit and a bidirectional recurrent neural network; Represents a multidimensional feature extraction function; Representing the pre-processed IPv6 traffic data; The dynamic scheduling module is used to push the IPv6 traffic prediction value of the future preset time period to the traffic scheduling controller to realize dynamic scheduling of network resources.

[0006] The beneficial effects of the machine learning-based IPv6 traffic prediction system of the present invention are as follows: The system of the present invention can solve the problems of low prediction accuracy, poor real-time performance and weak adaptability of traditional methods in IPv6 environment, improve the accuracy and real-time performance of network traffic prediction, enhance the adaptability to changes in IPv6 traffic patterns, and effectively assist in the rational allocation and efficient utilization of network resources.

[0007] Based on the above solution, the machine learning-based IPv6 traffic prediction system of the present invention can also be improved as follows.

[0008] Furthermore, the original traffic data includes: an IPv6 basic header and at least one extended header; the traffic collection module is specifically configured to: Parsing the IPv6 basic header to obtain a flow label field, and parsing each extended header to obtain a security encapsulation header type and routing extension header information; Extracting a protocol type identifier and routing path parameters according to the flow label field, the security encapsulation header type, and the routing extension header information; According to the protocol type identifier, IPv6 data packets belonging to the same service type in the original traffic data are aggregated into service type subsets, and data packet length normalization and timestamp alignment processing are performed on each service type subset to generate the preprocessed IPv6 traffic data.

[0009] Furthermore, the multidimensional feature extraction module is specifically used to: Extracting the pre-processed IPv6 traffic data through a sliding time window to obtain a statistical feature vector; Processing the pre-processed IPv6 traffic data based on the protocol type identifier and the routing path parameters to generate a protocol stack feature vector; Extracting the pre-processed IPv6 traffic data using discrete wavelet transform to obtain a multi-scale time series feature vector; After normalizing the statistical feature vector, the protocol stack feature vector, and the multi-scale time series feature vector, the fused feature vector is generated by vector concatenation. The expression of the fused feature vector is: ;in, Indicates the size of the sliding time window, represents the flow value of the i-th time window, represents the statistical eigenvector, represents the protocol stack feature vector, represents the one-hot encoding function, represents the multi-scale time series feature vector, represents discrete wavelet transform, represents the flow label field, represents the vector outer product operation, Indicates the routing path parameters.

[0010] Furthermore, the dynamic scheduling module is specifically used to: The predicted IPv6 traffic value for the future preset time period is pushed to the traffic scheduling controller, so that the traffic scheduling controller generates a bandwidth allocation instruction based on the predicted IPv6 traffic value for the future preset time period to achieve dynamic scheduling of network resources; the expression of the bandwidth allocation instruction is: ;in, represents the bandwidth allocation value of the future preset time period, Indicates the current bandwidth utilization. Represents the predicted value weight coefficient.

[0011] Furthermore, it also includes: An online incremental learning module is configured to dynamically update the parameters of the dynamic prediction model based on a time-decay weighting mechanism, wherein the time-decay weighting mechanism assigns training weights according to the timestamps of the preprocessed IPv6 traffic data, so that the dynamic prediction model adapts to IPv6 traffic feature drift; and the expression for dynamically updating the parameters of the dynamic prediction model is: , ;in, represents the parameters of the dynamic prediction model at time t, represents the learning rate of the dynamic prediction model, represents the loss function, express The time decay weight, represents the attenuation coefficient, Representation sample timestamp, Indicates the current timestamp, The dynamic prediction model The actual flow value of the sample, The dynamic prediction model The predicted flow value of samples.

[0012] Furthermore, the dynamic scheduling module is also used to: Before generating the bandwidth allocation instruction, the following anomaly detection logic is executed: when hour, ;in, Represents the historical traffic mean, represents the historical flow standard deviation, represents the abnormal threshold, Indicates the burst traffic expansion coefficient, .

[0013] Furthermore, the attenuation coefficient of the time attenuation weighting mechanism is dynamically adjusted according to the flow rate change rate, and the expression is: ;in, represents the basic attenuation coefficient, represents the first-order derivative of the flow prediction value over time, represents the hyperbolic tangent function.

[0014] In a second aspect, the present invention provides an IPv6 traffic prediction method based on machine learning, the technical solution of the method is as follows: The original traffic data of the IPv6 network interface is captured in real time by distributed probe nodes, and the original traffic data is preprocessed by protocol analysis and traffic classification to obtain preprocessed IPv6 traffic data; Performing multimodal feature extraction on the preprocessed IPv6 traffic data to generate a fusion feature vector; The fused feature vector is input into a dynamic prediction model to obtain the IPv6 traffic prediction value for a preset time in the future; the expression of the dynamic prediction model is: ;in, Indicates the preset duration in the future. represents the predicted IPv6 traffic value for the preset future time period, Represents the dynamic prediction model; the dynamic prediction model is: a machine learning model based on a lightweight inference engine deployment, combining a gated recurrent unit and a bidirectional recurrent neural network; Represents a multidimensional feature extraction function; Representing the pre-processed IPv6 traffic data; The predicted IPv6 traffic value of the future preset time period is pushed to the traffic scheduling controller to realize dynamic scheduling of network resources.

[0015] The beneficial effects of the IPv6 traffic prediction method based on machine learning of the present invention are as follows: The method of the present invention can solve the problems of low prediction accuracy, poor real-time performance and weak adaptability of traditional methods in IPv6 environment, improve the accuracy and real-time performance of network traffic prediction, enhance the adaptability to changes in IPv6 traffic patterns, and effectively contribute to the rational allocation and efficient utilization of network resources.

[0016] Based on the above scheme, the IPv6 traffic prediction method based on machine learning of the present invention can also be improved as follows.

[0017] Furthermore, the original traffic data includes: an IPv6 basic header and at least one extended header; the step of performing protocol parsing and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data further includes: Parsing the IPv6 basic header to obtain a flow label field, and parsing each extended header to obtain a security encapsulation header type and routing extension header information; Extracting a protocol type identifier and routing path parameters according to the flow label field, the security encapsulation header type, and the routing extension header information; According to the protocol type identifier, IPv6 data packets belonging to the same service type in the original traffic data are aggregated into service type subsets, and data packet length normalization and timestamp alignment processing are performed on each service type subset to generate the preprocessed IPv6 traffic data.

[0018] Furthermore, the step of performing multimodal feature extraction on the pre-processed IPv6 traffic data to generate a fusion feature vector further includes: Extracting the pre-processed IPv6 traffic data through a sliding time window to obtain a statistical feature vector; Processing the pre-processed IPv6 traffic data based on the protocol type identifier and the routing path parameters to generate a protocol stack feature vector; Extracting the pre-processed IPv6 traffic data using discrete wavelet transform to obtain a multi-scale time series feature vector; After normalizing the statistical feature vector, the protocol stack feature vector, and the multi-scale time series feature vector, the fused feature vector is generated by vector concatenation. The expression of the fused feature vector is: ;in, Indicates the size of the sliding time window, represents the flow value of the i-th time window, represents the statistical eigenvector, represents the protocol stack feature vector, represents the one-hot encoding function, represents the multi-scale time series feature vector, represents discrete wavelet transform, represents the flow label field, represents the vector outer product operation, Indicates the routing path parameters.

[0019] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0021] The accompanying drawings are only used to illustrate the embodiments and are not to be considered as limiting the present invention. In addition, the same reference symbols are used to represent the same components throughout the drawings. In the drawings: Figure 1 A schematic diagram of the structure of an embodiment of an IPv6 traffic prediction system based on machine learning of the present invention; Figure 2 The present invention is a flowchart of an embodiment of an IPv6 traffic prediction method based on machine learning. DETAILED DESCRIPTION

[0022] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0023] Figure 1 FIG1 shows a schematic diagram of an embodiment of an IPv6 traffic prediction system based on machine learning provided by the present invention. Figure 1 As shown, the system includes: a traffic collection module 110, a multi-dimensional feature extraction module 120, a real-time prediction module 130 and a dynamic scheduling module 140; The traffic collection module 110 is used to capture the original traffic data of the IPv6 network interface in real time through distributed probe nodes, and perform protocol analysis and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data; A multi-dimensional feature extraction module 120 is used to perform multi-modal feature extraction on the pre-processed IPv6 traffic data to generate a fusion feature vector; The real-time prediction module 130 is used to input the fused feature vector into a dynamic prediction model to obtain the IPv6 traffic prediction value for a preset time period in the future; the expression of the dynamic prediction model is: ;in, Indicates the preset duration in the future. represents the predicted IPv6 traffic value for the preset future time period, Represents the dynamic prediction model; the dynamic prediction model is: a machine learning model based on a lightweight inference engine deployment, combining a gated recurrent unit and a bidirectional recurrent neural network; Represents a multidimensional feature extraction function; Representing the pre-processed IPv6 traffic data; The dynamic scheduling module 140 is used to push the IPv6 traffic prediction value of the future preset time period to the traffic scheduling controller to achieve dynamic scheduling of network resources.

[0024] It should be noted that distributed probe nodes are data collection agents deployed at key links in an IPv6 network. They use mirrored ports or traffic replication to capture raw data packets flowing through network interfaces in real time. These nodes work collaboratively in a distributed architecture to ensure comprehensive collection of high-speed IPv6 traffic and achieve load balancing through consistent hashing. An IPv6 network interface is a physical or virtual interface on a network device that supports the IPv6 protocol stack and is used to send and receive IPv6 packets compliant with RFC 2460. IPv6 network interfaces include router 10 Gigabit Ethernet ports and virtual switch VXLAN tunnel endpoints. Their traffic carrying capacity must meet line-rate processing requirements exceeding 100 Gbps. Raw traffic data refers to the collection of unprocessed IPv6 packets captured by distributed probe nodes. It includes the basic IPv6 header (source / destination address, flow label, hop limit) and at least one extension header (such as a routing header and a security encapsulation header). Raw traffic data is stored in PCAP format, preserving complete protocol layer information and timestamp metadata.

[0025] In an optional manner, the traffic collection module 110 is specifically configured to: The IPv6 basic header is parsed to obtain a flow label field, and each extended header is parsed to obtain a security encapsulation header type and routing extended header information.

[0026] The protocol type identifier and routing path parameters are extracted according to the flow label field, the security encapsulation header type and the routing extension header information.

[0027] According to the protocol type identifier, IPv6 data packets belonging to the same service type in the original traffic data are aggregated into service type subsets, and data packet length normalization and timestamp alignment processing are performed on each service type subset to generate the preprocessed IPv6 traffic data.

[0028] It should be noted that the flow label field is a 20-bit field in the IPv6 basic header that identifies packets within the same service flow, ensuring that network devices can differentiate between specific service types (such as video streaming and real-time communication). The security encapsulation header type refers to the field in the IPv6 extension header that identifies the data encryption and authentication protocol type, including ESP (Encapsulating Security Payload, IPsec encryption) and AH (Authentication Header, IPsec integrity check). The routing extension header information refers to the cross-domain transmission path parameters recorded in the IPv6 routing extension header (type 0, RFC 8200), including the list of intermediate nodes, the number of remaining hops, and the routing type identifier. The protocol type identifier is a service type code generated by parsing the flow label field and the security encapsulation header, which is used to distinguish traffic from different application scenarios (such as video streaming, encrypted control signaling, and IoT sensor data). Routing path parameters are quantitative indicators extracted from the routing extension header information that describe the characteristics of the traffic transmission path, including the number of cross-domain hops, the frequency of path node changes, and the routing type (strict / loose routing). A service type subset refers to a collection of IPv6 data packets aggregated according to protocol type identifiers. Traffic within the same subset has the same service attributes and traffic patterns (for example, a video stream subset contains long packet lengths and periodic transmission characteristics).

[0029] In an optional manner, the multidimensional feature extraction module 120 is specifically configured to: Extracting the pre-processed IPv6 traffic data through a sliding time window to obtain a statistical feature vector; Processing the pre-processed IPv6 traffic data based on the protocol type identifier and the routing path parameters to generate a protocol stack feature vector; Extracting the pre-processed IPv6 traffic data using discrete wavelet transform to obtain a multi-scale time series feature vector; After normalizing the statistical feature vector, the protocol stack feature vector, and the multi-scale time series feature vector, the fused feature vector is generated by vector concatenation. The expression of the fused feature vector is: ;in, Indicates the size of the sliding time window, represents the flow value of the i-th time window, represents the statistical eigenvector, represents the protocol stack feature vector, represents the one-hot encoding function, represents the multi-scale time series feature vector, represents discrete wavelet transform, represents the flow label field, represents the vector outer product operation, Indicates the routing path parameters.

[0030] In an optional manner, the dynamic scheduling module 140 is specifically configured to: The predicted IPv6 traffic value for the future preset time period is pushed to the traffic scheduling controller, so that the traffic scheduling controller generates a bandwidth allocation instruction based on the predicted IPv6 traffic value for the future preset time period to achieve dynamic scheduling of network resources; the expression of the bandwidth allocation instruction is: ;in, represents the bandwidth allocation value of the future preset time period, Indicates the current bandwidth utilization. Represents the predicted value weight coefficient.

[0031] In an optional manner, the method further includes: An online incremental learning module is configured to dynamically update the parameters of the dynamic prediction model based on a time-decay weighting mechanism, wherein the time-decay weighting mechanism assigns training weights according to the timestamps of the preprocessed IPv6 traffic data, so that the dynamic prediction model adapts to IPv6 traffic feature drift; and the expression for dynamically updating the parameters of the dynamic prediction model is: , ;in, represents the parameters of the dynamic prediction model at time t, represents the learning rate of the dynamic prediction model, represents the loss function, express The time decay weight, represents the attenuation coefficient, Representation sample timestamp, Indicates the current timestamp, The dynamic prediction model The actual flow value of the sample, The dynamic prediction model The predicted flow value of samples.

[0032] In an optional manner, the dynamic scheduling module 140 is further configured to: Before generating the bandwidth allocation instruction, the following anomaly detection logic is executed: when hour, ;in, Represents the historical traffic mean, represents the historical flow standard deviation, represents the abnormal threshold, Indicates the burst traffic expansion coefficient, .

[0033] In an optional manner, the attenuation coefficient of the time attenuation weighting mechanism is dynamically adjusted according to the flow rate change rate, and the expression is: ;in, represents the basic attenuation coefficient, represents the first-order derivative of the flow prediction value over time, represents the hyperbolic tangent function.

[0034] The technical solution of this embodiment can solve the problems of low prediction accuracy, poor real-time performance, and weak adaptability of traditional methods in the IPv6 environment, improve the accuracy and real-time performance of network traffic prediction, enhance the adaptability to changes in IPv6 traffic patterns, and effectively assist in the rational allocation and efficient utilization of network resources.

[0035] Figure 2 FIG1 shows a flow chart of an embodiment of an IPv6 traffic prediction method based on machine learning provided by the present invention. Figure 2 As shown, the following steps are included: S1. Capturing the original traffic data of the IPv6 network interface in real time through distributed probe nodes, and performing protocol parsing and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data; S2. Performing multimodal feature extraction on the pre-processed IPv6 traffic data to generate a fusion feature vector; S3. Input the fused feature vector into a dynamic prediction model to obtain a predicted value of IPv6 traffic for a preset time in the future; the expression of the dynamic prediction model is: ;in, Indicates the preset duration in the future. represents the predicted IPv6 traffic value for the preset future time period, Represents the dynamic prediction model; the dynamic prediction model is: a machine learning model based on a lightweight inference engine deployment, combining a gated recurrent unit and a bidirectional recurrent neural network; Represents a multidimensional feature extraction function; Representing the pre-processed IPv6 traffic data; S4. Push the IPv6 traffic prediction value of the future preset time period to the traffic scheduling controller to achieve dynamic scheduling of network resources.

[0036] In an optional manner, the original traffic data includes: an IPv6 basic header and at least one extended header; the step of performing protocol parsing and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data further includes: Parsing the IPv6 basic header to obtain a flow label field, and parsing each extended header to obtain a security encapsulation header type and routing extension header information; Extracting a protocol type identifier and routing path parameters according to the flow label field, the security encapsulation header type, and the routing extension header information; According to the protocol type identifier, IPv6 data packets belonging to the same service type in the original traffic data are aggregated into service type subsets, and data packet length normalization and timestamp alignment processing are performed on each service type subset to generate the preprocessed IPv6 traffic data.

[0037] In an optional manner, the step of performing multimodal feature extraction on the preprocessed IPv6 traffic data to generate a fused feature vector further includes: Extracting the pre-processed IPv6 traffic data through a sliding time window to obtain a statistical feature vector; Processing the pre-processed IPv6 traffic data based on the protocol type identifier and the routing path parameters to generate a protocol stack feature vector; Extracting the pre-processed IPv6 traffic data using discrete wavelet transform to obtain a multi-scale time series feature vector; After normalizing the statistical feature vector, the protocol stack feature vector, and the multi-scale time series feature vector, the fused feature vector is generated by vector concatenation. The expression of the fused feature vector is: ;in, Indicates the size of the sliding time window, represents the flow value of the i-th time window, represents the statistical eigenvector, represents the protocol stack feature vector, represents the one-hot encoding function, represents the multi-scale time series feature vector, represents discrete wavelet transform, represents the flow label field, represents the vector outer product operation, Indicates the routing path parameters.

[0038] The technical solution of this embodiment can solve the problems of low prediction accuracy, poor real-time performance, and weak adaptability of traditional methods in the IPv6 environment, improve the accuracy and real-time performance of network traffic prediction, enhance the adaptability to changes in IPv6 traffic patterns, and effectively assist in the rational allocation and efficient utilization of network resources.

[0039] Furthermore, the above embodiments provide systems that implement their functions using only the division of the above functional modules as examples. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. Furthermore, the systems and method embodiments provided in the above embodiments share the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0040] The above description is merely a preferred embodiment of the present invention and an illustration of the underlying technical principles. Those skilled in the art should understand that the scope of the present invention is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned concepts. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in this invention.

[0041] It should be noted that the terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects and to define a specific order or precedence. Where appropriate, the order used for similar objects may be interchanged, such that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.

[0042] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A machine learning-based IPv6 traffic prediction system, characterized in that: The system comprises: The traffic collection module is used to capture the original traffic data of the IPv6 network interface in real time through distributed probe nodes, and perform protocol analysis and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data; A multi-dimensional feature extraction module is used to extract multi-modal features from the pre-processed IPv6 traffic data to generate a fused feature vector; The real-time prediction module is used to input the fused feature vector into a dynamic prediction model to obtain the IPv6 traffic prediction value for a preset time in the future; the expression of the dynamic prediction model is: ;in, Indicates the preset duration in the future. represents the predicted IPv6 traffic value for the preset future time period, Represents the dynamic prediction model; the dynamic prediction model is: a machine learning model based on a lightweight inference engine deployment, combining a gated recurrent unit and a bidirectional recurrent neural network; Represents a multidimensional feature extraction function; Representing the pre-processed IPv6 traffic data; The dynamic scheduling module is used to push the IPv6 traffic prediction value of the future preset time period to the traffic scheduling controller to realize dynamic scheduling of network resources.

2. The machine learning-based IPv6 traffic prediction system according to claim 1, characterized in that: The original traffic data includes: an IPv6 basic header and at least one extended header; the traffic collection module is specifically used to: Parsing the IPv6 basic header to obtain a flow label field, and parsing each extended header to obtain a security encapsulation header type and routing extension header information; Extracting a protocol type identifier and routing path parameters according to the flow label field, the security encapsulation header type, and the routing extension header information; According to the protocol type identifier, IPv6 data packets belonging to the same service type in the original traffic data are aggregated into service type subsets, and data packet length normalization and timestamp alignment processing are performed on each service type subset to generate the preprocessed IPv6 traffic data.

3. The machine learning-based IPv6 traffic prediction system according to claim 2, characterized in that: The multidimensional feature extraction module is specifically used for: Extracting the pre-processed IPv6 traffic data through a sliding time window to obtain a statistical feature vector; Processing the pre-processed IPv6 traffic data based on the protocol type identifier and the routing path parameters to generate a protocol stack feature vector; Extracting the pre-processed IPv6 traffic data using discrete wavelet transform to obtain a multi-scale time series feature vector; After normalizing the statistical feature vector, the protocol stack feature vector, and the multi-scale time series feature vector, the fused feature vector is generated by vector concatenation. The expression of the fused feature vector is: ;in, Indicates the size of the sliding time window, represents the flow value of the i-th time window, represents the statistical eigenvector, represents the protocol stack feature vector, represents the one-hot encoding function, represents the multi-scale time series feature vector, represents discrete wavelet transform, represents the flow label field, represents the vector outer product operation, Indicates the routing path parameters.

4. The machine learning-based IPv6 traffic prediction system according to claim 1, wherein: The dynamic scheduling module is specifically used for: The predicted IPv6 traffic value for the future preset time period is pushed to the traffic scheduling controller, so that the traffic scheduling controller generates a bandwidth allocation instruction based on the predicted IPv6 traffic value for the future preset time period to achieve dynamic scheduling of network resources; the expression of the bandwidth allocation instruction is: ;in, represents the bandwidth allocation value of the future preset time period, Indicates the current bandwidth utilization. Represents the predicted value weight coefficient.

5. The machine learning-based IPv6 traffic prediction system according to any one of claims 1 to 4, characterized in that: Also includes: An online incremental learning module is configured to dynamically update the parameters of the dynamic prediction model based on a time-decay weighting mechanism, wherein the time-decay weighting mechanism assigns training weights according to the timestamps of the preprocessed IPv6 traffic data, so that the dynamic prediction model adapts to IPv6 traffic feature drift; and the expression for dynamically updating the parameters of the dynamic prediction model is: ， in, represents the parameters of the dynamic prediction model at time t, represents the learning rate of the dynamic prediction model, represents the loss function, express The time decay weight, represents the attenuation coefficient, Representation sample timestamp, Indicates the current timestamp, The dynamic prediction model The actual flow value of the sample, The dynamic prediction model The predicted flow value of samples.

6. The machine learning-based IPv6 traffic prediction system according to claim 4, characterized in that: The dynamic scheduling module is also used to: Before generating the bandwidth allocation instruction, the following anomaly detection logic is executed: when hour, ;in, Represents the historical traffic mean, represents the historical flow standard deviation, represents the abnormal threshold, Indicates the burst traffic expansion coefficient, .

7. The machine learning-based IPv6 traffic prediction system according to claim 5, characterized in that: The attenuation coefficient of the time attenuation weighting mechanism is dynamically adjusted according to the flow rate change rate, and the expression is: ;in, represents the basic attenuation coefficient, represents the first-order derivative of the flow prediction value over time, represents the hyperbolic tangent function.

8. A method for predicting IPv6 traffic based on machine learning, characterized in that: The method comprises: The original traffic data of the IPv6 network interface is captured in real time by distributed probe nodes, and the original traffic data is preprocessed by protocol analysis and traffic classification to obtain preprocessed IPv6 traffic data; Performing multimodal feature extraction on the preprocessed IPv6 traffic data to generate a fusion feature vector; The fused feature vector is input into a dynamic prediction model to obtain the IPv6 traffic prediction value for a preset time in the future; the expression of the dynamic prediction model is: ;in, Indicates the preset duration in the future. represents the predicted IPv6 traffic value for the preset future time period, Represents the dynamic prediction model; the dynamic prediction model is: a machine learning model based on a lightweight inference engine deployment, combining a gated recurrent unit and a bidirectional recurrent neural network; Represents a multidimensional feature extraction function; Representing the pre-processed IPv6 traffic data; The predicted IPv6 traffic value of the future preset time period is pushed to the traffic scheduling controller to realize dynamic scheduling of network resources.

9. The IPv6 traffic prediction method based on machine learning according to claim 8, characterized in that: The original traffic data includes: an IPv6 basic header and at least one extended header; the step of performing protocol parsing and traffic classification preprocessing on the original traffic data to obtain preprocessed IPv6 traffic data further includes: Parsing the IPv6 basic header to obtain a flow label field, and parsing each extended header to obtain a security encapsulation header type and routing extension header information; Extracting a protocol type identifier and routing path parameters according to the flow label field, the security encapsulation header type, and the routing extension header information; According to the protocol type identifier, IPv6 data packets belonging to the same service type in the original traffic data are aggregated into service type subsets, and data packet length normalization and timestamp alignment processing are performed on each service type subset to generate the preprocessed IPv6 traffic data.

10. The IPv6 traffic prediction method based on machine learning according to claim 9, characterized in that: The step of performing multimodal feature extraction on the pre-processed IPv6 traffic data to generate a fusion feature vector further includes: Extracting the pre-processed IPv6 traffic data through a sliding time window to obtain a statistical feature vector; Processing the pre-processed IPv6 traffic data based on the protocol type identifier and the routing path parameters to generate a protocol stack feature vector; Extracting the pre-processed IPv6 traffic data using discrete wavelet transform to obtain a multi-scale time series feature vector; After normalizing the statistical feature vector, the protocol stack feature vector, and the multi-scale time series feature vector, the fused feature vector is generated by vector concatenation. The expression of the fused feature vector is: ;in, Indicates the size of the sliding time window, represents the flow value of the i-th time window, represents the statistical eigenvector, represents the protocol stack feature vector, represents the one-hot encoding function, represents the multi-scale time series feature vector, Represents discrete wavelet transform.

Images