An information concatenation sequence prediction method, device, equipment and medium

Abstract

The present application relates to the technical field of information processing, in particular to a kind of information cascade sequence prediction method, device, equipment and medium, by extracting the propagation trajectory of target information, the historical behavior trajectory of user and the diffusion trajectory of associated information, construct multi-level data association;Subsequently by iteratively executing the steps of information influence update and user influence update, make information propagation value and user propagation capacity mutually enhance in cycle:In each iteration, based on the updated user influence re-evaluates information propagation potential, while based on the updated information influence re-evaluates user propagation efficiency, form the feedback loop of continuous optimization.Effectively overcome the information loss problem of traditional one-way modeling, can simultaneously capture the dynamic interaction of propagation content features, propagator attribute features and network structure features, improve the effect of information propagation prediction.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and specifically to a method, apparatus, device, and medium for predicting information cascade sequences. Background Technology

[0002] Information cascade sequence prediction refers to predicting the popularity of a piece of information (such as microblogs, news articles, or academic papers) on social networks or citation networks. This technology is core to public opinion monitoring, trending content discovery, and influence assessment, and its accurate predictions can provide relevant departments with strong decision-making support. Each timestamp corresponds to an information dissemination event, reflecting the dynamic interaction process of system behavior over time.

[0003] Existing cascading sequence prediction models mainly include point process models and dynamic graph models. Point process models model the intensity of event occurrence, with representative models including TREND and CASPER. TREND designs a message-passing network to estimate the Hawkes intensity function, but its drawback is that it only focuses on the interaction characteristics and intensity between node variables, ignoring the information diffusion process. CASPER proposes a prediction model with analytical solutions, using historical timestamps and the number of time-related nodes to estimate the mean and variance of the cascading diffusion scale. However, its cost is the oversimplification of complex real-world interactions, compressing the rich diffusion process, composed of heterogeneous variable couplings, into a single scale-growing statistical model, leading to severe information loss. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, device, and medium for predicting information cascade sequences, thereby solving the problems in the prior art.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a method for predicting information concatenation sequences, including:

[0007] Obtain the initial set of users who forwarded the target information before the prediction time, and the forwarding behavior trajectory of each initial user before the prediction time, wherein the forwarding behavior trajectory represents the data record of all forwarded information forwarded by the initial user, and the forwarding behavior trajectory includes the forwarding of the target information;

[0008] For each forwarding information in each forwarding behavior trajectory, the forwarding information diffusion trajectory before the predicted time is obtained. The forwarding information diffusion trajectory is used to characterize the data records of all forwarding diffusion users forwarding the forwarding information. The initial forwarding user is included among all forwarding diffusion users.

[0009] For each forwarded message, based on the forwarding message diffusion trajectory of the forwarded message and the initialized forwarding user influence of each forwarding diffusion user, the forwarding message influence of the forwarded message is updated to obtain the updated forwarding message influence of each forwarded message.

[0010] For each initial forwarding user, based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, the forwarding user influence of the initial forwarding user is updated to obtain the updated forwarding user influence of each initial forwarding user.

[0011] The forwarding user influence updated by each initial forwarding user is taken as the initial forwarding user influence of each initial forwarding user among all forwarding diffusion users. Asynchronous loop optimization is performed according to the steps of updating the forwarding information influence and updating the user influence until the convergence condition is met. The forwarding information influence of the target information is finally updated as the prediction result.

[0012] Preferably, obtaining the initial set of forwarding users of the target information before the prediction time, and the forwarding behavior trajectory of each initial forwarding user before the prediction time, includes:

[0013] Extract all user identifiers that have forwarded the target information from the historical forwarding records of the target information and whose forwarding time is before the predicted time, and use the set of user identifiers as the initial forwarding user set;

[0014] For each user in the initial set of forwarding users, extract all forwarding records of the user before the predicted time, and each forwarding record contains at least the information identifier of the forwarded information and the forwarding timestamp;

[0015] The forwarding records are arranged in chronological order according to their forwarding timestamps to form a forwarding behavior trajectory.

[0016] Preferably, obtaining the forwarding information diffusion trajectory before the predicted time includes:

[0017] Extract the historical records of the forwarding information that were forwarded by all users before the predicted time, with each historical record containing at least the forwarding user identifier and the forwarding timestamp;

[0018] The historical records are arranged in chronological order of forwarding timestamps to form a trajectory of the spread of forwarded information.

[0019] Preferably, the initialized forwarding user influence of each forwarding diffusion user is obtained according to the following method:

[0020] Calculate the initial influence of each forwarding user based on the static attribute data and / or dynamic behavior statistics of each forwarding and diffusion user;

[0021] The static attribute data includes at least one of the following: account permission level, number of followers, number of following accounts, and account registration duration.

[0022] The dynamic behavior statistics include at least one of the following: total historical forwarding count, average historical forwarding interval, and average popularity of historical forwarded information.

[0023] Preferably, the step of updating the forwarding influence of the forwarding information based on the forwarding information diffusion trajectory and the initialized forwarding user influence of each forwarding diffusion user, to obtain the updated forwarding influence of each forwarding information, includes:

[0024] The diffusion trajectory of the forwarded information is encoded, and the diffusion pattern features of the forwarded information itself are extracted;

[0025] Aggregate the initial forwarding user influence of each forwarding and diffusion user to generate user influence association features;

[0026] By integrating the diffusion pattern features and the user influence association features, and calculating the forwarding information influence of the forwarding information update through a preset first update model, the forwarding information update is output.

[0027] Preferably, the step of updating the forwarding user influence of the initial forwarding user based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each updated forwarding information, to obtain the updated forwarding user influence of each initial forwarding user, includes:

[0028] Encode the forwarding behavior trajectory of the initial forwarding user and extract the behavioral pattern features of the initial forwarding user;

[0029] Aggregate the influence of each forwarding message corresponding to the forwarding behavior trajectory to generate a forwarding information influence association feature;

[0030] By integrating the behavioral pattern features with the forwarding information influence association features, and calculating through a preset second update model, the updated forwarding user influence of the initial forwarding user is output.

[0031] Preferably, the convergence condition includes at least one of the following:

[0032] In two consecutive asynchronous loop optimizations, the change in the forwarding information influence of the target information is less than the first threshold.

[0033] The asynchronous loop optimization count reaches the preset maximum number of rounds.

[0034] Secondly, embodiments of the present invention provide an information concatenation sequence prediction device, comprising:

[0035] The first acquisition module is used to acquire the initial set of forwarding users of the target information before the prediction time, and the forwarding behavior trajectory of each initial forwarding user before the prediction time, wherein the forwarding behavior trajectory represents the data record of all forwarded information forwarded by the initial forwarding user, and the forwarding behavior trajectory includes the forwarding of the target information;

[0036] The second acquisition module is used to acquire the forwarding information diffusion trajectory before the predicted time for each forwarding information in each forwarding behavior trajectory. The forwarding information diffusion trajectory is used to characterize the data records of all forwarding diffusion users forwarding the forwarding information. The initial forwarding user is included among all forwarding diffusion users.

[0037] The first update module is used to update the forwarding influence of each forwarding message based on the forwarding information diffusion trajectory of the forwarding message and the initial forwarding user influence of each forwarding diffusion user, so as to obtain the updated forwarding influence of each forwarding message.

[0038] The second update module is used to update the forwarding user influence of each initial forwarding user based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, so as to obtain the updated forwarding user influence of each initial forwarding user.

[0039] The output module is used to take the forwarding user influence updated by each initial forwarding user as the initial forwarding user influence of each initial forwarding user among all forwarding diffusion users, and perform asynchronous loop optimization according to the steps of updating the forwarding information influence and updating the user influence until the convergence condition is met, and output the forwarding information influence of the target information finally updated as the prediction result.

[0040] Thirdly, embodiments of the present invention provide an electronic device, including: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method of the first aspect described above.

[0041] Fourthly, embodiments of the present invention provide a storage medium storing computer program instructions, which, when executed by a processor, implement the method of the first aspect described above.

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0043] This method constructs a two-layer diffusion network consisting of target information, initial forwarding users, and their associated information. It also designs an asynchronous cyclic optimization mechanism that alternately updates information influence and user influence, enabling dynamic collaborative modeling of the information dissemination process. Specifically, the system first extracts the propagation trajectory of the target information, the historical behavior trajectory of users, and the diffusion trajectory of associated information to construct multi-level data associations. Then, by iteratively updating information influence and user influence, the system ensures that the value of information dissemination and the user's dissemination ability mutually reinforce each other in a loop. In each iteration, the system reassesses the information dissemination potential based on the updated user influence, and simultaneously reassesses the user's dissemination effectiveness based on the updated information influence, forming a continuously optimizing feedback loop. This mechanism effectively overcomes the information loss problem of traditional unidirectional modeling, simultaneously capturing the dynamic interaction of dissemination content characteristics, disseminator attribute characteristics, and network structure characteristics. By setting convergence conditions to control the iteration depth, the system reaches a stable state within a finite number of rounds, ultimately outputting the information influence after multiple rounds of collaborative optimization as the prediction result. This achieves more comprehensive and robust prediction performance in scenarios such as social network information diffusion and news dissemination scale prediction. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0045] Figure 1 A flowchart illustrating the information cascade sequence prediction method provided by this invention;

[0046] Figure 2 This is a schematic diagram of the initial state in the example provided by the present invention;

[0047] Figure 3 This is a schematic diagram of the result after the first iteration in the case provided by the present invention;

[0048] Figure 4 This is a schematic diagram of the result after the second iteration in the case provided by the present invention;

[0049] Figure 5 This is a schematic diagram of the structure of the information cascade sequence prediction device provided by the present invention;

[0050] Figure 6This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0053] It should be noted that all actions involving the acquisition of signals, information, or data in this invention are carried out in compliance with the relevant data protection laws and regulations of the locality and with authorization from the owner of the relevant device.

[0054] Example 1

[0055] Please see Figure 1 This invention provides a method for predicting information concatenation sequences, comprising:

[0056] S1. Obtain the initial set of forwarding users of the target information before the prediction time, and the forwarding behavior trajectory of each initial forwarding user before the prediction time, wherein the forwarding behavior trajectory represents the data record of all forwarded information forwarded by the initial forwarding user, and the forwarding behavior trajectory includes the forwarding of the target information;

[0057] Specifically, this step involves the systematic collection and structuring of the initial state of a specific information dissemination process. The target information refers to the digital content entity whose future dissemination scale is to be predicted, such as a tweet with a specific hashtag on a social media platform, a news article on a news client, or an original video on a short video platform. The prediction time is a set point in time; the system needs to estimate the cumulative dissemination scale of the information at the current time at the prediction time. The process of obtaining the initial set of forwarding users is as follows: using the unique content identifier of the target information as the search key, the system queries the platform's historical behavior database for all user account identifiers whose forwarding timestamps are earlier than the prediction time. The set of these identifiers is defined as the initial set of forwarding users. For example, when the target information is a news report about technology trends, the initial set of forwarding users includes all reader accounts that forwarded the article before the prediction time. For each user in this set, the system further extracts their complete behavioral records within the same time frame, arranging them in ascending chronological order to form a forwarding behavior trajectory. This trajectory is an ordered sequence of (forwarded information identifier, forwarding timestamp) tuples; for example, a user might have forwarded sports news, entertainment videos, and the target technology article sequentially over time. This structured representation not only records the key behaviors of users forwarding target information, but more importantly, it captures the user's content selection preferences and behavioral rhythm patterns throughout the observation period, providing a multi-dimensional data foundation for subsequent analysis of user influence.

[0058] S2. For each forwarding information in each forwarding behavior trajectory, obtain the forwarding information diffusion trajectory before the predicted time. The forwarding information diffusion trajectory is used to characterize the data records of all forwarding diffusion users forwarding the forwarding information. The initial forwarding user is included among all forwarding diffusion users.

[0059] Specifically, this step establishes a two-way mapping relationship from the user behavior dimension to the information dissemination dimension. Each forwarding behavior trajectory obtained in step S1 is analyzed, with each forwarding information identifier in the trajectory representing independent digital content that the user has historically disseminated. For example, a technology user's forwarding behavior trajectory may contain multiple product release messages from different technology companies. For each forwarding message in the trajectory, the system performs independent diffusion network reconstruction: using the unique identifier of the information as the retrieval key, it queries the platform's historical database for all user accounts whose forwarding time is earlier than the predicted time and their forwarding timestamps, arranging them chronologically to form the complete propagation chain of the information. For example, for a smartphone review video in a user trajectory, its diffusion trajectory records a time-series propagation path formed from the initial publisher through multiple layers of user forwarding. In particular, since the initial forwarding user's forwarding behavior trajectory contains forwarding records of the information, the forwarding node of the initial forwarding user can necessarily be found in the retrieved diffusion trajectory. This design ensures the closure of the user-information association network. This two-way trajectory acquisition mechanism constructs a cross-network between user behavior and information dissemination, enabling a single user's forwarding behavior to be linked to the overall diffusion pattern of the information it disseminates, while the diffusion trajectory of a single piece of information can reflect the participation characteristics of different disseminators, providing structured data support for subsequent collaborative optimization.

[0060] S3. For each forwarding message, based on the forwarding information diffusion trajectory of the forwarding information and the initial forwarding user influence of each forwarding diffusion user, update the forwarding information influence of the forwarding information to obtain the updated forwarding information influence of each forwarding message.

[0061] Specifically, this step performs a dynamic recalculation of information influence. For the diffusion trajectory of each forwarded message obtained in step S2, the system performs a comprehensive analysis based on the initial influence status of each user involved in the dissemination of that information. For example, the diffusion trajectory of a science video shows that it has been forwarded and spread by users from different fields, including science bloggers, general viewers, and other types of users. The system presets an initial forwarding user influence parameter for each user, which may be calculated based on dimensions such as the user's historical authority indicators (e.g., certification level) and dissemination effectiveness indicators (e.g., historical forwarding adoption rate). The forwarding information influence update process adopts a fusion calculation model. This model takes the time series characteristics of the diffusion trajectory as input and the user influence of each node in the dissemination path as a weighting coefficient. Through a time-series attention mechanism or a dynamic weight allocation algorithm, it calculates the influence estimate of the information at the current dissemination stage. Specifically, when a message is forwarded earlier by a high-influence user, its influence update value will receive a greater gain; conversely, if the dissemination mainly occurs among low-influence user groups, the influence update may be relatively conservative. This update mechanism allows the assessment of information influence to consider not only the breadth of dissemination (length of the diffusion trajectory) but also the quality of dissemination (the distribution of influence at the dissemination nodes), forming a multi-dimensional quantification of the potential for information dissemination.

[0062] S4. For each initial forwarding user, based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, update the forwarding user influence of the initial forwarding user to obtain the updated forwarding user influence of each initial forwarding user.

[0063] Specifically, this step iteratively optimizes user influence. Based on the complete behavioral trajectory of each initial forwarding user obtained in step S1, the system performs a reverse evaluation by combining the latest influence status of each piece of information forwarded by that user in the past (updated by step S3). For example, the 20 pieces of information forwarded by a user in the past three months have obtained updated influence estimates after calculation in step S3. The user influence update process uses a correlation analysis model, which analyzes the content selection patterns (such as preference for technology content) and time behavior characteristics (such as high-frequency forwarding periods) reflected in the user's behavioral trajectory. At the same time, it uses the influence estimate of each piece of information in the trajectory as input features and recalculates the user's influence parameters at the current moment through weighted aggregation or pattern matching algorithms. In particular, when the information forwarded by a user in the past generally has a high updated influence value, the user's influence parameters will increase accordingly, because this correlation indicates that the user has the ability to discover and spread high-potential information; conversely, if the information forwarded by a user generally has a low influence, their influence parameters may be adjusted accordingly. This design realizes a positive feedback mechanism from information quality to disseminator credibility, so that the evaluation of user influence is not only based on its static attributes, but also reflects the effectiveness of its dynamic behavioral choices.

[0064] S5. Take the forwarding user influence updated by each initial forwarding user as the initial forwarding user influence of each initial forwarding user among all forwarding diffusion users, and perform asynchronous loop optimization according to the steps of forwarding information influence update and user influence update until the convergence condition is met, and output the forwarding information influence of the target information finally updated as the prediction result.

[0065] Specifically, this step establishes a closed-loop iterative optimization system. The system reassigns the updated user influence parameter calculated in step S4 to the corresponding initial forwarding user. These users will use the updated influence parameter for calculation when acting as forwarding and diffusion nodes in other information propagation networks in subsequent iterations. For example, in the first round of calculation, user A's initial influence is 0.6, which becomes 0.72 after being updated in step S4; in the second round of calculation, when the diffusion trajectory of other information involves user A, 0.72 will be used as its influence parameter. The iterative optimization process is achieved by alternately executing steps S3 (information influence update) and S4 (user influence update): in each iteration, the influence values ​​of all information are first updated using the current user influence parameter, and then the user influence parameter is recalculated using the updated information influence value, forming an alternating optimization flow of information influence, user influence, and information influence. Convergence judgment is based on monitoring the parameter change rate. For example, the loop terminates when the rate of change of the target information's influence is less than a set threshold (e.g., 0.5%) in two consecutive iterations, or when the preset maximum number of iterations (e.g., 50 rounds) is reached. The final output, representing the target information's forwarding influence, undergoes multiple rounds of mutually reinforcing calculations, integrating the information's own propagation characteristics and network quality factors, and is transformed into a predicted value for future propagation scale. This asynchronous cyclical structure simulates the dynamic process of information value and communicator reputation mutually shaping each other in real-world communication, gradually approaching the system's stable state through alternating parameter updates.

[0066] In some implementations, obtaining the initial set of forwarding users prior to the prediction time and the forwarding behavior trajectory of each initial forwarding user prior to the prediction time includes:

[0067] Extract all user identifiers that have forwarded the target information from the historical forwarding records of the target information and whose forwarding time is before the predicted time, and use the set of user identifiers as the initial forwarding user set;

[0068] For each user in the initial set of forwarding users, extract all forwarding records of the user before the predicted time, and each forwarding record contains at least the information identifier of the forwarded information and the forwarding timestamp;

[0069] The forwarding records are arranged in chronological order according to their forwarding timestamps to form a forwarding behavior trajectory.

[0070] In some embodiments, obtaining the forwarding information diffusion trajectory before the predicted time includes:

[0071] Extract the historical records of the forwarding information that were forwarded by all users before the predicted time, with each historical record containing at least the forwarding user identifier and the forwarding timestamp;

[0072] The historical records are arranged in chronological order of forwarding timestamps to form a trajectory of the spread of forwarded information.

[0073] In some implementations, the initialized forwarding user influence of each forwarding diffusion user is obtained in the following manner:

[0074] Calculate the initial influence of each forwarding user based on the static attribute data and / or dynamic behavior statistics of each forwarding and diffusion user;

[0075] The static attribute data includes at least one of the following: account permission level, number of followers, number of following accounts, and account registration duration.

[0076] The dynamic behavior statistics include at least one of the following: total historical forwarding count, average historical forwarding interval, and average popularity of historical forwarded information.

[0077] Specifically, the system calculates the initial influence of each forwarding user (i.e., all users involved in the diffusion trajectory in step S2). This calculation process integrates two types of data features: static attribute data reflects the user's platform identity characteristics, including account permission level (such as ordinary user, certified user, official media, etc.), number of followers, number of followers, and account registration time, etc., which are long-term stable attributes; dynamic behavioral statistics reflect the user's historical behavior patterns, including historical total number of forwards, historical average forwarding interval, and historical average popularity of forwarded information, etc., which are behavioral derivative indicators. The initial influence calculation can be implemented using models such as weighted linear combination, multilayer perceptron, or rule-based scoring systems. For example, the number of followers can be logarithmically normalized, the account permission level can be mapped to a weight coefficient, and the historical average popularity of forwarded information can be used as a quality factor. Through feature engineering and model calculation, a standardized initial influence value between 0 and 1 can be generated. In particular, for the indicator of historical average popularity of forwarded information, the system needs to trace all the information forwarded by the user in the past and calculate the average number of forwards obtained by these information within their respective observation periods as a reference benchmark. This initialization mechanism provides a reasonable starting point for subsequent iterative optimization, avoiding the instability that may result from completely random initialization, while also preserving the space to further improve the accuracy of parameters through subsequent collaborative updates.

[0078] In some implementations, updating the forwarding influence of the forwarding information based on the forwarding information diffusion trajectory and the initialized forwarding user influence of each forwarding diffusion user, to obtain the updated forwarding influence of each forwarding information, includes:

[0079] The diffusion trajectory of the forwarded information is encoded, and the diffusion pattern features of the forwarded information itself are extracted;

[0080] Aggregate the initial forwarding user influence of each forwarding and diffusion user to generate user influence association features;

[0081] By integrating the diffusion pattern features and the user influence association features, and calculating the forwarding information influence of the forwarding information update through a preset first update model, the forwarding information update is output.

[0082] Specifically, the first step is to encode the spread trajectory of the forwarded information. This trajectory is a time series composed of user identifiers and forwarding timestamps. The encoding process uses time-series modeling techniques to extract multi-dimensional spread pattern features, including the temporal distribution characteristics of forwarding events, the structural features of the propagation path, and the evolutionary pattern of forwarding intensity. Simultaneously, for each user involved in the spread trajectory, initialized user influence parameters are obtained. These parameters are calculated based on the user's static attribute data (such as account permission level and number of followers) and dynamic behavioral statistics (such as total historical forwarding count and average popularity of historical forwarded information). By aggregating these influence parameters, user influence correlation features reflecting the overall quality and structural characteristics of the propagation network are generated. Subsequently, the spread pattern features and user influence correlation features are fused in multiple dimensions, and the fused feature vector is input into a pre-defined first update model. This model learns the interaction patterns between the spread features and user influence features to calculate and output an updated forwarding information influence value. This update process ensures that the assessment of information influence not only considers the temporal dynamic characteristics of its propagation process but also combines the quality distribution of user nodes in the propagation network, forming a comprehensive quantitative representation of the information's propagation potential.

[0083] The first update model can adopt a point-process-based diffusion modeling architecture. This model treats the forwarding information diffusion trajectory as a point-process event sequence, by defining a conditional strength function. To model the probability of a forwarding event occurring, we use Ht to represent the set of historical events up to time t. The specific form of the intensity function can be designed as follows: ,in Represents the basic propagation rate. The influence coefficient, For the moment of forwarding historical events, This is a time-decay kernel function (such as an exponential function or a power-law function). For forwarding users Initialize influence weights. User influence association features are determined through... Parameters are incorporated into the model, while diffusion mode features are encoded using a parameterized form of the time decay kernel function. Model parameters can be optimized using maximum likelihood estimation, and the final quantitative estimate of information influence is obtained through the integral intensity function. ,in This refers to the observation time window.

[0084] The first update model can employ a temporal fusion network based on an attention mechanism. This network includes a diffusion feature encoding layer, an influence feature projection layer, and a cross-attention computation layer. The diffusion feature encoding layer receives the temporal feature vector obtained from the diffusion trajectory encoding and captures the importance differences between different time segments through a temporal attention module. The influence feature projection layer maps the user influence-related features to a dimensional space that matches the diffusion features. The cross-attention computation layer establishes a dynamic weighted association between the two types of features and calculates the weighted fused context representation using a multi-head attention mechanism with the diffusion features as the query vector and the influence features as the key-value vector, respectively. Finally, the influence update value in scalar form is output through a multilayer perceptron regression head.

[0085] In some implementations, updating the forwarding user influence of the initial forwarding user based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each updated forwarding information, to obtain the updated forwarding user influence of each initial forwarding user, includes:

[0086] Encode the forwarding behavior trajectory of the initial forwarding user and extract the behavioral pattern features of the initial forwarding user;

[0087] Aggregate the influence of each forwarding message corresponding to the forwarding behavior trajectory to generate a forwarding information influence association feature;

[0088] By integrating the behavioral pattern features with the forwarding information influence association features, and calculating through a preset second update model, the updated forwarding user influence of the initial forwarding user is output.

[0089] Specifically, this step achieves iterative reassessment of user influence. First, the forwarding behavior trajectory of the initial forwarding user is encoded. This trajectory is a time series composed of tuples containing (forwarded information identifier, forwarding timestamp). The encoding process extracts multi-dimensional behavioral pattern features using behavioral modeling techniques, including the temporal distribution characteristics of forwarded content, the changing patterns of forwarding frequency, the distribution characteristics of forwarded information types, and the periodic patterns of behavioral activity. Simultaneously, for each forwarded piece of information recorded in the forwarding behavior trajectory, its forwarding information influence value, updated in step S3, is obtained, forming a dynamically updated set of influence parameters. By aggregating these influence parameters, a forwarding information influence correlation feature reflecting the historical distribution of user forwarded content quality is generated. Specifically, the weighted average, time-series weighted sum, and influence growth trend of the forwarding information influence can be calculated. Subsequently, the behavioral pattern features and the forwarding information influence correlation features are fused in multiple dimensions. The fused feature vector is input into a preset second update model. This model learns the correlation between user behavioral features and the quality features of their forwarded content, and calculates and outputs the updated forwarding user influence value. This update process allows the assessment of user influence to be based not only on the temporal dynamics of their historical behavioral patterns, but also on the quality evolution of the content they disseminate, forming a dynamic quantitative representation of a user's dissemination capabilities. The second update model can employ the same or different machine learning architectures as the first update model, including but not limited to feature fusion networks based on attention mechanisms, hybrid models of graph neural networks and behavioral sequences, or ensemble learning models such as gradient boosting decision trees.

[0090] The following example illustrates this embodiment.

[0091] I. Case Background and Data Initialization

[0092] The system needs to predict the spread of information m in the future time T, where the current time t < T. The system identifies users who forwarded information m before time t, forming an initial forwarding set {user 1, user 2}. User behavior trajectories are constructed based on historical data: user 1's trajectory is [(information m, t1), (information 1, t2)], and user 2's trajectory is [(information m, t3), (information 2, t4)], where t1 to t4 are timestamps less than t. Further, the spread trajectories of related information are extracted: information 1's spread trajectory is [(user 1, t2), (user 3, t5)], and information 2's spread trajectory is [(user 2, t4), (user 4, t6)].

[0093] II. Influence Initialization and Asynchronous Loop Optimization

[0094] The initial influence of all users is calculated based on static attributes (such as number of followers and verification status) and dynamic behaviors (such as average popularity of historical reposts). The normalized initial values ​​for users 1, 2, 3, and 4 are 0.72, 0.35, 0.45, and 0.28, respectively; the initial value for the influence of all information is set to 0.5. The system enters asynchronous loop optimization, with the initial state as follows. Figure 2 As shown in the diagram. In the first iteration, based on the diffusion trajectory and the initial user influence, the first update model calculated the updated information influence of information m, information 1, and information 2 to be 0.47, 0.68, and 0.42, respectively. Subsequently, based on the user behavior trajectory and the updated information influence, the second update model calculated the updated user influence of user 1 and user 2 to be 0.75 and 0.38, respectively. The results of the first iteration are shown in the diagram. Figure 3 As shown.

[0095] In the second iteration, the first update model recalculates the information influence using the updated user influence, adjusting the information influences of information m, information 1, and information 2 to 0.48, 0.71, and 0.43, respectively. The second update model uses the updated information influence from this iteration to calculate the user influences of user 1 and user 2 to be 0.77 and 0.41, respectively. The results of the second iteration are as follows: Figure 4 As shown in the diagram, this process is used for multiple iterations, with each iteration alternately updating information influence and user influence. In the diagram, the user influence of users 3 and 4 remains constant in each iteration. The system checks the convergence condition after each iteration; in this case, it is set to a rate of change of less than 1% in the information influence of information m. After five iterations, the information influence of information m stabilizes at 0.47, and the rate of change decreases to 0.8%, satisfying the convergence condition. Finally, the information influence of information m is output as 0.47.

[0096] Example 2

[0097] Please see Figure 5 This invention provides an information concatenation sequence prediction device, comprising:

[0098] The first acquisition module is used to acquire the initial set of forwarding users of the target information before the prediction time, and the forwarding behavior trajectory of each initial forwarding user before the prediction time, wherein the forwarding behavior trajectory represents the data record of all forwarded information forwarded by the initial forwarding user, and the forwarding behavior trajectory includes the forwarding of the target information;

[0099] The second acquisition module is used to acquire the forwarding information diffusion trajectory before the predicted time for each forwarding information in each forwarding behavior trajectory. The forwarding information diffusion trajectory is used to characterize the data records of all forwarding diffusion users forwarding the forwarding information. The initial forwarding user is included among all forwarding diffusion users.

[0100] The first update module is used to update the forwarding influence of each forwarding message based on the forwarding information diffusion trajectory of the forwarding message and the initial forwarding user influence of each forwarding diffusion user, so as to obtain the updated forwarding influence of each forwarding message.

[0101] The second update module is used to update the forwarding user influence of each initial forwarding user based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, so as to obtain the updated forwarding user influence of each initial forwarding user.

[0102] The output module is used to take the forwarding user influence updated by each initial forwarding user as the initial forwarding user influence of each initial forwarding user among all forwarding diffusion users, and perform asynchronous loop optimization according to the steps of updating the forwarding information influence and updating the user influence until the convergence condition is met, and output the forwarding information influence of the target information finally updated as the prediction result.

[0103] It should be noted that each module and unit in the information cascade sequence prediction device in this embodiment corresponds one-to-one with each step in the information cascade sequence prediction method in the aforementioned embodiment. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned information cascade sequence prediction method, and will not be repeated here.

[0104] Example 3

[0105] Please see Figure 6 This embodiment provides an electronic device, including at least one processor 601 and a memory 602. Optionally, the device further includes a communication component 603. The processor 601, memory 602, and communication component 603 are connected via a bus 604.

[0106] In a specific implementation, at least one processor 601 executes computer execution instructions stored in memory 602, causing at least one processor 601 to perform the above-described method.

[0107] The specific implementation process of processor 601 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0108] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0109] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0110] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0111] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0112] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0113] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0114] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0115] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0116] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0117] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0118] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0119] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0120] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting information concatenation sequences, characterized in that, include: Obtain the initial set of users who forwarded the target information before the prediction time, and the forwarding behavior trajectory of each initial user before the prediction time, wherein the forwarding behavior trajectory represents the data record of all forwarded information forwarded by the initial user, and the forwarding behavior trajectory includes the forwarding of the target information; For each forwarding information in each forwarding behavior trajectory, the forwarding information diffusion trajectory before the predicted time is obtained. The forwarding information diffusion trajectory is used to characterize the data records of all forwarding diffusion users forwarding the forwarding information. The initial forwarding user is included among all forwarding diffusion users. For each forwarded message, based on the forwarding message diffusion trajectory of the forwarded message and the initialized forwarding user influence of each forwarding diffusion user, the forwarding message influence of the forwarded message is updated to obtain the updated forwarding message influence of each forwarded message. For each initial forwarding user, based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, the forwarding user influence of the initial forwarding user is updated to obtain the updated forwarding user influence of each initial forwarding user. The forwarding user influence updated by each initial forwarding user is taken as the initial forwarding user influence of each initial forwarding user among all forwarding and diffusion users. Asynchronous loop optimization is performed according to the steps of updating the forwarding information influence and updating the user influence until the convergence condition is met. The forwarding information influence of the target information finally updated is output as the prediction result. The initialized forwarding user influence of each forwarding and diffusion user is obtained in the following way: Calculate the initial influence of each forwarding user based on the static attribute data and / or dynamic behavior statistics of each forwarding and diffusion user; The static attribute data includes at least one of the following: account permission level, number of followers, number of following accounts, and account registration duration. The dynamic behavior statistics include at least one of the following: total historical forwarding count, average historical forwarding interval, and average popularity of historical forwarded information; The step of updating the forwarding influence of the forwarding information based on the forwarding information diffusion trajectory and the initialized forwarding user influence of each forwarding diffusion user, to obtain the updated forwarding influence of each forwarding information, includes: The diffusion trajectory of the forwarded information is encoded, and the diffusion pattern features of the forwarded information itself are extracted; Aggregate the initial forwarding user influence of each forwarding and diffusion user to generate user influence association features; By integrating the diffusion pattern features and the user influence association features, the forwarding information influence of the forwarding information update is calculated and output through a preset first update model. The step of updating the forwarding user influence of the initial forwarding user based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, to obtain the updated forwarding user influence of each initial forwarding user, includes: Encode the forwarding behavior trajectory of the initial forwarding user and extract the behavioral pattern features of the initial forwarding user; Aggregate the influence of each forwarding message corresponding to the forwarding behavior trajectory to generate a forwarding information influence association feature; By integrating the behavioral pattern features with the forwarding information influence association features, and calculating through a preset second update model, the updated forwarding user influence of the initial forwarding user is output.

2. The information concatenation sequence prediction method according to claim 1, characterized in that, The acquisition of the initial set of forwarding users prior to the prediction time, and the forwarding behavior trajectory of each initial forwarding user prior to the prediction time, includes: Extract all user identifiers that have forwarded the target information from the historical forwarding records of the target information and whose forwarding time is before the predicted time, and use the set of user identifiers as the initial forwarding user set; For each user in the initial set of forwarding users, extract all forwarding records of the user before the predicted time, and each forwarding record contains at least the information identifier of the forwarded information and the forwarding timestamp; The forwarding records are arranged in chronological order according to their forwarding timestamps to form a forwarding behavior trajectory.

3. The information concatenation sequence prediction method according to claim 1, characterized in that, The step of obtaining the forwarding information diffusion trajectory before the predicted time includes: Extract the historical records of the forwarding information that were forwarded by all users before the predicted time, with each historical record containing at least the forwarding user identifier and the forwarding timestamp; The historical records are arranged in chronological order of forwarding timestamps to form a trajectory of the spread of forwarded information.

4. The information concatenation sequence prediction method according to claim 1, characterized in that, The convergence condition includes at least one of the following: In two consecutive asynchronous loop optimizations, the change in the forwarding information influence of the target information is less than the first threshold. The asynchronous loop optimization count reaches the preset maximum number of rounds.

5. An information cascade sequence prediction device, characterized in that, include: The first acquisition module is used to acquire the initial set of forwarding users of the target information before the prediction time, and the forwarding behavior trajectory of each initial forwarding user before the prediction time, wherein the forwarding behavior trajectory represents the data record of all forwarded information forwarded by the initial forwarding user, and the forwarding behavior trajectory includes the forwarding of the target information; The second acquisition module is used to acquire the forwarding information diffusion trajectory before the predicted time for each forwarding information in each forwarding behavior trajectory. The forwarding information diffusion trajectory is used to characterize the data records of all forwarding diffusion users forwarding the forwarding information. The initial forwarding user is included among all forwarding diffusion users. The first update module is used to update the forwarding influence of each forwarding message based on the forwarding information diffusion trajectory of the forwarding message and the initial forwarding user influence of each forwarding diffusion user, so as to obtain the updated forwarding influence of each forwarding message. The second update module is used to update the forwarding user influence of each initial forwarding user based on the forwarding behavior trajectory of the initial forwarding user and the forwarding information influence of each forwarding information update, so as to obtain the updated forwarding user influence of each initial forwarding user. The output module is used to take the forwarding user influence updated by each initial forwarding user as the initial forwarding user influence of each initial forwarding user among all forwarding diffusion users, and perform asynchronous loop optimization according to the steps of the forwarding information influence update and the user influence update until the convergence condition is met, and output the forwarding information influence of the target information finally updated as the prediction result. The initialized forwarding user influence of each forwarding and diffusion user is obtained in the following way: Calculate the initial influence of each forwarding user based on the static attribute data and / or dynamic behavior statistics of each forwarding and diffusion user; The static attribute data includes at least one of the following: account permission level, number of followers, number of following accounts, and account registration duration. The dynamic behavior statistics include at least one of the following: total historical forwarding count, average historical forwarding interval, and average popularity of historical forwarded information; The first update module is also used for: The diffusion trajectory of the forwarded information is encoded, and the diffusion pattern features of the forwarded information itself are extracted; Aggregate the initial forwarding user influence of each forwarding and diffusion user to generate user influence association features; By integrating the diffusion pattern features and the user influence association features, the forwarding information influence of the forwarding information update is calculated and output through a preset first update model. The second update module is also used for: Encode the forwarding behavior trajectory of the initial forwarding user and extract the behavioral pattern features of the initial forwarding user; Aggregate the influence of each forwarding message corresponding to the forwarding behavior trajectory to generate a forwarding information influence association feature; By integrating the behavioral pattern features with the forwarding information influence association features, and calculating through a preset second update model, the updated forwarding user influence of the initial forwarding user is output.

6. An electronic device, characterized in that, include: At least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method as described in any one of claims 1-4.

7. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, The method as described in any one of claims 1-4 is implemented when the computer program instructions are executed by the processor.

Images