Dynamic social network-oriented adaptive influence propagation path planning method and system
By combining dynamic graph neural networks and reinforcement learning models with user dynamic embedding and information content, the problem of personalization and adaptation in propagation path planning in dynamic social networks is solved, and efficient optimization of information propagation paths is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-23
- Publication Date
- 2026-03-20
AI Technical Summary
Existing technologies cannot effectively reflect user personalization differences and the impact of information content in dynamic social networks, and lack real-time feedback and adaptive adjustment mechanisms, resulting in a decline in the effectiveness of information dissemination path planning in dynamic environments.
We employ dynamic graph neural networks to process multi-source heterogeneous social network data, construct a propagation probability prediction model based on an attention mechanism and a reinforcement learning strategy model, and adjust the propagation path in real time by combining user dynamic embedding and information content.
It achieves accurate capture of user state changes in dynamic social networks, improves the accuracy of propagation probability estimation, obtains globally optimal path planning, and has good robustness and adaptability.
Smart Images
Figure CN121707549A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of information technology and social network analysis, and particularly relates to a dynamic social network-oriented adaptive influence propagation path planning method and system. BACKGROUND
[0002] With the popularity of the Internet and the wide application of intelligent terminals, social networks have become the core carrier of information dissemination. In application scenarios such as commercial marketing, the "influence maximization" problem is involved, that is, how to select appropriate initial propagation node sets so that information can cover as wide a user group as possible with the least investment. Therefore, influence propagation path planning in social networks has become a research hotspot.
[0003] Existing researches usually model information diffusion based on the independent cascade (IC) model or the linear threshold (LT) model, and determine the propagation nodes through greedy algorithms, centrality algorithms and other methods. However, most of these methods are based on the assumption of static network snapshots, and cannot reflect the characteristics of real social networks evolving over time, resulting in a significant decrease in the effectiveness of propagation path planning in dynamic environments. In addition, existing methods often simplify the propagation probability between users as a global constant, ignoring the individual differences of users and the influence of information content itself, and are difficult to truly simulate the propagation rules of information among different groups of people.
[0004] On the other hand, existing influence propagation path planning is usually a "one-time" offline decision, lacking real-time feedback and adjustment mechanisms during the propagation process. Once the propagation environment changes, the existing scheme cannot adaptively correct the strategy, which easily leads to a significant reduction in the range and efficiency of information diffusion.
[0005] In summary, the existing technology generally has problems such as static, roughness and open-loop, and is difficult to meet the efficient information propagation demand in real social network environment. Therefore, there is an urgent need for an influence propagation path planning method and system that can combine the dynamic evolution characteristics of social networks, accurately model the individualized propagation probability of users, and adaptively adjust according to real-time feedback during the propagation process. SUMMARY
[0006] In order to solve the above technical problems in the prior art, the present application provides a dynamic social network-oriented adaptive influence propagation path planning method and system to solve the above technical problems.
[0007] According to a first aspect of the present application, a dynamic social network-oriented adaptive influence propagation path planning method is provided, comprising:
[0008] S1: fusing and processing multi-source heterogeneous social network data collected in real time based on a dynamic graph neural network to generate a dynamic embedding vector capable of representing a user's structural position, activity level, and interest preference;
[0009] S2: constructing a propagation probability prediction model, combining user dynamic embedding and information content embedding, and outputting personalized propagation probability;
[0010] S3: modeling the information propagation process as a sequential decision problem, constructing a reinforcement learning strategy model, and selecting a propagation path according to the current network state and the predicted probability;
[0011] S4: online updating the propagation probability prediction model and the reinforcement learning strategy model according to actual propagation feedback, thereby dynamically and adaptively adjusting the subsequent propagation path.
[0012] In some specific embodiments, the multi-source heterogeneous social network data includes user relationship data, user behavior data, content interaction data, and user attribute data.
[0013] In some specific embodiments, the dynamic graph neural network is used to dynamically update the addition or cancellation of user relationships, changes in user activity, and drifts in user interests.
[0014] In some specific embodiments, the propagation probability prediction model is a deep learning model based on an attention mechanism, taking the source user node embedding vector, the target user node embedding vector, and the content embedding vector of the information to be propagated as input, and outputting the propagation probability of the source user to the target user under a specific information.
[0015] In some specific embodiments, the reinforcement learning strategy model is solved by modeling the influence propagation process as a Markov decision process, wherein the agent is a central path planning server, the environment is composed of a dynamic social network and a propagation probability prediction model, the state is the current set of activated users, the corresponding dynamic embedding vector, and the network topology state, the action is to select one or more users from the set of activated users and propagate information to their neighbor nodes, the reward is the sum of the number of newly activated users and the estimate of long-term propagation potential, and the agent learns the strategy of selecting the optimal propagation action according to the network dynamic state through interaction with the environment.
[0016] In some specific embodiments, S4 specifically includes comparing the user forwarding behavior monitored in the actual propagation process with the output result of the propagation probability prediction model, generating an error signal, and adjusting the parameters of the propagation probability prediction model and the reinforcement learning strategy model accordingly.
[0017] According to a second aspect of the invention, a computer-readable storage medium is provided on which one or more computer programs are stored, which, when executed by a computer processor, implement the method described above.
[0018] According to a third aspect of the present invention, an adaptive influence propagation path planning system for dynamic social networks is proposed, comprising:
[0019] The dynamic representation module is configured to fuse real-time acquired multi-source heterogeneous social network data based on a dynamic graph neural network to generate dynamic embedding vectors that can represent user structural location, activity level, and interest preferences.
[0020] The propagation probability prediction module is configured to combine user-dynamic embedding with information content embedding to output personalized propagation probabilities.
[0021] The path planning module is configured to model the information propagation process as a sequential decision problem, construct a reinforcement learning policy model, and select a propagation path based on the current network state and predicted probabilities.
[0022] The online learning module is configured to compare the user forwarding behavior monitored during the actual propagation process with the output results of the propagation probability prediction model, generate error signals, and adjust the parameters of the propagation probability prediction model and the reinforcement learning strategy model accordingly, thereby dynamically and adaptively adjusting the subsequent propagation path.
[0023] In some specific embodiments, multi-source heterogeneous social network data includes user relationship data, user behavior data, content interaction data, and user attribute data. Dynamic graph neural networks are used to dynamically update the addition or cancellation of user relationships, changes in user activity, and shifts in user interests.
[0024] In some specific embodiments, the propagation probability prediction model is a deep learning model based on an attention mechanism. It takes the source user node embedding vector, the target user node embedding vector, and the content embedding vector of the information to be propagated as inputs, and outputs the propagation probability of the source user to the target user under specific information. The reinforcement learning strategy model solves the influence propagation process by modeling it as a Markov decision process. In this model, the agent is a central path planning server, the environment consists of a dynamic social network and the propagation probability prediction model, the state is the currently activated user set, the corresponding dynamic embedding vector, and the network topology state, the action is to select one or more users from the activated user set and propagate the information to their neighbor nodes, and the reward is the sum of the number of newly activated users and the prediction of the long-term propagation potential. The agent learns the strategy of selecting the optimal propagation action based on the dynamic state of the network through interaction with the environment.
[0025] This invention proposes an adaptive influence propagation path planning method and system for dynamic social networks, which has the following beneficial effects:
[0026] First, this invention introduces a dynamic graph neural network to model social network relationships, user behavior, and interest drift in real time, which can accurately capture the dynamic changes in network structure and user state, effectively overcoming the problem of propagation path failure caused by traditional methods based on static snapshot modeling.
[0027] Secondly, the propagation probability prediction model based on the attention mechanism constructed in this invention combines user dynamic embedding with information content embedding, which can output differentiated propagation probabilities for different users and different information content, greatly improving the accuracy of propagation probability estimation. Compared with the method of setting the propagation probability as a global constant, its results are more in line with the real propagation law.
[0028] Furthermore, this invention models the propagation path planning problem as a Markov decision process (MDP) and uses a reinforcement learning policy model to solve it, so that path selection is no longer limited to the local optimum of the greedy policy, but can obtain the globally optimal path planning scheme based on the long-term propagation potential.
[0029] Finally, the real-time feedback and online update mechanism set up in this invention can continuously monitor the actual diffusion effect during the propagation process and use the feedback to correct the prediction model and strategy model, thereby forming a closed-loop adaptive optimization system. This allows the propagation path to be continuously optimized as the network environment and user behavior change, exhibiting good robustness and adaptability. Attached Figure Description
[0030] The accompanying drawings are included to provide a further understanding of the embodiments and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments and, together with the description, serve to explain the principles of the invention. Other embodiments and many anticipated advantages of the embodiments will be readily recognized as they become better understood through reference to the following detailed description. Other features, objects, and advantages of this application will become more apparent from reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings:
[0031] Figure 1 This is a flowchart of an embodiment of the adaptive influence propagation path planning method for dynamic social networks in this application;
[0032] Figure 2 This is a flowchart of an adaptive influence propagation path planning method for dynamic social networks, which is a specific embodiment of this application;
[0033] Figure 3This is a framework diagram of an adaptive influence propagation path planning system for dynamic social networks, according to one embodiment of this application.
[0034] Figure 4 This is a general architecture diagram of an adaptive influence propagation path planning system for dynamic social networks, which is a specific embodiment of this application.
[0035] Figure 5 This is a schematic diagram of the structure of a computer system used to implement the electronic device of the present application. Detailed Implementation
[0036] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.
[0037] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0038] Figure 1 A flowchart illustrating an adaptive influence propagation path planning method for dynamic social networks according to an embodiment of this application is shown. Figure 1 As shown, the method includes the following steps:
[0039] S101: Based on dynamic graph neural networks, real-time collected multi-source heterogeneous social network data is fused and processed to generate dynamic embedding vectors that can represent user structural location, activity level and interest preferences.
[0040] In specific embodiments, multi-source heterogeneous social network data includes user relationship data, user behavior data, content interaction data, and user attribute data. Dynamic graph neural networks are used to dynamically update the addition or removal of user relationships, changes in user activity, and shifts in user interests.
[0041] S102: Construct a propagation probability prediction model, combining user dynamic embedding and information content embedding to output personalized propagation probabilities.
[0042] In a specific embodiment, the propagation probability prediction model is a deep learning model based on the attention mechanism. It takes the source user node embedding vector, the target user node embedding vector, and the content embedding vector of the information to be propagated as input, and outputs the propagation probability of the source user to the target user under specific information.
[0043] S103: Model the information propagation process as a sequential decision problem, construct a reinforcement learning policy model, and select the propagation path based on the current network state and predicted probability.
[0044] In a specific embodiment, the reinforcement learning strategy model solves the influence propagation process by modeling it as a Markov decision process. The agent is a central path planning server, the environment consists of a dynamic social network and a propagation probability prediction model, the state is the currently activated user set, the corresponding dynamic embedding vector, and the network topology state, the action is to select one or more users from the activated user set and propagate the information to their neighbor nodes, and the reward is the sum of the number of newly activated users and the prediction of long-term propagation potential. The agent learns through interaction with the environment to obtain a strategy for selecting the optimal propagation action based on the dynamic state of the network.
[0045] S104: Update the propagation probability prediction model and reinforcement learning strategy model online based on actual propagation feedback, thereby dynamically and adaptively adjusting the subsequent propagation path.
[0046] In a specific embodiment, the user forwarding behavior monitored during the actual propagation process is compared with the output results of the propagation probability prediction model to generate an error signal, and the parameters of the propagation probability prediction model and the reinforcement learning strategy model are adjusted accordingly.
[0047] Figure 2 A flowchart illustrating an adaptive influence propagation path planning method for dynamic social networks, as shown in a specific embodiment of this application, is provided. Figure 2 As shown, it includes the following steps:
[0048] S201: Multidimensional Dynamic Network Representation Learning, using Dynamic Graph Neural Networks (Dynamic GNNs) to process real-time data streams.
[0049] In a specific embodiment, dynamic data streams from the target social network are collected in real time, including: network structure data: follow / friend relationships between users and their changes; user behavior data: sequences of user posting, forwarding, liking, and commenting behaviors; content interaction data: the history of user information interaction with specific topics and keywords; and user attribute data: user tags, self-descriptions, and geographic location information. A Dynamic Graph Neural Network (Dynamic GNN) is used to fuse the above multi-source heterogeneous data, generating a dynamic embedding vector for each user node at a specific moment. This vector comprehensively represents the user's network structure position, current activity level, and interest preferences. The Dynamic GNN provides an integrated, embedded, real-time representation of the constantly changing social network structure (such as adding or unfollowing) and user dynamic attributes (such as real-time activity and interest drift). This differs from traditional methods that process network structure and node attributes separately or use static snapshots.
[0050] S202: Personalized propagation probability prediction, integrating node and content vector output probabilities.
[0051] In a specific embodiment, a deep learning model based on an attention mechanism is constructed. This model takes the source user node embedding vector and target user node embedding vector obtained in S201 above, as well as the content embedding vector of the information to be propagated, as input, and outputs a personalized propagation probability value P(v|u,M), representing the probability that user u successfully influences user v under specific information M. This model is trained using historical propagation data. The input of the attention-based deep learning model includes not only the embedding vectors of the source and target users, but also, in particular, the content embedding vector of the information to be propagated (Message) itself. This makes the propagation probability prediction personalized (varies from person to person) and highly correlated with the content (varies from information to information), rather than a global constant.
[0052] S203: Adaptive path planning, where the RL agent selects the best action based on the state.
[0053] In a specific embodiment, the influence propagation process is modeled as a sequential decision problem and solved using a reinforcement learning framework. The agent is a central path planning server; the environment is the dynamic social network and prediction model defined in S201 and S202; the state is the environment state St at time t, representing the currently activated user set, their dynamic embedding vectors, and the current network topology; the action is At, which at time t selects one or more users from the activated user set to perform the next propagation action (i.e., propagating information to their neighbors); the reward is defined as the number of newly activated users after an action (immediate reward) plus the estimated long-term propagation potential (long-term reward). The agent (using algorithms such as PPO or DQN) learns an optimal policy π(At|St) through continuous interaction with the environment. This policy adaptively selects the best action to expand the propagation range based on the current network dynamic state (i.e., which activated user to choose for propagation is most effective). The reinforcement learning agent defines "which node to choose for propagation next" as its action, "the current set of activated nodes and network state" as its state, and "the number and quality of newly activated nodes" as its reward. This modeling approach transforms a discrete combinatorial optimization problem into a learnable sequential decision problem, thereby achieving adaptive path planning with global optimization.
[0054] S204: Policy execution and feedback collection, performing propagation and collecting results in a real network.
[0055] In a specific implementation, during the actual execution of information dissemination, the system continuously monitors the real dissemination effect (e.g., which users actually forwarded the message). This real feedback data is compared with the prediction results of the S202 model to generate an error signal. This error signal is then used to fine-tune the S202 prediction model and the S203 policy model online, enabling them to quickly adapt to network changes and improve their generalization performance and stability in dynamic environments. Based on a real-time feedback online collaborative learning mechanism, a closed-loop system is established, using the feedback data generated from real online dissemination to simultaneously update the dissemination probability prediction model and the reinforcement learning policy model. This allows both models to continuously self-correct and optimize in practical applications, achieving a "the more you use it, the smarter it becomes" effect and possessing true adaptive capabilities.
[0056] Figure 3 This application presents a framework diagram of an adaptive influence propagation path planning system for dynamic social networks, as shown in the following embodiment. Figure 3As shown, the system includes a dynamic representation module 301, a propagation probability prediction module 302, a path planning module 303, and an online learning module 304. The dynamic representation module 301 is configured to fuse real-time collected multi-source heterogeneous social network data based on a dynamic graph neural network to generate dynamic embedding vectors that characterize user structural location, activity level, and interest preferences. The propagation probability prediction module 302 is configured to combine user dynamic embeddings with information content embeddings to output personalized propagation probabilities. The path planning module 303 is configured to model the information propagation process as a sequential decision problem, construct a reinforcement learning strategy model, and select a propagation path based on the current network state and predicted probabilities. The online learning module 304 is configured to compare the monitored user forwarding behavior during actual propagation with the output of the propagation probability prediction model, generate error signals, and adjust the parameters of the propagation probability prediction model and the reinforcement learning strategy model accordingly, thereby dynamically and adaptively adjusting subsequent propagation paths.
[0057] Figure 4 This application illustrates the overall architecture diagram of an adaptive influence propagation path planning system for dynamic social networks, based on a specific embodiment of the present application. Figure 4As shown, in the data acquisition layer, dynamic data streams, including user relationship data, user behavior data, content interaction data, and user attribute data, are acquired in real time. These dynamic data streams are input to the dynamic representation module 301 in the system of this invention. The dynamic representation module 301 (dynamic graph neural network) receives the dynamic data streams, performs fusion processing, and outputs dynamic node embedding vectors. On one hand, these dynamic node embedding vectors are provided as input to the probability prediction module 302 (attention mechanism model); on the other hand, they are also transmitted as network state information to the policy learning module 303 (reinforcement learning agent). Externally, the information to be propagated is first processed by an information encoder to obtain an embedded representation of the information content. This information content embedding, along with the dynamic node embedding vector, is input to the probability prediction module. The probability prediction module 302 (attention mechanism model) fuses the source user embedding, target user embedding, and information content embedding to output a personalized propagation probability. This propagation probability is output as a result and, along with the network state provided by the dynamic representation module, is input to the policy learning module 303. The policy learning module 303 (reinforcement learning agent), after receiving the network state and personalized propagation probabilities, models the information propagation process as an environment model and outputs a decision action, namely, selecting the set of nodes for propagation. This decision result is passed to the policy execution module 305. The policy execution module 305, based on the action selection output by the policy learning module, executes the propagation in a real social network environment and collects feedback data after propagation, including node activation status and propagation range. The feedback data is sent back to the online learning module 304. The online learning module 304 compares the actual propagation result with the prediction result of the probability prediction module, generates an error signal, and updates the parameters of the probability prediction module and the policy learning module accordingly, thereby correcting the propagation probability estimation and policy selection. The updated model is then applied to subsequent propagation rounds, forming a continuously optimizing loop.
[0058] The following is for reference. Figure 5 It shows a schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application. Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.
[0059] like Figure 5 As shown, the computer system includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 502 or programs loaded from storage section 508 into random access memory (RAM) 503. RAM 503 also stores various programs and data required for the operation of system 500. CPU 501, ROM 502, and RAM 503 are interconnected via bus 504. Input / output (I / O) interface 505 is also connected to bus 504.
[0060] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a liquid crystal display (LCD) and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card and a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed.
[0061] Specifically, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs the functions defined in the methods of this application. It should be noted that the computer-readable storage medium of this application can be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer-readable storage medium other than a computer-readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Program code contained on a computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0062] Computer program code for performing the operations of this application can be written in one or more programming languages or a combination thereof. Programming languages include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0063] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0064] The modules described in the embodiments of this application can be implemented in software or in hardware.
[0065] In another aspect, this application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable storage medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: perform fusion processing on real-time collected multi-source heterogeneous social network data based on a dynamic graph neural network to generate dynamic embedding vectors that can characterize user structural location, activity level, and interest preferences; construct a propagation probability prediction model, combining user dynamic embedding and information content embedding to output personalized propagation probabilities; model the information propagation process as a sequential decision problem, construct a reinforcement learning strategy model, and select a propagation path based on the current network state and predicted probabilities; and update the propagation probability prediction model and reinforcement learning strategy model online based on actual propagation feedback, dynamically and adaptively adjusting subsequent propagation paths.
[0066] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.
Claims
1. An adaptive influence propagation path planning method for dynamic social networks, characterized in that, include: S1: Based on dynamic graph neural networks, real-time collected multi-source heterogeneous social network data is fused and processed to generate dynamic embedding vectors that can represent user structural location, activity level and interest preferences. S2: Construct a propagation probability prediction model, combining user dynamic embedding and information content embedding, and output personalized propagation probabilities; S3: Model the information propagation process as a sequential decision problem, construct a reinforcement learning policy model, and select the propagation path based on the current network state and predicted probability; S4: Update the propagation probability prediction model and reinforcement learning strategy model online based on actual propagation feedback, thereby dynamically and adaptively adjusting the subsequent propagation path.
2. The adaptive influence propagation path planning method for dynamic social networks according to claim 1, characterized in that, The multi-source heterogeneous social network data includes user relationship data, user behavior data, content interaction data, and user attribute data.
3. The adaptive influence propagation path planning method for dynamic social networks according to claim 1, characterized in that, The dynamic graph neural network is used to dynamically update user relationships, changes in user activity, and shifts in user interests.
4. The adaptive influence propagation path planning method for dynamic social networks according to claim 1, characterized in that, The propagation probability prediction model is a deep learning model based on the attention mechanism. It takes the source user node embedding vector, the target user node embedding vector, and the content embedding vector of the information to be propagated as input, and outputs the propagation probability of the source user to the target user under the predetermined information.
5. The adaptive influence propagation path planning method for dynamic social networks according to claim 1, characterized in that, The reinforcement learning strategy model solves the influence propagation process by modeling it as a Markov decision process. The agent is a central path planning server, the environment consists of a dynamic social network and a propagation probability prediction model, the state is the currently activated user set, the corresponding dynamic embedding vector, and the network topology state, the action is to select one or more users from the activated user set and propagate the information to their neighboring nodes, and the reward is the sum of the number of newly activated users and the prediction of long-term propagation potential. The agent learns the strategy of selecting the optimal propagation action based on the network dynamic state through interaction with the environment.
6. The adaptive influence propagation path planning method for dynamic social networks according to claim 1, characterized in that, S4 specifically includes comparing the user forwarding behavior monitored during the actual propagation process with the output results of the propagation probability prediction model, generating an error signal, and adjusting the parameters of the propagation probability prediction model and the reinforcement learning strategy model accordingly.
7. A computer-readable storage medium having one or more computer programs stored thereon, characterized in that, When the one or more computer programs are executed by a computer processor, they perform the method according to any one of claims 1-6.
8. An adaptive influence propagation path planning system for dynamic social networks, characterized in that, include: The dynamic representation module is configured to fuse real-time acquired multi-source heterogeneous social network data based on a dynamic graph neural network to generate dynamic embedding vectors that can represent user structural location, activity level, and interest preferences. The propagation probability prediction module is configured to combine user-dynamic embedding with information content embedding to output personalized propagation probabilities. The path planning module is configured to model the information propagation process as a sequential decision problem, construct a reinforcement learning policy model, and select a propagation path based on the current network state and predicted probabilities. The online learning module is configured to compare the user forwarding behavior monitored during the actual propagation process with the output results of the propagation probability prediction model, generate an error signal, and adjust the parameters of the propagation probability prediction model and the reinforcement learning strategy model accordingly, so as to dynamically and adaptively adjust the subsequent propagation path.
9. The adaptive influence propagation path planning system for dynamic social networks according to claim 8, characterized in that, The multi-source heterogeneous social network data includes user relationship data, user behavior data, content interaction data, and user attribute data. The dynamic graph neural network is used to dynamically update the addition or cancellation of user relationships, changes in user activity, and shifts in user interests.
10. The adaptive influence propagation path planning system for dynamic social networks according to claim 8, characterized in that, The propagation probability prediction model is a deep learning model based on the attention mechanism. It takes the source user node embedding vector, the target user node embedding vector, and the content embedding vector of the information to be propagated as input, and outputs the propagation probability of the source user to the target user under the predetermined information. The reinforcement learning strategy model solves the influence propagation process by modeling it as a Markov decision process. The agent is a central path planning server, the environment consists of a dynamic social network and a propagation probability prediction model, the state is the currently activated user set, the corresponding dynamic embedding vector, and the network topology state, the action is to select one or more users from the activated user set and propagate the information to their neighboring nodes, and the reward is the sum of the number of newly activated users and the prediction of long-term propagation potential. The agent learns the strategy of selecting the optimal propagation action based on the network dynamic state through interaction with the environment.