Topic heat prediction method, device and equipment and computer storage medium

Abstract

This invention relates to the field of internet technology and discloses a method, apparatus, device, and computer storage medium for predicting topic popularity. The method includes: inputting network feature information of sub-topics corresponding to a topic to be predicted into a preset topic evolution prediction model to obtain topic evolution feature information corresponding to the topic to be predicted; wherein the sub-topics are obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes regarding the topic to be predicted; and the predicted popularity of the topic to be predicted is determined based on the topic evolution feature information. Through the above method, this invention improves the precision of topic popularity prediction.

Description

Technical Field

[0001] This invention relates to the field of Internet technology, specifically to a method, apparatus, device, and computer storage medium for predicting topic popularity. Background Technology

[0002] Traditional methods for predicting topic popularity can be divided into two categories: The first category uses a mapping function from topic-related Weibo posts and user influence to topic popularity, then builds a model using the factor values ​​from the previous t time points to predict the factor values ​​at time t+1, and finally calculates the topic popularity at time t+1 using the mapping function. The second category uses time series values ​​of topic popularity to fit an evolution model of topic popularity, thereby predicting the topic popularity at any given time.

[0003] The inventors of this application discovered during the implementation of the embodiments of this method that existing topic popularity prediction has the problem of low prediction precision. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention provide a method, apparatus, device, and computer storage medium for predicting topic popularity, in order to solve the problem of low prediction precision in the prior art.

[0005] According to one aspect of the present invention, a method for predicting the popularity of a topic is provided, the method comprising:

[0006] The network feature information of the sub-topic corresponding to the topic to be predicted is input into a preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted.

[0007] The predicted popularity of the topic to be predicted is determined based on the topic evolution characteristics information.

[0008] In one alternative approach, the method further includes: periodically collecting topic network snapshot samples within a historical time interval;

[0009] For each of the topic network snapshot samples, the topic network snapshot sample is divided into multiple sub-topic samples based on the interaction behavior of each user node within the topic network snapshot sample.

[0010] The topic evolution feature samples are determined based on the changes in the network feature information of each sub-topic sample;

[0011] The preset model is trained based on the sub-topic samples and the topic evolution feature samples corresponding to the sub-topic samples to obtain the topic evolution prediction model.

[0012] In one alternative approach, an edge in the topic network sample corresponds to a pair of user nodes that have interacted with the topic network sample; the method further includes:

[0013] The connection density between multiple user node pairs is determined based on the number of edges and edge weights between multiple user nodes within the topic network snapshot sample; wherein, the edge weights are determined based on the intensity of the interaction behavior of the user node pairs;

[0014] The topic snapshot sample is divided into multiple sub-topic samples according to the connection tightness; wherein, each sub-topic sample includes at least one user node pair.

[0015] In an alternative approach, the method further includes:

[0016] For each sub-topic sample in each of the topic network snapshot samples, a matching group network sample corresponding to the sub-topic sample is determined from the sub-topic sample set; wherein, the sub-topic sample set includes all the sub-topic samples in all the topic network snapshot samples;

[0017] The topic evolution feature sample is determined based on the comparison results of the network feature information of the matching group network sample and the sub-topic sample.

[0018] In one optional approach, the network feature information includes at least one of node attribute features and network structure features; wherein, the node attribute features are determined based on the node attributes of the user node in a preset social circle; the network structure features are determined based on the network density of the sub-topic samples; the social circle is obtained by dividing the social network based on the degree of association between the user node and the topic to be predicted; and the social network is constructed based on the social relationships between the user nodes.

[0019] In one optional approach, the topic evolution feature samples include topic network evolution pattern samples and topic network evolution degree samples; the method further includes:

[0020] The topic network evolution pattern sample is determined based on the pairwise comparison results between each of the sub-topic samples within the topic network snapshot sample;

[0021] When the topic network evolution pattern sample is determined to represent the network size change of the sub-topic sample, the topic network evolution degree sample is determined according to the degree of network size change of each sub-topic sample.

[0022] In one alternative approach, the topic evolution prediction model includes an evolution pattern prediction network and an evolution degree prediction network; wherein the evolution pattern prediction network is trained with the sub-topic samples as input and the topic network evolution pattern samples as output; and the evolution degree prediction network is trained with the topic network evolution pattern samples as input and the topic network evolution degree samples as output.

[0023] According to another aspect of the present invention, a topic popularity prediction device is provided, comprising:

[0024] The first determining module is used to input the network feature information of the sub-topic corresponding to the topic to be predicted into a preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted.

[0025] The second determining module is used to determine the predicted popularity of the topic to be predicted based on the topic evolution feature information.

[0026] According to another aspect of the present invention, a topic popularity prediction device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0027] The memory is used to store at least one executable instruction that causes the processor to perform operations as described in any of the embodiments of the topic popularity prediction method.

[0028] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing at least one executable instruction that causes a topic popularity prediction device to perform operations as described in any of the topic popularity prediction method embodiments.

[0029] This invention provides an embodiment of the invention that inputs the network feature information of sub-topics corresponding to a topic to be predicted into a preset topic evolution prediction model to obtain topic evolution feature information corresponding to the topic to be predicted. The sub-topics are obtained by dividing the topic network corresponding to the topic to be predicted. The topic network is constructed based on the interaction behavior of multiple user nodes with respect to the topic to be predicted. The predicted popularity of the topic to be predicted is determined based on the topic evolution feature information. Unlike existing technologies that quantify topic popularity based on topic-related content such as Weibo posts, articles, news, and participating users, this invention does not require estimating topic popularity based on multiple topic-related factors. Instead, it predicts topic popularity based on the evolution trend of sub-topics' interaction behavior with respect to the topic to be predicted. Sub-topics are constructed based on the interaction behavior of multiple users with respect to the current topic to be predicted, linking users together due to their shared interaction with the topic. Therefore, the evolution of each sub-topic in the topic network over time can be used to predict the overall trend of topic popularity. This not only provides the overall popularity of the topic (represented by the number of users participating in the topic) but also allows for the prediction of the popularity changes within each sub-topic, improving the precision of topic popularity prediction.

[0030] The above description is merely an overview of the technical solutions of the embodiments of the present invention. In order to better understand the technical means of the embodiments of the present invention and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0031] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0032] Figure 1 A flowchart illustrating the topic popularity prediction method provided in an embodiment of the present invention is shown.

[0033] Figure 2 This diagram illustrates the structure of a sub-topic in the topic popularity prediction method provided in an embodiment of the present invention.

[0034] Figure 3 This diagram illustrates the structure of the topic evolution prediction model in the topic popularity prediction method provided in this embodiment of the invention.

[0035] Figure 4 A schematic diagram of the topic popularity prediction device provided in an embodiment of the present invention is shown;

[0036] Figure 5 A schematic diagram of the topic popularity prediction device provided in an embodiment of the present invention is shown. Detailed Implementation

[0037] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.

[0038] Before describing the embodiments of the present invention, the prior art and its problems will be further explained:

[0039] Traditional methods for predicting topic popularity can be divided into two categories: One category designs a mapping function from topic-related Weibo posts and user influence to topic popularity. Then, it builds a model using the factor values ​​from the previous t time points to predict the factor values ​​at time t+1. Finally, it calculates the topic popularity at time t+1 using the mapping function. For example, it uses random walks to calculate user influence related to the topic, designing a mapping function based on three factors: the total influence of users who posted topic-related Weibo posts, the number of topic-related Weibo posts, and the topic's popularity at the previous time point. Next, it uses wavelet transform to convert the topic popularity at each time point into a signal set, and trains an ARIMA regression model based on the signal sets from the previous t time points to obtain the signal set at time t+1. Finally, it uses the signal set at time t+1 to perform wavelet reconstruction to obtain the topic popularity at time t+1. The other category uses time-series values ​​of topic popularity to fit an evolution model of topic popularity, thereby predicting the topic popularity at any given time. For each post related to a topic, a function is designed to calculate the topic popularity score based on user opinion trends from comment data. Then, a gamma prediction model function is constructed and trained based on the topic popularity scores of each post for each time period. This gamma prediction model function can then be used to obtain the predicted topic popularity value for any given time point. For example, first, (time, topic popularity value) pairs are counted for the topic at t time points (where topic popularity is the total number of comments). Then, the covariance matrix of a Gaussian process model is calculated based on the relationship between historical topic popularity statistics time points. Finally, a Gaussian process prediction model is constructed based on the covariance matrix and historical topic popularity values. For a given time point, the topic popularity value at that time point is calculated using the difference vector between that time point and the historical topic popularity statistics time points, combined with the historical topic popularity vector, and the constructed Gaussian process prediction model.

[0040] The inventors have discovered that existing topic prediction methods have at least the following problems:

[0041] Existing methods quantify topic popularity by considering factors such as related Weibo posts, articles, news articles, and participating users, or simply by using the total number of related comments. The former, requiring estimation of topic popularity, introduces some bias. Using only the number of comments is also inaccurate. Furthermore, existing methods rely on predictive models built from historical trends in topic popularity to forecast topic popularity at a specific point in time, providing only a macro-level value and failing to offer finer-grained insights into changes in topic popularity. Therefore, a more accurate and granular method for predicting topic popularity is needed.

[0042] Figure 1 A flowchart of a topic popularity prediction method provided by an embodiment of the present invention is shown. This method is executed by a computer processing device. The computer processing device may include a mobile phone, a laptop computer, etc. Figure 1 As shown, the method includes the following steps:

[0043] Step 10: Input the network feature information of the sub-topic corresponding to the topic to be predicted into the preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted.

[0044] In one embodiment of the present invention, the interactive behaviors of user nodes towards the topic to be predicted may include behaviors such as following, commenting, liking, forwarding, and collecting. A topic network is constructed based on whether connections are established between multiple user nodes due to their interactive behaviors towards the topic to be predicted. A sub-topic represents a connection relationship between multiple users who are closely linked by their interactive behaviors towards the topic. For example, two users may like the same sports event and have many interactive behaviors related to that sports event, such as commenting, collecting, and liking each other; these two users belong to a sub-topic corresponding to that sports event. Specifically, a node in a sub-topic corresponds to a user, and the interactions between users towards the topic to be predicted (such as liking, commenting, replying, forwarding, etc.) are represented as an edge between the corresponding nodes.

[0045] For example, two users might like the same sporting event (such as the World Cup) and engage in at least one interactive behavior related to that sporting event, such as commenting on each other's posts, saving them, or liking them. In this case, the two users are connected in the topic network.

[0046] Furthermore, the topic network can be a weighted network, where the weights of the edges connecting user nodes within the topic network can be determined based on the type and frequency of interaction behavior. When partitioning the topic network, the usual method for partitioning weighted networks can be used. Furthermore, it should be considered that the density and strength of connections between nodes in the topic network (mapped to the weights of the edges connecting users) exhibit regional variations within the network.

[0047] Therefore, to further improve the accuracy and precision of predicting topic popularity based on network structure evolution, the topic network can be divided into multiple sub-topics based on its network feature information. This network feature information characterizes the connections between user nodes within each sub-topic, such as the number of user nodes and connecting edges, and the density of the network structure.

[0048] The topic evolution prediction model is used to predict the evolution of each sub-topic in a topic network over time. Based on the evolution of all sub-topics, the network structure information of the topic network at a predetermined future time can be obtained. Since sub-topics are constructed based on user interaction behavior, and user interaction behavior directly reflects the topic's popularity, the future popularity of the topic can be predicted based on the network structure information at that future time. By dividing multiple users into corresponding sub-topics, the topic popularity at the user group level can be predicted more granularly by observing changes in user behavior and connections within each sub-topic. Simultaneously, the topic popularity at the topic network level can be predicted based on multiple or all sub-topics as a whole, thereby improving the precision of topic popularity prediction. Furthermore, when using the number of user participations as a metric for topic popularity, the granular popularity prediction of this embodiment is also more accurate.

[0049] Therefore, in one embodiment of the present invention, the method further includes the following step before step 10:

[0050] Step 101: Periodically collect topic network snapshot samples within historical time intervals.

[0051] Specifically, topic network snapshot samples are used to characterize the network performance of topic network samples at historical collection times. The topic network samples are divided according to time intervals (e.g., 1 day, 1 hour), resulting in topic network snapshot samples NetE1, NetE2, ..., NetE corresponding to each collection time (denoted as 1-T). T .

[0052] Step 102: For each topic snapshot sample, divide the topic snapshot sample into multiple sub-topic samples according to the interaction behavior of each user node in the topic snapshot sample.

[0053] Specifically, sub-topic samples are used to characterize the relationships between user nodes that interact and connect due to the current topic. That is, different interaction patterns for a topic affect the tightness of user connections within that topic. Therefore, for each topic's snapshot sample, NetE... t (t = 1, ..., T), based on the tightness of connections and the intensity of interaction among user nodes in the topic snapshot sample, the topic snapshot sample is divided into multiple sub-topic samples, such as... Figure 2 As shown, a subtopic sample corresponds to a pair of user connections where users interact closely within the current topic. The strength of the interaction can be determined by weighting the type and frequency of the interaction, and the tightness of the connection between user nodes can be determined by the number of connection edges between multiple users. The more connection edges between multiple users, the tighter the internal connection of the subtopic formed by those users.

[0054] Therefore, in one embodiment of the present invention, an edge in the topic network sample corresponds to a pair of user nodes that have interactive behavior with respect to the topic network sample;

[0055] Step 102 further includes: Step 1021: Determine the connection tightness between multiple user node pairs based on the number of edges and edge weights between multiple user nodes in the topic network snapshot sample; wherein, the edge weights are determined based on the intensity of the interaction behavior of the user node pairs.

[0056] Specifically, the strength of an interaction can be determined based on the type and frequency of the interaction between two connected user nodes in the topic network sample. Different types of interaction have different weights, which can be obtained by weighting and summing the interaction types and corresponding frequencies between user nodes. For example, the weight of an edge between user nodes that have liked a post is 1, the weight of an edge between user nodes that have commented, replied, or forwarded a post is 2, and the weight of an edge between user nodes that have both commented (or replied, forwarded) and liked a post is 3. The tightness of the connections between user nodes in the network can be determined by the number of connecting edges between multiple users; the more connecting edges between multiple users, the tighter the connections within the sub-topic network formed by those users.

[0057] Step 1022: Divide the topic snapshot sample into multiple sub-topic samples according to the connection tightness; wherein, each sub-topic sample includes at least one user node pair.

[0058] Specifically, user node pairs with a connection density greater than a preset density threshold can be divided into a sub-topic sample. Optionally, a topic snapshot sample can also be divided into multiple sub-topic samples based on the edge weights between node pairs and the connection density between multiple nodes. Each sub-topic sample includes several closely connected user nodes whose network connection edge weights are greater than a preset threshold. The closeness of the connection can be determined based on whether the connection density is greater than the aforementioned preset density threshold.

[0059] Step 103: Determine the topic evolution feature sample based on the change information of the network feature information of each sub-topic sample.

[0060] In one embodiment of the present invention, step 103 further includes:

[0061] Step 1031: For each sub-topic sample in each topic snapshot sample, determine the matching group network sample corresponding to the sub-topic sample from the sub-topic sample set; wherein, the sub-topic sample set includes all the sub-topic samples in all topic snapshot samples.

[0062] Specifically, the matching group network samples refer to the pre-evolution or post-evolution states of sub-topic samples in other collection periods within the historical time interval. Matching can be performed based on the node identity information of user nodes within the user group samples to obtain the various network states presented by the same sub-topic sample during its evolution throughout the historical time interval. Considering that new nodes may be added or nodes may disappear during the evolution of sub-topic samples, a match can be determined when the matching degree of user node identity information between two sub-topic samples exceeds a preset matching degree threshold. Optionally, the similarity of the group network structures of two sub-topic samples can also be used to determine whether they match.

[0063] Step 1032: Determine the topic evolution feature sample based on the comparison results of the network feature information of the matching group network sample and the sub-topic sample.

[0064] In one embodiment of the present invention, the network feature information includes at least one of node attribute features and network structure features; wherein, the node attribute features are determined based on the node attributes of the user node in a preset social circle; the network structure features are determined based on the network density of the sub-topic samples; the social circle is obtained by dividing the social network according to the degree of association between the user node and the topic to be predicted; and the social network is constructed based on the social relationships between the user nodes.

[0065] Specifically, social networks are constructed based on social connections between users, which can include friendships; one user node in a social network corresponds to one user. Social circles are used to associate social networks with specific topics, representing user groups with shared interests and behaviors related to a particular topic. Specifically, based on the similarity of user nodes' preferences for a given topic, the edges connecting user nodes in the social network can be weighted, and the social network can be divided according to the edge weights to obtain multiple different social circles. A social network is a network that reflects the social relationships between people, where one person is represented as a node, and the relationship between people is represented as an edge between corresponding nodes. For example, if users A and B are friends or mutually following each other, then there is an edge between user nodes A and B. Optionally, each user node has a set of tags representing its interests and hobbies. For example, if user node A's tag set is travel and music, then it means that user A likes travel and music. For a given topic, if there is an edge AB between user nodes A and B in a social network, then the weight of edge AB in the social circle is 1+sim(A,B,T), where sim(A,B,T) represents the similarity between the intersection of the labels of user nodes A and B and the given topic.

[0066] Specifically, network feature information can include 12 features as shown in Table 1. Among them, node attribute features include the number of nodes, node affiliation degree, node persistence, node activity, bridging degree of nodes in the social network, density of the social circle to which the node belongs, degree of nodes in the social network, and edge attribute features corresponding to the edges between nodes; edge attribute features can include the number of internal edges, internal weights, and external weights; network structure features include structural density and structural separation degree. The specific meanings of the above features are shown in Table 1.

[0067]

[0068]

[0069] Table 1 Network Feature Information Table

[0070] In another embodiment of the present invention, the topic evolution feature sample includes a topic network evolution pattern sample and a topic network evolution degree sample; wherein, the topic network evolution pattern sample is used to characterize the evolution type of each sub-topic in the topic network, specifically such as the continuity, growth, shrinkage, or disappearance of the network size, as well as merging or splitting with other sub-topics in the topic network, etc., and the topic network evolution degree sample is used to characterize the degree of evolution and evolution direction information of the sub-topic under its corresponding evolution type, such as the degree of change in network size or the information of the sub-topics it merged into or split into, etc.

[0071] Therefore, step 1032 further includes: step 1041: determining the topic network evolution pattern sample based on the comparison results between each pair of the sub-topic samples within the topic network snapshot sample.

[0072] Specifically, the changes in sub-topic samples are compared pairwise according to the snapshot collection time, thereby obtaining the changes in each sub-topic sample within the historical time interval. Then, the changes in each sub-topic sample are aggregated to obtain the topic network evolution pattern sample.

[0073] Step 1042: When it is determined that the topic network evolution pattern sample represents the network size change of the sub-topic sample, the topic network evolution degree sample is determined according to the degree of network size change of each sub-topic sample.

[0074] Specifically, network size includes the number of user nodes and / or connection edges in the network. When the network size of a subtopic sample changes, such as increasing or decreasing, the topic network evolution sample includes the amount of change in each subtopic sample, such as the amount of increase or decrease.

[0075] Optionally, when the evolution pattern of a subtopic is merging, it can be further determined which subtopics are merged together based on the social network structure. The determination steps are as follows: If the evolution patterns corresponding to subtopics A and B are both merging, then determine whether the nodes in A and B satisfy the following conditions in the social network: 1) they are connected and the edge weight is greater than 1, or 2) they belong to the same social circle. If these conditions are met, then A and B are more likely to have merged into one subtopic; otherwise, A and B are less likely to have merged into one subtopic.

[0076] Step 1033: Train the preset model based on the sub-topic samples and the topic evolution feature samples corresponding to the sub-topic samples to obtain the topic evolution prediction model.

[0077] Specifically, the preset model can be a neural network model. Using sub-topic samples as input and the corresponding topic evolution feature samples as input, the preset model is trained to obtain a topic evolution prediction model. The sub-topic samples are obtained by extracting network features of a preset dimension from data of at least one topic network within a historical time interval; the topic evolution feature samples are obtained by processing data on the changes in the network features of the at least one topic network over time.

[0078] In another embodiment of the present invention, the topic evolution prediction model includes an evolution pattern prediction network and an evolution degree prediction network; wherein, the evolution pattern prediction network is trained with the sub-topic samples as input and the topic network evolution pattern samples as output; the evolution degree prediction network is trained with the topic network evolution pattern samples as input and the topic network evolution degree samples as output.

[0079] Specifically, the structure of the topic evolution prediction model can be found by referring to... Figure 3 ,like Figure 3 As shown, firstly, the feature vectors corresponding to each sub-topic sample included in the sub-topic sample (specifically including the 12-dimensional features in Table 1 above) are used as the input of the evolution pattern prediction network. After processing by the evolution pattern prediction network, the topic network evolution pattern sample including the network evolution pattern of each sub-topic sample is obtained.

[0080] It should be noted that, as Figure 3 As shown, when the network evolution pattern is growth or shrinkage, the output of the aforementioned evolution pattern prediction network is further input into the evolution degree prediction network to obtain the corresponding increment or decrement. The network structures of the evolution pattern prediction network and the evolution degree prediction network can be constructed according to specific circumstances, such as a multi-layer neural network; this embodiment of the invention does not impose any limitations.

[0081] Step 20: Determine the predicted popularity of the topic to be predicted based on the topic evolution feature information.

[0082] Corresponding to the topic evolution prediction model training process in the aforementioned steps, the topic evolution feature information includes the topic network evolution pattern and the topic network evolution degree. The topic network evolution pattern represents the evolution pattern of each sub-topic in the topic network corresponding to the topic to be predicted, and the topic network evolution degree includes the evolution degree of each of the aforementioned sub-topics.

[0083] For each sub-topic in the topic network corresponding to the topic to be predicted, based on the evolution pattern of the topic network corresponding to that sub-topic, and combined with the degree of topic network evolution, the evolution calculation of the topic network of the topic to be predicted at the current moment is performed to obtain the sub-topic at the next moment. Finally, based on all the sub-topics at the next moment, the network state of the topic network corresponding to the topic to be predicted at the next moment is determined, and the topic popularity of the topic to be predicted at the next moment is determined based on the network state. The network state may include the number of user nodes, the number of connections, etc. This embodiment of the invention uses the connection status between all users participating in the topic as a representation of topic popularity, while also considering topic-related comments, reposts, and likes, and does not require estimating topic popularity. Therefore, this embodiment of the invention has smaller bias and higher efficiency. Furthermore, this embodiment of the invention divides the topic network into multiple sub-topics based on user interaction behavior on a given topic. Based on the changes in the sub-topics, more granular topic popularity can be predicted, such as the distribution of the topic in the social network (specifically, which user groups in the social network are paying attention to the topic) and the changes that have occurred in each distribution (i.e., the sub-topics in this embodiment of the invention) (a certain distribution attracts more and more users to participate, two distributions merge into one distribution, etc.), thereby improving the accuracy of topic prediction.

[0084] This invention provides an embodiment of the invention that inputs the network feature information of sub-topics corresponding to a topic to be predicted into a preset topic evolution prediction model to obtain topic evolution feature information corresponding to the topic to be predicted. The sub-topics are obtained by dividing the topic network corresponding to the topic to be predicted. The topic network is constructed based on the interaction behavior of multiple user nodes with respect to the topic to be predicted. The predicted popularity of the topic to be predicted is determined based on the topic evolution feature information. Unlike existing technologies that quantify topic popularity based on topic-related content such as Weibo posts, articles, news, and participating users, this invention does not require estimating topic popularity based on multiple topic-related factors. Instead, it predicts topic popularity based on the evolution trend of sub-topics' interaction behavior with respect to the topic to be predicted. Sub-topics are constructed based on the interaction behavior of multiple users with respect to the current topic to be predicted, linking users together due to their shared interaction with the topic. Therefore, the evolution of each sub-topic in the topic network over time can be used to predict the overall trend of topic popularity. This not only provides the overall popularity of the topic (represented by the number of users participating in the topic) but also allows for the prediction of the popularity changes within each sub-topic, improving the precision of topic popularity prediction.

[0085] Figure 4 A schematic diagram of the topic popularity prediction device provided in an embodiment of the present invention is shown. Figure 4 As shown, the device 30 includes: a first determining module 301 and a second determining module 302. Wherein,

[0086] The first determining module 301 is used to input the network feature information of the sub-topic corresponding to the topic to be predicted into a preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted.

[0087] The second determining module 302 is used to determine the predicted popularity of the topic to be predicted based on the topic evolution feature information.

[0088] The operation process of the topic popularity prediction device provided in this embodiment of the invention is largely the same as that of the aforementioned method embodiment, and will not be described again.

[0089] The topic popularity prediction device provided in this embodiment of the invention obtains topic evolution feature information corresponding to the topic to be predicted by inputting the network feature information of the sub-topic corresponding to the topic to be predicted into a preset topic evolution prediction model; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted; and the predicted popularity of the topic to be predicted is determined based on the topic evolution feature information. Unlike existing technologies that quantify topic popularity based on topic-related content such as Weibo posts, news articles, and participating users, this invention does not require estimating topic popularity based on multiple topic-related factors. Instead, it predicts topic popularity based on the evolution trend of interaction behavior between sub-topics and the topic to be predicted. Sub-topics are constructed based on the interaction behavior of multiple users with the current topic to be predicted, so that users are associated because they have common interactions with the topic to be predicted. Thus, the evolution of each sub-topic in the topic network over time can be used to predict the overall trend of topic popularity. This not only provides the overall popularity of the topic (represented by the number of users participating in the topic), but also allows for the prediction of the popularity changes of the topic in each sub-topic, improving the precision of topic popularity.

[0090] Figure 5 The diagram shows a schematic of the topic popularity prediction device provided in an embodiment of the present invention. The specific implementation of the topic popularity prediction device is not limited by the specific embodiments of the present invention.

[0091] like Figure 5 As shown, the topic popularity prediction device may include: processor 402, communication interface 404, memory 406, and communication bus 408.

[0092] The processor 402, communication interface 404, and memory 406 communicate with each other via communication bus 408. Communication interface 404 is used to communicate with other network elements, such as clients or other servers. The processor 402 executes program 410, specifically performing the relevant steps described above in the embodiment of the topic popularity prediction method.

[0093] Specifically, program 410 may include program code, which includes computer-executable instructions.

[0094] Processor 402 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The topic popularity prediction device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0095] Memory 406 is used to store program 410. Memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0096] Specifically, program 410 can be called by processor 402 to cause the topic popularity prediction device to perform the following operations:

[0097] The network feature information of the sub-topic corresponding to the topic to be predicted is input into a preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted; the predicted popularity of the topic to be predicted is determined based on the topic evolution feature information.

[0098] The operation process of the topic popularity prediction device provided in this embodiment of the invention is largely the same as that of the aforementioned method embodiment, and will not be repeated here.

[0099] The topic popularity prediction device provided in this embodiment of the invention obtains topic evolution feature information corresponding to the topic to be predicted by inputting the network feature information of the sub-topic corresponding to the topic to be predicted into a preset topic evolution prediction model; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted; and the predicted popularity of the topic to be predicted is determined based on the topic evolution feature information. Unlike existing technologies that quantify topic popularity based on topic-related content such as Weibo posts, news articles, and participating users, this invention does not require estimating topic popularity based on multiple topic-related factors. Instead, it predicts topic popularity based on the evolution trend of interaction behavior between sub-topics and the topic to be predicted. Sub-topics are constructed based on the interaction behavior of multiple users with the current topic to be predicted, so that users are associated because they have common interactions with the topic to be predicted. Thus, the evolution of each sub-topic in the topic network over time can be used to predict the overall trend of topic popularity. This not only provides the overall popularity of the topic (represented by the number of users participating in the topic), but also allows for the prediction of the popularity changes of the topic in each sub-topic, improving the precision of topic popularity.

[0100] This invention provides a computer-readable storage medium storing at least one executable instruction. When the executable instruction is executed on a topic popularity prediction device, the topic popularity prediction device performs the topic popularity prediction method in any of the above method embodiments.

[0101] Specifically, the executable instructions can be used to cause the topic popularity prediction device to perform the following operations:

[0102] The network feature information of the sub-topic corresponding to the topic to be predicted is input into a preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted; the predicted popularity of the topic to be predicted is determined based on the topic evolution feature information.

[0103] The operation process of storing executable instructions on the computer storage medium provided in this embodiment of the invention is largely the same as that in the aforementioned method embodiments, and will not be described again.

[0104] The executable instructions stored in the computer storage medium provided in this embodiment of the invention obtain topic evolution feature information corresponding to the topic to be predicted by inputting the network feature information of the sub-topic corresponding to the topic to be predicted into a preset topic evolution prediction model; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted; and the predicted popularity of the topic to be predicted is determined based on the topic evolution feature information. Unlike existing technologies that quantify topic popularity based on topic-related content such as Weibo posts, news articles, and participating users, this invention does not require estimating topic popularity based on multiple topic-related factors. Instead, it predicts topic popularity based on the evolution trend of interaction behavior between sub-topics and the topic to be predicted. Sub-topics are constructed based on the interaction behavior of multiple users with the current topic to be predicted, so that users are associated because they have common interactions with the topic to be predicted. Thus, the evolution of each sub-topic in the topic network over time can be used to predict the overall trend of topic popularity. This not only provides the overall popularity of the topic (represented by the number of users participating in the topic), but also allows for the prediction of the popularity changes of the topic in each sub-topic, improving the precision of topic popularity.

[0105] This invention provides a topic popularity prediction device for executing the above-described topic popularity prediction method.

[0106] This invention provides a computer program that can be called by a processor to cause a topic popularity prediction device to execute the topic popularity prediction method in any of the above method embodiments.

[0107] This invention provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed on a computer, cause the computer to perform the topic popularity prediction method in any of the above method embodiments.

[0108] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of the present invention are not directed to any particular programming language. It should be understood that the content of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0109] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0110] Similarly, it should be understood that, in order to streamline the invention and aid in understanding one or more of the various aspects of the invention, features of the embodiments of the invention are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of the invention. However, this disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.

[0111] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0112] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A topic heat prediction method, characterized in that, The method includes: A topic network snapshot sample is periodically collected within a historical time interval. For each topic network snapshot sample, the connection density between multiple user node pairs is determined based on the number of edges and edge weights between multiple user nodes within the topic network snapshot sample. Each edge in the topic network sample corresponds to a user node pair that interacts with the topic network sample, and the edge weight is determined based on the intensity of the interaction behavior of the user node pair. The topic network snapshot sample is divided into multiple sub-topic samples based on the connection density, wherein each sub-topic sample includes at least one user node pair. For each sub-topic sample in each topic network snapshot sample, a matching group network sample corresponding to the sub-topic sample is determined from a set of sub-topic samples, wherein the set of sub-topic samples includes all sub-topic samples within all topic network snapshot samples. A topic evolution feature sample is determined based on the comparison results of the network feature information of the matching group network sample and the sub-topic sample. A preset model is trained based on the sub-topic sample and the corresponding topic evolution feature sample to obtain a topic evolution prediction model. The network feature information of the sub-topic corresponding to the topic to be predicted is input into a preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein, the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted. The predicted popularity of the topic to be predicted is determined based on the topic evolution characteristics information.

2. The method of claim 1, wherein, The network feature information includes at least one of node attribute features and network structure features; wherein, the node attribute features are determined based on the node attributes of the user node in the preset social circle; the network structure features are determined based on the network density of the sub-topic samples; the social circle is obtained by dividing the social network based on the degree of association between the user node and the topic to be predicted; the social network is constructed based on the social relationships between the user nodes.

3. The method of claim 1, wherein, The topic evolution feature samples include topic network evolution pattern samples and topic network evolution degree samples; the step of determining the topic evolution feature samples based on the comparison results of the network feature information of the matching group network samples and sub-topic samples further includes: The topic network evolution pattern sample is determined based on the pairwise comparison results between each of the sub-topic samples within the topic network snapshot sample; When the topic network evolution pattern sample is determined to represent the network size change of the sub-topic sample, the topic network evolution degree sample is determined according to the degree of network size change of each sub-topic sample.

4. The method of claim 3, wherein, The topic evolution prediction model includes an evolution pattern prediction network and an evolution degree prediction network; wherein, the evolution pattern prediction network is trained with the sub-topic samples as input and the topic network evolution pattern samples as output; the evolution degree prediction network is trained with the topic network evolution pattern samples as input and the topic network evolution degree samples as output. 5.A topic heat prediction apparatus characterized by comprising: The device includes: The first determining module is used to periodically collect topic network snapshot samples within a historical time interval; for each topic network snapshot sample, the connection density between multiple user node pairs is determined based on the number of edges and edge weights between multiple user nodes within the topic network snapshot sample, wherein one edge in the topic network sample corresponds to a user node pair that has interactive behavior with respect to the topic network sample, and the edge weight is determined based on the intensity of the interactive behavior of the user node pair; the topic network snapshot sample is divided into multiple sub-topic samples based on the connection density, wherein each sub-topic sample includes at least one user node pair; for each sub-topic sample in each topic network snapshot sample, the sub-topic sample is determined from the sub-topic sample set. The topic sample is a matching group network sample, wherein the sub-topic sample set includes all sub-topic samples within all topic network snapshot samples; topic evolution feature samples are determined based on the comparison results of the network feature information of the matching group network samples and the sub-topic samples; a preset model is trained based on the sub-topic samples and the topic evolution feature samples corresponding to the sub-topic samples to obtain a topic evolution prediction model; the network feature information of the sub-topic corresponding to the topic to be predicted is input into the preset topic evolution prediction model to obtain the topic evolution feature information corresponding to the topic to be predicted; wherein the sub-topic is obtained by dividing the topic network corresponding to the topic to be predicted; the topic network is constructed based on the interaction behavior of multiple user nodes on the topic to be predicted. The second determining module is used to determine the predicted popularity of the topic to be predicted based on the topic evolution feature information. 6.A topic heat prediction device characterized by comprising: include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation of the topic popularity prediction method as described in any one of claims 1-4.

7. A computer readable storage medium characterized in that, The storage medium stores at least one executable instruction, which, when executed on the topic popularity prediction device, causes the topic popularity prediction device to perform the operation of the topic popularity prediction method as described in any one of claims 1-4.

Images