A large model-based intelligent heterogeneous network fusion optimization method and device

Abstract

The application discloses an intelligent heterogeneous network fusion optimization method and device based on a large model, wherein the method comprises the following steps: collecting real-time network state data, service demand data and user equipment data of various heterogeneous networks and preprocessing; extracting key features capable of reflecting real-time changes of network state and service demand from the preprocessed data; adopting an adaptive multi-modal fusion algorithm to automatically identify the modal characteristics of the heterogeneous network data and dynamically adjust the weight and parameters in the fusion process to achieve the optimal fusion effect; combining a cross-network context perception optimization strategy to realize the construction of a deep learning driven large model; and dynamically selecting the best network path and transmission strategy according to the output result of the large model by the intelligent fusion module. The method and device realize the intelligent fusion of the network based on the deep learning large model, the adaptive multi-modal fusion algorithm and the cross-network context perception optimization strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of network convergence, and in particular to an intelligent heterogeneous network convergence optimization method and device based on a large model. BACKGROUND

[0002] With the rapid development of information technology, various heterogeneous networks (such as wide area networks, wireless local area networks, etc.) coexist, each with its unique advantages and limitations. Traditional network convergence methods often use fixed fusion strategies, which cannot be dynamically adjusted according to real-time network status and service requirements. Therefore, how to realize intelligent network convergence to fully utilize the advantages of various networks and improve overall network performance has become a problem to be solved. SUMMARY

[0003] To solve the above problems existing in traditional network convergence methods, the present application provides an intelligent heterogeneous network convergence optimization method and device based on a large model, which builds a deep learning large model that can analyze the network status and service requirements of different types of networks in real time and develop intelligent fusion strategies accordingly to realize intelligent network convergence; uses adaptive multi-modal fusion algorithm and cross-network context perception optimization strategy to fully utilize the advantages of various networks to achieve more efficient and flexible network performance optimization, thereby improving cross-network data transmission efficiency and reducing network latency.

[0004] To achieve the above purpose, the present application adopts the following technical solutions:

[0005] In an embodiment of the present application, an intelligent heterogeneous network convergence optimization method based on a large model is proposed, which comprises:

[0006] Collecting real-time network status data, service requirement data and user equipment data of various heterogeneous networks and preprocessing them;

[0007] Extracting key features from the preprocessed data that can reflect real-time changes in network status and service requirements; using adaptive multi-modal fusion algorithm to automatically identify the modal characteristics of heterogeneous network data and dynamically adjust the weights and parameters in the fusion process to achieve optimal fusion effect; and combining with the cross-network context perception optimization strategy to realize the construction of a deep learning driven large model;

[0008] The intelligent fusion module dynamically selects the best network path and transmission strategy according to the output results of the large model.

[0009] Further, the implementation steps of the adaptive multi-modal fusion algorithm are as follows:

[0010] The adaptive multi-modal fusion algorithm performs modal recognition on the input heterogeneous network data through a pre-trained modal recognition module and extracts respective feature vectors that can capture key features of the heterogeneous network data.

[0011] During the fusion process, the adaptive multi-modal fusion algorithm dynamically adjusts the fusion weights between different modal data according to the features of the heterogeneous network data and the task requirements, using attention mechanisms or reinforcement learning techniques.

[0012] The adaptive multi-modal fusion algorithm captures the correlation and complementary information between different modal data through the design of a multi-modal interaction layer, which utilizes the strong learning ability of deep neural networks to generate a unified representation after fusion.

[0013] Further, the implementation steps of the cross-network context-aware optimization strategy are as follows:

[0014] The cross-network context-aware optimization strategy uses graph neural network technology to construct an association graph between heterogeneous networks; the association and dependency relationships between heterogeneous networks and data are represented by nodes and edges in the graph, thereby capturing cross-network context information.

[0015] During the fusion process, the captured cross-network context information is input as an important feature into the adaptive multi-modal fusion algorithm for fusion processing along with the original data.

[0016] According to the fused data and task requirements, the cross-network context-aware optimization strategy dynamically adjusts the optimization objective function to ensure that the fusion result can maximize the satisfaction of actual requirements.

[0017] Further, the intelligent fusion module monitors the changes in network status and business requirements in real time and adjusts the fusion strategy accordingly, including network path selection algorithms, transmission strategy optimization, load balancing, and abnormality grooming.

[0018] In an embodiment of the present application, an intelligent heterogeneous network fusion optimization device based on a large model is also proposed, which comprises:

[0019] A data acquisition module for collecting real-time network status information, business requirement information, and user equipment information of various heterogeneous networks and sending them to a data preprocessing module.

[0020] A data preprocessing module for preprocessing the collected data and sending the processed data to a large model.

[0021] The large model module is used to extract key features that can reflect real-time changes in network status and business demand from pre-processed data; an adaptive multi-modal fusion algorithm is adopted to automatically identify the modal characteristics of heterogeneous network data and dynamically adjust the weights and parameters in the fusion process to achieve optimal fusion results; combined with the cross-network context perception optimization strategy, the construction of a deep learning-driven large model is realized;

[0022] The intelligent fusion module is used to dynamically select the best network path and transmission strategy according to the output results of the large model.

[0023] Further, the implementation steps of the adaptive multi-modal fusion algorithm are as follows:

[0024] The adaptive multi-modal fusion algorithm uses a pre-trained modal recognition module to recognize the modalities of the input heterogeneous network data and extract their respective feature vectors, which can capture the key features of the heterogeneous network data;

[0025] During the fusion process, the adaptive multi-modal fusion algorithm uses attention mechanisms or reinforcement learning techniques to dynamically adjust the fusion weights between different modal data based on the characteristics of the heterogeneous network data and task requirements;

[0026] The adaptive multi-modal fusion algorithm uses a multi-modal interaction layer to capture the correlation and complementary information between different modal data and generate a unified representation after fusion.

[0027] Further, the implementation steps of the cross-network context perception optimization strategy are as follows:

[0028] The cross-network context perception optimization strategy uses graph neural network technology to construct a correlation graph between heterogeneous networks; the nodes and edges in the graph represent the correlation and dependency between heterogeneous networks and data, thereby capturing the cross-network context information;

[0029] During the fusion process, the captured cross-network context information is input into the adaptive multi-modal fusion algorithm as an important feature, along with the original data, for fusion processing;

[0030] According to the fused data and task requirements, the cross-network context perception optimization strategy dynamically adjusts the optimization objective function to ensure that the fusion results can maximize the actual requirements.

[0031] Further, the intelligent fusion module monitors the changes in network status and business demand in real time and adjusts the fusion strategy accordingly, including network path selection algorithm, transmission strategy optimization, load balancing, and exception grooming.

[0032] In an embodiment of the present application, a computer device is also provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the foregoing large model-based intelligent heterogeneous network fusion optimization method when executing the computer program.

[0033] In an embodiment of the present application, a computer readable storage medium is also provided, which stores a computer program for executing the large model-based intelligent heterogeneous network fusion optimization method.

[0034] Advantages:

[0035] 1. The large model of the present application adopts an adaptive multi-modal fusion algorithm and a cross-network context perception optimization strategy, which can learn and adjust itself according to real-time data to adapt to the changing network environment.

[0036] 2. The present application can comprehensively understand the network state by analyzing data in multiple dimensions such as network performance, traffic characteristics, and user behavior.

[0037] 3. The present application can dynamically adjust the fusion strategy according to real-time analysis results to achieve optimal data transmission efficiency and network performance.

[0038] 4. The present application adopts modular design, which is easy to extend and integrate into existing network architecture, improving practicality and flexibility. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a large model-based intelligent heterogeneous network fusion optimization method flowchart of an embodiment of the present application;

[0040] Figure 2 is a large model-based intelligent heterogeneous network fusion optimization device structure diagram of an embodiment of the present application;

[0041] Figure 3 is a computer device structure diagram of the present application. DETAILED DESCRIPTION

[0042] The principles and spirits of the present application will be described below with reference to a number of exemplary embodiments, it should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0043] Those skilled in the art know that the embodiments of the present application can be implemented as an apparatus, a device, an equipment, a method or a computer program product. Therefore, the present disclosure can be embodied in the form of an entirely hardware, an entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0044] According to the embodiments of the present application, an intelligent heterogeneous network fusion optimization method based on large model is proposed. By constructing a deep learning large model, using adaptive multi-modal fusion algorithm and cross-network context perception optimization strategy, the network state and business demand of different types of networks can be analyzed, and intelligent fusion strategy can be formulated accordingly, so as to realize intelligent fusion of networks.

[0045] The principles and spirits of the present application will be explained in detail below with reference to several representative embodiments of the present application.

[0046] Figure 1 is the flowchart of the intelligent heterogeneous network fusion optimization method based on large model of an embodiment of the present application. As shown in Figure 1 , the method comprises:

[0047] 1. Adaptive data collection and preprocessing

[0048] The real-time state data, business demand data and user equipment information of heterogeneous networks such as wireless network, metropolitan area network and wide area network are collected through the data collection module. These data include network state information such as network bandwidth, delay and packet loss rate, as well as business demand information such as data transmission volume, data type and transmission priority. At the same time, the type, location and supported network type of user equipment and other information need to be collected.

[0049] After the data collection is completed, the data needs to be preprocessed. Preprocessing includes data cleaning, feature extraction and normalization, etc. so that the subsequent model can better understand and utilize these data.

[0050] 2. Deep learning driven large model construction

[0051] The key part of the large model architecture supports heterogeneous network data processing and fusion, and adopts adaptive multi-modal fusion algorithm and cross-network context perception optimization strategy.

[0052] Key features are extracted from the preprocessed heterogeneous network data (features cover multiple dimensions such as network performance, business traffic characteristics, user behavior, etc.). These features can reflect the real-time changes of network state and business demand.

[0053] Adaptive multi-modal fusion algorithm automatically identifies the modal characteristics (such as text, image, time series, etc.) of heterogeneous network data, and dynamically adjusts the weights and parameters in the fusion process to achieve the optimal fusion effect.

[0054] The cross-network context-aware optimization strategy emphasizes fully considering the cross-network context information in the fusion process to improve the fusion effect and decision accuracy.

[0055] Through the combination of the adaptive multi-modal fusion algorithm and the cross-network context-aware optimization strategy, the construction of a deep learning-driven large model is realized.

[0056] 3. Dynamic intelligent fusion strategy implementation

[0057] After the construction of the large model, the intelligent fusion module is used to realize the intelligent fusion of the network. According to the output results of the large model, the module dynamically selects the best network path and transmission strategy to realize the efficient transmission of data. At the same time, the module can also monitor the changes of network state and business demand in real time, and adjust the fusion strategy (including path selection algorithm, transmission strategy optimization, load balancing, abnormality grooming, etc.) accordingly to adapt to the changing network environment.

[0058] It should be noted that although the operations of the method of the present application are described in a specific order in the above embodiments and drawings, this does not require or imply that the operations must be performed in this specific order, or that all of the shown operations must be performed to achieve the desired results. Additionally or alternatively, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be divided into multiple steps.

[0059] In order to more clearly explain the above-mentioned large model-based intelligent heterogeneous network fusion optimization method, a specific embodiment will be described below, however, it should be noted that this embodiment is only used to better illustrate the present application and does not constitute an improper limitation on the present application.

[0060] Embodiment:

[0061] The specific implementation steps of the present application are as follows:

[0062] I. Data collection and preprocessing

[0063] 1. The data collection module collects real-time state data, business demand data and user equipment information of different types of networks in real time, and sends these data to the data preprocessing module.

[0064] The specific data are as follows:

[0065] (1) Network state data: including real-time information such as network bandwidth, delay, packet loss rate, etc.

[0066] (2) Business demand data: including data transmission volume, data type (such as video, file, real-time data, etc.), transmission priority, etc.

[0067] (3) User device information: including device type (such as mobile phone, computer, server, etc.), location, supported network type, etc.

[0068] 2. The data preprocessing module cleans, extracts features, and normalizes the collected data, and sends the processed data to the subsequent process.

[0069] import numpy as np

[0070] from sklearn.preprocessing import StandardScaler,MinMaxScaler

[0071] from sklearn.decomposition import PCA

[0072] from sklearn.feature_selection import SelectKBest,chi2

[0073] # Assuming X is the original data, y is the label (only for feature selection)

[0074] X = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])

[0075] y = np.array([0, 1, 0])

[0076] # Data normalization

[0077] min_max_scaler = MinMaxScaler()

[0078] X_normalized = min_max_scaler.fit_transform(X)

[0079] # Data standardization

[0080] standard_scaler = StandardScaler()

[0081] X_standardized = standard_scaler.fit_transform(X)

[0082] # Dimension reduction

[0083] pca = PCA(n_components=2)

[0084] X_pca = pca.fit_transform(X)

[0085] # Feature Selection

[0086] selector = SelectKBest(chi2, k=2)

[0087] X_selected = selector.fit_transform(X, y)

[0088] II. Adaptive Multi-modal Fusion Algorithm

[0089] 1. Modality Recognition and Representation: The algorithm uses a pre-trained modality recognition module to identify the input heterogeneous network data and extract their respective representation vectors. These representation vectors capture the key features of the heterogeneous network data, providing a foundation for subsequent fusion.

[0090]

[0091]

[0092] 2. Adaptive Weight Adjustment: During the fusion process, the algorithm uses attention mechanisms or reinforcement learning techniques to dynamically adjust the fusion weights between different modal data based on the characteristics of the heterogeneous network data and task requirements. This adaptive adjustment ensures that the fusion results better match the actual situation, improving fusion accuracy.

[0093]

[0094]

[0095] 3. Multi-modal Interaction and Fusion: The algorithm designs a multi-modal interaction layer to realize the interaction and fusion between different modal data. This layer uses deep neural networks (such as Transformer, LSTM, etc.) with strong learning capabilities to capture the correlation and complementary information between different modal data, generating a unified representation after fusion.

[0096]

[0097] III. Cross-network Context Awareness Optimization Strategy

[0098] 1. Context Information Construction: This strategy first uses techniques such as Graph Neural Networks (GNN) to construct an association graph between heterogeneous networks. Through the nodes and edges in the graph, the association and dependency relationships between heterogeneous networks and data are represented, capturing cross-network context information.

[0099] import torch

[0100] from torch_geometric.nn import GCNConv

[0101] from torch_geometric.data import Data

[0102] # Assume we have the following node features and edge indices for a heterogeneous network

[0103] node_features = torch.tensor([[1, 2], [3, 4], [5, 6]], dtype=torch.float)

[0104] edge_index = torch.tensor([[0, 1], [1, 2]], dtype=torch.long)

[0105] # Create a graph data object

[0106] graph_data = Data(x=node_features, edge_index=edge_index)

[0107] # Define a graph convolutional network layer

[0108] gcn_layer = GCNConv(in_channels=2, out_channels=2)

[0109] # Perform forward propagation using the graph convolutional network layer

[0110] output = gcn_layer(graph_data.x, graph_data.edge_index)

[0111] print("Output of the GCN layer:", output)

[0112] 2、Context information fusion: During the fusion process, the captured cross-network context information is input as an important feature into the adaptive multi-modal fusion algorithm, along with the original data for fusion processing. This fusion method can make the fusion result more comprehensive and accurate, taking into account the mutual influence and constraint relationship between heterogeneous networks.

[0113] The mutual influence and constraint relationship between heterogeneous networks is crucial for the final task completion. Through context information fusion, these complex relationships can be better understood and processed, thereby improving the accuracy and effectiveness of the task.

[0114] import torch

[0115] from torch_geometric.data import Data

[0116] from torch_geometric.nn import GCNConv

[0117] # Assuming we have the following node features and edge indices for a heterogeneous network

[0118] node_features = torch.tensor([[1, 2], [3, 4], [5, 6]], dtype=torch.float)

[0119] edge_index = torch.tensor([[0, 1], [1, 2]], dtype=torch.long)

[0120] # Create a graph data object

[0121] graph_data = Data(x=node_features, edge_index=edge_index)

[0122] # Define a graph convolutional network layer

[0123] gcn_layer = GCNConv(in_channels=2, out_channels=2)

[0124] # Perform forward propagation using the graph convolutional network layer to get context information

[0125] context_info = gcn_layer(graph_data.x, graph_data.edge_index)

[0126] # Assume we also have original data (represented here by a simple random tensor)

[0127] original_data = torch.randn(3, 2)

[0128] # Fuse the context information with the original data (here simply by adding them)

[0129] fused_data = context_info + original_data

[0130] print("Fused data:", fused_data)

[0131] 3. Optimization target adjustment: Based on the fused data and task requirements, the strategy also dynamically adjusts the optimization target function to ensure that the fusion result can maximize the satisfaction of actual requirements. This adjustment of the target function can make the algorithm more flexible and effective in complex network environments.

[0132]

[0133]

[0134] Four, after building a large model, realize the intelligent fusion of the network through the intelligent fusion module. According to the output results of the large model, the module dynamically selects the best network path and transmission strategy. The intelligent fusion module also monitors the changes of network state and business demand in real time, and adjusts the fusion strategy accordingly, including network path selection algorithm, transmission strategy optimization, load balancing and abnormality analysis.

[0135] For example:

[0136] For example, there is an online video live platform that needs to provide high-definition, low-latency video streaming services to users. In order to ensure the quality of service, the intelligent heterogeneous network fusion optimization device of the large model is adopted.

[0137] 1. Data collection and processing

[0138] Real-time monitoring: Collect network state data in real time, including bandwidth, delay, packet loss rate and other key indicators.

[0139] Business demand analysis: Also receive demand information from the business system, such as the number of current online users, video resolution, encoding format, etc.

[0140] 2. Building of adaptive multi-modal fusion algorithm

[0141] Modality recognition and representation: Use the pre-trained modality recognition module to recognize the modalities of the obtained data and extract their respective representation vectors.

[0142] Adaptive weight adjustment: According to the characteristics of the data and the task requirements (such as whether the user is currently watching video, whether emergency response is needed, etc.), use attention mechanism or reinforcement learning technology to dynamically adjust the fusion weight between different modal data.

[0143] Multi-modal interaction and fusion: Design a multi-modal interaction layer, use deep neural networks (such as Transformer, LSTM, etc.) to capture the correlation and complementary information between different modalities, and generate a unified representation after fusion.

[0144] 3. Cross-network context-aware optimization strategy

[0145] Context information construction: Use technologies such as graph neural networks (GNN) to construct association graphs between heterogeneous networks, capturing cross-network context information (such as device association, user behavior patterns, etc.).

[0146] Context information fusion: Capture the context information as an important input feature, and fuse it with the original data to improve the accuracy and efficiency of data transmission.

[0147] Optimization target adjustment: According to the fused data and task requirements (such as the current demand priority of the user, network status, etc.), dynamically adjust the optimization target function (such as transmission path selection, rate control, priority sorting, etc.) to ensure that the fusion result can maximize the actual demand.

[0148] 4, Large model generation

[0149] Based on the output results of the above steps, develop a fusion strategy for data transmission, including selecting transmission paths, determining transmission rates and priorities, etc.

[0150] Five, support real-time monitoring of network state and business demand changes, and feedback these changes to adjust the fusion strategy in real time.

[0151] Based on the same inventive concept, the present application also proposes an intelligent heterogeneous network fusion optimization device based on a large model. The implementation of this device can refer to the implementation of the above-mentioned method, and the repeated parts will not be repeated. The term "module" used below can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware implementation is also possible and contemplated.

[0152] Figure 2 is an intelligent heterogeneous network fusion optimization device based on a large model according to an embodiment of the present application. As shown in Figure 2 , the device includes:

[0153] The data acquisition module 101 is used to collect real-time network state information, business demand information and user equipment information of various heterogeneous networks, and send them to the data preprocessing module.

[0154] The data preprocessing module 102 is used to preprocess the collected data and send the processed data to the large model.

[0155] The large model module 103 is used to extract key features that can reflect real-time changes in network status and business demand from pre-processed data; an adaptive multi-modal fusion algorithm is used to automatically identify the modal characteristics of heterogeneous network data and dynamically adjust the weights and parameters in the fusion process to achieve optimal fusion results; combined with the cross-network context perception optimization strategy, the construction of a large model driven by deep learning is realized.

[0156] The implementation steps of the adaptive multi-modal fusion algorithm are as follows:

[0157] The adaptive multi-modal fusion algorithm uses a pre-trained modal recognition module to identify the modalities of the input heterogeneous network data and extract their respective feature vectors, which can capture the key features of the heterogeneous network data.

[0158] During the fusion process, the adaptive multi-modal fusion algorithm uses attention mechanisms or reinforcement learning techniques to dynamically adjust the fusion weights between different modal data based on the characteristics of the heterogeneous network data and task requirements.

[0159] The adaptive multi-modal fusion algorithm uses a multi-modal interaction layer to capture the correlation and complementary information between different modal data and generate a unified representation after fusion.

[0160] The implementation steps of the cross-network context perception optimization strategy are as follows:

[0161] The cross-network context perception optimization strategy uses graph neural network technology to construct a correlation graph between heterogeneous networks; the nodes and edges in the graph represent the correlation and dependency between heterogeneous networks and data, thereby capturing cross-network context information.

[0162] During the fusion process, the captured cross-network context information is input into the adaptive multi-modal fusion algorithm as an important feature, along with the original data, for fusion processing.

[0163] According to the fused data and task requirements, the cross-network context perception optimization strategy dynamically adjusts the optimization objective function to ensure that the fusion results can maximize the actual requirements.

[0164] The intelligent fusion module 104 is used to dynamically select the best network path and transmission strategy based on the output results of the large model; this module also monitors the changes in network status and business demand in real time and adjusts the fusion strategy accordingly, including network path selection algorithms, transmission strategy optimization, load balancing, and abnormality analysis.

[0165] It should be noted that although several modules of the large model-based intelligent heterogeneous network fusion optimization apparatus are mentioned in the foregoing detailed description, such division is merely exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules described above can be embodied in one module. Conversely, the features and functions of one module described above can be further divided into several modules embodied by several modules.

[0166] Based on the foregoing inventive concept, as shown in Figure 3 The present application also proposes a computer device 200, comprising a memory 210, a processor 220, and a computer program 230 stored on the memory 210 and executable on the processor 220, wherein the processor 220 implements the foregoing large model-based intelligent heterogeneous network fusion optimization method when executing the computer program 230.

[0167] Based on the foregoing inventive concept, the present application also proposes a computer-readable storage medium storing a computer program for executing the foregoing large model-based intelligent heterogeneous network fusion optimization method.

[0168] The large model-based intelligent heterogeneous network fusion optimization method and apparatus proposed by the present application have the following highlights:

[0169] 1. Adaptive learning capability: The large model of the present application adopts an adaptive multi-modal fusion algorithm and a cross-network context perception optimization strategy, which can learn and adjust itself according to real-time data to adapt to the changing network environment.

[0170] 2. Multi-dimensional data analysis: By comprehensively analyzing data in multiple dimensions such as network performance, traffic characteristics, and user behavior, the present application can more comprehensively understand the network state.

[0171] 3. Dynamic fusion strategy: The present application can dynamically adjust the fusion strategy according to real-time analysis results to achieve optimal data transmission efficiency and network performance.

[0172] 4. Modular design: The present application adopts modular design, which is easy to extend and integrate into existing network architecture, improving practicality and flexibility.

[0173] Although the spirit and principles of the present application have been described with reference to several specific embodiments, it should be understood that the present application is not limited to the disclosed specific embodiments, and the division of aspects does not mean that the features in these aspects cannot be combined for benefit, but only for the convenience of expression. The present application is intended to cover various modifications and equivalent arrangements contained in the spirit and scope of the appended claims.

[0174] The skilled in the art should understand that the modifications or changes made on the basis of the technical scheme of the present application without paying creative labor are still within the protection scope of the present application.

Claims

1. A large model-based intelligent heterogeneous network fusion optimization method, characterized in that, The method comprises: Collecting real-time network state data, service demand data and user equipment data of various heterogeneous networks and preprocessing; From the preprocessed data, key features reflecting the real-time changes of network state and service demand are extracted; an adaptive multi-modal fusion algorithm is used to automatically identify the modal characteristics of heterogeneous network data and dynamically adjust the weights and parameters in the fusion process to achieve the optimal fusion effect; combined with the cross-network context perception optimization strategy, a deep learning driven large model is constructed; The implementation steps of the adaptive multi-modal fusion algorithm are as follows: The adaptive multi-modal fusion algorithm identifies the modal of the input heterogeneous network data through the pre-trained modal recognition module, and extracts the respective feature vectors, which can capture the key features of the heterogeneous network data; In the fusion process, the adaptive multi-modal fusion algorithm uses attention mechanism or reinforcement learning technology to dynamically adjust the fusion weights between different modal data according to the characteristics of the heterogeneous network data and the task requirements; The adaptive multi-modal fusion algorithm captures the correlation and complementary information between different modal data through the design of a multi-modal interaction layer, which uses the strong learning ability of deep neural networks to generate a unified representation after fusion; The implementation steps of the cross-network context perception optimization strategy are as follows: The cross-network context perception optimization strategy uses graph neural network technology to construct the association graph between heterogeneous networks; the association and dependency relationship between heterogeneous networks and data is represented by nodes and edges in the graph, so as to capture the cross-network context information; In the fusion process, the captured cross-network context information is input into the adaptive multi-modal fusion algorithm as an important feature, and is fused with the original data; According to the fused data and task requirements, the cross-network context perception optimization strategy dynamically adjusts the optimization objective function to ensure that the fusion result can maximize the actual requirements; The intelligent fusion module dynamically selects the best network path and transmission strategy according to the output result of the large model; the intelligent fusion module monitors the changes of network state and service demand in real time, and adjusts the fusion strategy accordingly, including network path selection algorithm, transmission strategy optimization, load balancing and abnormality sorting.

2. A large model-based intelligent heterogeneous network fusion optimization device, characterized in that, The device comprises: A data acquisition module for collecting real-time network state information, service demand information and user equipment information of various heterogeneous networks and sending them to the data preprocessing module; A data preprocessing module for preprocessing the collected data and sending the processed data to the large model; A large model module for extracting key features reflecting the real-time changes of network state and service demand from the preprocessed data; using an adaptive multi-modal fusion algorithm, automatically identifying the modal characteristics of heterogeneous network data and dynamically adjusting the weights and parameters in the fusion process to achieve the optimal fusion effect; combined with the cross-network context perception optimization strategy, a deep learning driven large model is constructed; The implementation steps of the adaptive multi-modal fusion algorithm are as follows: The adaptive multi-modal fusion algorithm performs modal recognition on the input heterogeneous network data through a pre-trained modal recognition module and extracts respective feature vectors that can capture key features of the heterogeneous network data; During the fusion process, the adaptive multi-modal fusion algorithm dynamically adjusts the fusion weights between different modal data according to the features of the heterogeneous network data and the task requirements, using attention mechanisms or reinforcement learning techniques; The adaptive multi-modal fusion algorithm captures the correlation and complementary information between different modal data through the design of a multi-modal interaction layer, which utilizes the strong learning ability of deep neural networks to generate a unified representation after fusion; The implementation steps of the cross-network context-aware optimization strategy are as follows: The cross-network context-aware optimization strategy uses graph neural network technology to construct an association graph between heterogeneous networks; the association and dependency relationships between heterogeneous networks and data are represented by nodes and edges in the graph, thereby capturing cross-network context information; During the fusion process, the captured cross-network context information is input into the adaptive multi-modal fusion algorithm as an important feature, along with the original data, for fusion processing; According to the fused data and task requirements, the cross-network context-aware optimization strategy dynamically adjusts the optimization objective function to ensure that the fusion result can maximize the satisfaction of actual requirements; The intelligent fusion module is used to dynamically select the best network path and transmission strategy according to the output results of the large model; real-time monitoring of network status and changes in business requirements is performed, and the fusion strategy is adjusted accordingly, including network path selection algorithm, transmission strategy optimization, load balancing, and abnormality grooming.

3. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of claim 1.

4. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program for executing the method of claim 1.

Images