Social Relationship Prediction Methods, Devices, Equipment and Media

Abstract

This application relates to artificial intelligence, and more particularly to a method, apparatus, device, and medium for predicting social relationships. The method includes: acquiring a social network to be processed, the social network including users and social relationships between users; calculating the similarity between different users in the social network; determining an initial prediction result based on the similarity to determine whether a future social relationship exists between the users; calculating the probability that a future social relationship exists between the users and the probability that a future social relationship does not exist between the users based on the initial prediction result and the similarity; and determining whether a future social relationship exists between the users based on the probability that a future social relationship exists between the users and the probability that a future social relationship does not exist between the users. This method can improve accuracy. It should be noted that the social relationship prediction method, apparatus, device, and medium of this application can be used in the financial field or other fields.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a social relationship prediction method, a product recommendation method, an apparatus, a device, and a medium. Background Technology

[0002] In recent years, the internet has expanded from a technological level to the commercial and social levels, with numerous online brand communities emerging. Based on these platforms, user interaction has intensified, and friends' opinions and suggestions play an increasingly important role in influencing other users' purchasing decisions and companies' attitudes. Therefore, it is crucial for online marketers, as it often encourages purchases. Generally, users with higher influence have a stronger ability to radiate and influence those around them. High-influence individuals are often high-profile figures on online social media or opinion leaders and event organizers within brand user groups; these groups represent a relatively small percentage of the total user base. To accurately assess social influence, Sinan Ami studied the usage of a mobile service by 27 million Yahoo! Messenger users, specifically examining how user usage and recommendations influenced their friends' choices of that service. The results showed that traditional models overestimated the role of influence by as much as seven times, and approximately half of the "influence" was simply the result of friend influence. Therefore, leveraging the high-influence friends of target customers within online brand communities for display marketing can further enhance marketing effectiveness.

[0003] Algorithms for identifying high-influence users in networks are generally categorized into ranking methods based on node neighbor centrality, path centrality, iterative optimization, and node position. However, these studies emphasize the importance of each node without considering network evolution. Link prediction plays a crucial role in friend recommendation and network evolution. Later, a link prediction method for evolving networks was proposed. To predict future node similarity scores, the ARIMA time series prediction model, based on past node similarity scores, was used.

[0004] However, these methods for predicting links in evolutionary networks are all based on modeling changes in past node similarities and external factors. But these studies are mainly based on the one-way nature of the links, that is, they directly calculate whether social relationships exist between users. This single perspective leads to reduced accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, device, and medium for predicting social relationships by separately calculating the probability of future social relationships between users and the probability of future social relationships not existing, and by determining whether future social relationships exist between users based on the probability of future social relationships and the probability of future social relationships not existing.

[0006] Firstly, this application provides a method for predicting social relationships, the method comprising:

[0007] Obtain the social network to be processed, which includes users and the social relationships between users;

[0008] Calculate the similarity between different users in the social network to be processed;

[0009] An initial prediction result is determined based on the similarity to determine whether there will be a social relationship between the users in the future;

[0010] Based on the initial prediction results and the similarity, the probability that users will have a social relationship in the future and the probability that they will not have a social relationship in the future are calculated respectively.

[0011] Based on the probability that social relationships will exist between users in the future and the probability that social relationships will not exist between users in the future, determine whether social relationships will exist between users in the future.

[0012] In one embodiment, calculating the similarity between different users in the social network to be processed includes:

[0013] Obtain the associated users of each user in the social network to be processed;

[0014] The similarity between different users in the social network to be processed is calculated based on the associated users.

[0015] In one embodiment, calculating the similarity between different users in the social network to be processed based on the associated users includes:

[0016] The similarity between different users in the social network to be processed is calculated based on the associated users and at least one similarity calculation rule.

[0017] In one embodiment, after determining whether a future social relationship exists between the users based on the probability that such a relationship will exist and the probability that such a relationship will not exist, the process includes:

[0018] The social network to be processed is evolved based on whether there are future social relationships between users to obtain the target social network;

[0019] Based on the social network to be processed and the target social network, determine the target user corresponding to each user;

[0020] The corresponding target user is pushed to each user.

[0021] In one embodiment, determining the target user corresponding to each user based on the social network to be processed and the target social network includes:

[0022] Calculate the first social breadth of each user in the social network to be processed;

[0023] Calculate the second social breadth of each user in the target social network;

[0024] Users are categorized according to the size of their social breadth based on the first social breadth and the second social breadth.

[0025] The target user corresponding to each user is determined from the classification.

[0026] In one embodiment, the initial prediction result for determining whether a future social relationship exists between the users based on the similarity includes:

[0027] At least one pre-trained base classifier calculates an initial prediction of whether a future social relationship exists between the users based on the similarity.

[0028] In one embodiment, calculating the probability of future social relationships between users and the probability of future non-social relationships based on the initial prediction result and the similarity, respectively, includes:

[0029] The probability of future social relationships between users and the probability of future social relationships are calculated using a pre-trained Bayesian classifier, based on the initial prediction results and the similarity scores.

[0030] In one embodiment, the training method of the base classifier includes:

[0031] Obtain a sample network, which carries annotation results;

[0032] The similarity between sample users is calculated based on the sample network.

[0033] Each base classifier is obtained by training based on the sample user similarity and the annotation results.

[0034] In one embodiment, the training method of the meta-classifier includes:

[0035] The sample user similarity is input into the trained base classifier to obtain the initial prediction result of the sample;

[0036] A Bayesian classifier is obtained by training based on the initial prediction results of each sample, the sample user similarity, and the annotation results.

[0037] Secondly, this application also provides a target user determination device, the target user determination device comprising:

[0038] The first network acquisition module is used to acquire the social network to be processed, which includes users and the social relationships between users.

[0039] The first similarity calculation module is used to calculate the similarity between different users in the social network to be processed;

[0040] The first prediction module is used to determine an initial prediction result based on the similarity to determine whether there is a future social relationship between the users;

[0041] The second prediction module is used to calculate the probability that users will have a social relationship in the future and the probability that they will not have a social relationship in the future based on the initial prediction result and the similarity, respectively.

[0042] The target user setting module is used to determine whether there will be a social relationship between users in the future based on the probability that there will be a social relationship between them in the future and the probability that there will be no social relationship between them in the future.

[0043] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the above embodiments.

[0044] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in any of the above embodiments.

[0045] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the above embodiments.

[0046] The aforementioned social relationship prediction method, apparatus, device, and medium, after acquiring the social network to be processed, calculate the similarity between nodes and first determine an initial prediction result based on the similarity to determine whether a social relationship will exist between users in the future. Based on the initial prediction result and the similarity, the probability of a future social relationship between users and the probability of a future social relationship will not be calculated respectively. The prediction is made from the perspective of the bidirectionality of the link (i.e., the probability of linking or not linking). In other words, this calculation is performed from two perspectives, which is more comprehensive. Therefore, when judging whether a social relationship exists, information from two perspectives is considered, making it more accurate. Attached Figure Description

[0047] Figure 1 This is a diagram illustrating the application environment of a social relationship prediction method in one embodiment.

[0048] Figure 2 This is a flowchart illustrating a social relationship prediction method in one embodiment;

[0049] Figure 3 This is a flowchart illustrating the social relationship prediction method in another embodiment;

[0050] Figure 4 This is a structural block diagram of a social relationship prediction device in one embodiment;

[0051] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0053] The target user determination method and product recommendation method provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on another network server.

[0054] Server 104 can acquire the social network to be processed, which includes users and their social relationships. Server 104 calculates the similarity between different users in the social network to determine an initial prediction of whether future social relationships exist between users based on the similarity. Based on the initial prediction and similarity, it calculates the probability of future social relationships between users and the probability of future non-social relationships. This determination of future social relationships based on the bidirectional nature of the link (i.e., the probability of linking versus not linking) improves accuracy. In other words, this calculation considers two perspectives, providing a more comprehensive view and making the judgment of social relationship existence more accurate. Subsequently, server 104 determines the target user for each user based on the social network to be processed and the evolved target social network; it then pushes the target user to the corresponding user, thereby improving the accuracy of the recommendation.

[0055] The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle systems. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0056] In one embodiment, such as Figure 2 As shown, a method for determining target users is provided, which is then applied to... Figure 1 Taking the server in the example, the following steps are included:

[0057] S202: Obtain the social network to be processed, which includes users and the social relationships between users.

[0058] Specifically, the social network to be processed can refer to the social network between users, which includes users and the social relationships between users. In a general social network, users are represented by nodes, and the social relationships between users are represented by the links between nodes that represent social relationships.

[0059] In practical applications, each social network to be processed can be pre-generated, or it can be obtained based on the user's request or the server's scheduled task after a product recommendation request is triggered. For example, the corresponding social network to be processed can be obtained based on the user's request or the user identifier in the scheduled task.

[0060] S204: Calculate the similarity between different users in the social network to be processed.

[0061] It should be noted that the similarity between users, i.e., the similarity between nodes, includes similarity based on local information, similarity based on global information, and similarity based on random walks. However, the similarity algorithms based on global information and random walks require the topological structure information of the entire network, and therefore are not suitable for networks with a large number of nodes, such as online brand communities. Therefore, this embodiment uses similarity based on local information to ensure computational efficiency.

[0062] Specifically, similarity can be calculated based on different similarity indices, including at least one of the following: Salton similarity index, Sorenson similarity index, HPI similarity index, HDI similarity index, LHN similarity index, PA similarity index, RA similarity index, AA similarity index, and Jaccard similarity index.

[0063] The AA similarity index characterizes the degree values ​​of common neighbors of overlapping node pairs from different products. It assigns the reciprocal of the logarithm of the node's degree (1 / (lg k)) as a weight to each overlapping node. The HDI and HPI similarity indices assume a relationship between the number of common neighbors of a link and an overlapping node pair and the degree of the overlapping node pair itself. The numerator is always the number of common neighbors, and the denominators are the larger and smaller degrees of the overlapping nodes, respectively. Similarly, the Sorenson, Salton, and LHN similarity indices also represent the relationship between the number of common neighbors of a link and an overlapping node pair and the degree of the overlapping node pair itself. The Jaccard similarity index represents the number of common neighbors of an overlapping node pair divided by the union of the overlapping node pair's neighbors. It's important to note that the union of the overlapping node pair's neighbors is not the sum of the degrees of the node pairs, but rather the sum of the degrees of the node pairs minus the number of common neighbors. The RA similarity index, based on the degree values ​​of the common neighbors of overlapping node pairs, considers common neighbors as a medium of transmission from a resource perspective. The overlapping nodes are assigned values ​​using the reciprocal of the degree of their common neighbor nodes. The PA similarity index indicates that the link is closely related to the degree of each overlapping node itself. The higher the degree of each of the two overlapping nodes in the network, the higher the probability that the two overlapping nodes will form a link.

[0064] S206: Initial prediction results based on similarity to determine whether there will be a social relationship between users in the future.

[0065] The initial prediction result is a preliminary prediction of whether social relationships will exist between users in the future. It is a rough prediction and can be processed using relatively simple methods.

[0066] In one embodiment, determining the initial prediction result of whether there is a future social relationship between users based on similarity includes: calculating the initial prediction result of whether there is a future social relationship between users based on similarity using at least one pre-trained base classifier.

[0067] Specifically, the base classifier makes a coarse prediction based on similarity, which can also be seen as an initial prediction of whether social relationships exist between users. This prediction determines how social relationships between users change over time. The initial prediction results include whether social relationships exist between users over time, that is, whether there are links between nodes or not.

[0068] The base classifier can include at least one of KNN, SVM, random forest, logistic regression, discriminant analysis, etc.

[0069] S208: Calculate the probability that users will have a social relationship in the future and the probability that they will not have a social relationship in the future based on the initial prediction results and similarity.

[0070] S210: Determine whether a social relationship will exist between users in the future based on the probability that a social relationship will exist between users in the future and the probability that a social relationship will not exist between users in the future.

[0071] The process involves calculating the probability of future social relationships between users and the probability of future non-social relationships based on the initial prediction results and similarity scores. This second, and more precise, prediction is made regarding the existence of social relationships between users. To improve accuracy, the server calculates the probabilities of future social relationships and non-social relationships separately. Specifically, it first calculates the probability of future social relationships based on the initial prediction results and similarity scores, and then calculates the probability of future non-social relationships based on the same scores. In other embodiments, these two processes can be performed in parallel.

[0072] In one embodiment, the probability of future social relationships between users and the probability of future non-social relationships are calculated based on the initial prediction results and similarity, respectively, including: using a pre-trained Bayesian classifier, and calculating the probability of future social relationships between users and the probability of future non-social relationships based on the initial prediction results and similarity, respectively.

[0073] Specifically, the meta-classifier can be a Bayesian classifier. In order to further improve the prediction accuracy and predict links from multiple perspectives, a Bayesian classifier is added in this embodiment to calculate the probability of linking (Yes) and not linking (No).

[0074] In this way, the server determines whether a social relationship will exist between users in the future based on the probability that a social relationship will exist between users in the future and the probability that a social relationship will not exist between users in the future. For example, when P(Yes) > P(No), the link is predicted to exist, that is, a social relationship will exist between users in the future; otherwise, the link will not exist, that is, a social relationship will not exist between users in the future.

[0075] In this process, the server inputs the initial prediction results and similarity scores from each base classifier into the meta-classifier to obtain the target prediction result. This involves a second prediction to calculate the probability of linking or not linking users, which corresponds to the probability of future social relationships between users (as mentioned above) and the probability of future non-social relationships. In other embodiments, new input data can be generated based on the initial prediction results and similarity scores of the base classifiers, and then this new input data is input into the meta-classifier to obtain the target prediction result.

[0076] Specifically, in combination Figure 3 As shown, the evolution trend of the entire network is demonstrated through bidirectional prediction of links. The model in this embodiment includes a two-layer classifier: level-0 is the base classifier, and level-1 is the meta-classifier. The output of the base classifier is used as the input of the meta-classifier, whose task is to reasonably combine the output set and correct the classification errors of the base classifier. Therefore, the first step in determining the target user is to use the base learner to predict the dataset. Next, the output of the base classifier is used as the input of the meta-classifier, that is, the predicted information from the dataset output and the true classification results of the training data are integrated into a single dataset. This new dataset is then used as the training dataset for a new learner, and the meta-classifier is used to solve this problem. Specifically, combined with... Figure 3First, a simpler classifier is used to achieve fast classification. This chapter selects five classifiers as base classifiers: KNN, SVM, Random Forest, Logistic Regression, and Discriminant Analysis. Then, the initial prediction results are written into the original dataset as new input. Since the Bayesian classifier can calculate the probability of an object belonging to a certain class, this chapter uses a Bayesian classifier to perform a second-level prediction on the new dataset, calculating the probability of links being connected or not connected. Furthermore, because the model building process is relatively complex, a similarity metric based on local information is used in the learning process. Figure 3 The diagram shows the structure of the decision support system for node selection, where the dashed line represents BLPM. When P(Yes) > P(No), the link is predicted to exist; otherwise, it is not. From the perspective of network evolution, three types of influential users who can be used for friend display marketing are identified.

[0077] The aforementioned target user identification method, after acquiring the social network to be processed, calculates the similarity between nodes and first determines an initial prediction result based on the similarity to determine whether there will be a social relationship between users in the future. Based on the initial prediction result and the similarity, it calculates the probability that there will be a social relationship between users in the future and the probability that there will not be a social relationship in the future, respectively. The prediction is made from the perspective of the bidirectionality of the link (i.e., the probability of linking or not linking). In other words, it performs calculations from two perspectives, which is more comprehensive. Therefore, when judging whether there is a social relationship, it also considers information from two perspectives, thus making it more accurate.

[0078] In one embodiment, calculating the similarity between different users in the social network to be processed includes: obtaining the associated users of each user in the social network to be processed; and calculating the similarity between different users in the social network to be processed based on the associated users.

[0079] In one embodiment, calculating the similarity between different users in the social network to be processed based on associated users includes: calculating the similarity between different users in the social network to be processed based on associated users and at least one similarity calculation rule.

[0080] Specifically, associated users refer to users who have social relationships with other users. In the social network to be processed, associated users are represented by neighboring nodes, which are nodes currently connected to other nodes in the social network. The degree of a node is the number of its neighboring nodes. The server can calculate the similarity between nodes, i.e., the similarity between users, based on the degree of each node or its neighboring nodes.

[0081] Similarity calculation rules can refer to the calculation formulas for various similarity indicators. The server can calculate the similarity between users based on these formulas. Furthermore, multiple similarity calculation rules exist, allowing the server to calculate multiple similarity scores between two users, thus considering various aspects of the problem and improving the accuracy of prediction results.

[0082] The Salton similarity index is calculated using the following formula:

[0083]

[0084] Where Γ(x) represents the neighboring nodes of node x in the social network to be processed, Γ(y) represents the neighboring nodes of node y in the social network to be processed, k(x) represents the degree of node x, and k(y) represents the degree of node y. Γ(x)∩Γ(y) represents the common neighboring nodes of nodes x and y.

[0085] The Sorenson similarity index is calculated using the following formula:

[0086]

[0087] The formula for calculating the HPI similarity index is as follows:

[0088]

[0089] Here, min{k(x),k(y)} is the minimum degree of node x and node y.

[0090] The formula for calculating the HDI similarity index is as follows:

[0091]

[0092] Among them, max{k(x),k(y)} is the minimum degree of node x and node y.

[0093] The formula for calculating the LHN similarity index is as follows:

[0094]

[0095] The formula for calculating the PA similarity index is as follows:

[0096]

[0097] The formula for calculating the RA similarity index is as follows:

[0098]

[0099] Where z is a common neighbor of node x and node y.

[0100] The formula for calculating the AA similarity index is as follows:

[0101]

[0102] in, This is a weight value assigned to each node based on the degree of the common neighbor nodes of node x and node y. This weight value is equal to one-logarithm of the degree of the common neighbor node.

[0103] The formula for calculating the Jaccard similarity index is as follows:

[0104]

[0105] Where |Γ(x)∪Γ(y)| is the union of the neighboring nodes of node x and node y.

[0106] In one embodiment, after determining whether a social relationship exists between users in the future based on the probability that a social relationship will exist between users in the future and the probability that a social relationship will not exist in the future, the process includes: evolving the social network to be processed based on whether a social relationship exists between users in the future to obtain a target social network; determining the target user corresponding to each user based on the social network to be processed and the target social network; and pushing the corresponding target user to each user.

[0107] In one embodiment, determining the target user corresponding to each user based on the social network to be processed and the target social network includes: calculating the first social breadth of each user in the social network to be processed; calculating the second social breadth of each user in the target social network; classifying each user according to the size of the social breadth based on the first social breadth and the second social breadth; and determining the target user corresponding to each user from the classification.

[0108] The target social network is generated based on the predicted future social relationships between users, which is the evolution of the social network to be processed.

[0109] Specifically, considering that overlapping nodes contain more information than other nodes, and that overlapping nodes are more familiar with the products to be displayed and marketed and share common interests with the target customers, all influential nodes are defined as overlapping nodes, i.e., target users. Comparing the network structure before and after server evolution, node degree is a commonly used and important metric for node performance. However, high-degree, high-influence display users are not necessarily friends with the target customers. Therefore, in this embodiment, influential display users are divided into three categories: general influential display users, potential influential display users, and currently popular influential display users. Based on their influence, the users most likely to become friends with the target customers are selected from these three categories and recommended to them, achieving personalized marketing through the influence of high-influence friends.

[0110] Specifically, determining the target user corresponding to each user based on the social network to be processed and the target social network includes: calculating the first social breadth of each user in the social network to be processed; calculating the second social breadth of each user in the target social network; classifying each user according to the size of the social breadth based on the first social breadth and the second social breadth; and determining the target user corresponding to each user from the classification.

[0111] The three types of users are shown below, where μ is the threshold:

[0112] The first category is the set of users displaying general influence, E1: before evolution, degree d1 < μ; after evolution, degree d2 < μ. The second category is the set of users displaying potential influence, E2: before evolution, degree d1 < μ; after evolution, degree d2 ≥ μ. The third category is the set of users displaying current popularity, E3: before evolution, degree d1 ≥ μ.

[0113] In this embodiment, high-influence users within the brand community who are likely to become friends with the target customer are identified. Generally, nodes with higher degrees are considered to have greater influence. Overlapping nodes in the brand community that like two or more product categories act as bridges between these categories, demonstrating familiarity with the products being promoted and sharing similar interests with the target customer. Therefore, in this embodiment, overlapping nodes with higher degrees before and after evolution are defined as high-influence nodes. The target user determination model combines local similarity metrics between node pairs with classification algorithms such as KNN, SVM, random forest, logistic regression, and discriminant analysis. A Bayesian classifier is used to predict links from both linked and unlinked perspectives. The network before and after evolution is compared to identify high-influence users among the overlapping nodes in the brand community. Display nodes that are likely to become friends with the target customer and possess high influence are recommended to these users, achieving personalized marketing based on the influence of high-influence friends.

[0114] In the above embodiments, it was found that influential users in overlapping nodes of the brand community can help influence target consumers through friend-based marketing for a certain type of product. Recommending these highly influential users to target customers can more accurately improve their influence on target customers and increase purchases. To address this issue, this embodiment identifies three types of influential display users in the brand community. First, the similarity score of node pairs is calculated based on the similarity of network links. Combining five base classification models—KNN, SVM, Random Forest, Logistic Regression, and Discriminant Analysis—five classification prediction results for the links are calculated. Then, the five results are applied to a Bayesian classifier to obtain the probability of predicting a link (Yes) and no link (No). When P(Yes) > P(No), the link is predicted to exist; otherwise, it is not linked. Based on the comparison of the network structure before and after the evolution of the display node selection decision support system, the three types of influential users are identified from the perspective of network evolution.

[0115] In one embodiment, determining the initial prediction result of whether there is a future social relationship between users based on similarity includes: calculating the initial prediction result of whether there is a future social relationship between users based on similarity using at least one pre-trained base classifier.

[0116] Specifically, the base classifier processes similarity metrics to obtain initial prediction results, and the base classifier is trained using KNN, SVM, random forest, logistic regression, and discriminant analysis.

[0117] The basic idea of ​​KNN is that if a sample's k nearest neighbors in the feature space mostly belong to a certain class, then that sample also belongs to that class and possesses the characteristics of samples in that class. The specific steps of KNN are as follows: During prediction, in the training sample set S... T* Find the predicted sample S p* The K most similar nearest neighbor links are determined here using Euclidean distance. To improve processing speed, K is set to 3. Other values ​​can be chosen in other embodiments; no specific limitation is made here. Then, the set Y = {Y1, Y2, Y3} of these three nearest neighbors is found, and the most similar Y is selected using a voting principle. i As S p* The prediction results. Assume S p* The prediction result based on the KNN classifier is w1.

[0118] In an SVM classifier, for a given training sample set S T* The hyperplane is denoted as (w·x) + b = 0. For the linearly inseparable case, to improve algorithm efficiency, this embodiment uses a nonlinear Gaussian function (RBF) K(x). i ,x)=exp(-‖xxi || 2 / δ 2 Using ) as the kernel function, for an input vector z, the optimal classification function can be obtained as: Where a, b, and δ are constants. Assume S p* The prediction result based on the SVM classifier is w2, where w and b are parameters in the hyperplane, and the solution to the optimal problem at the saddle point satisfies that the partial derivatives of w and b are 0. x is a training sample, where x i Let y represent the i-th training sample. i Let x be the label of the i-th training sample, Φ(z) be the transformation that converts the training sample x from the input space to the feature space, and K(x) be the label of the i-th training sample. i ,z)=Φ(x i )·Φ(x).

[0119] Random forest is an ensemble classifier consisting of a set of decision tree classifiers. The steps to generate a random forest are as follows: From S T* In this embodiment, a random resampling method is applied to randomly sample K new samples with replacement, and K decision trees are constructed from these samples. In this embodiment, K is chosen to be 50, but other values ​​can be selected in other embodiments. The decision trees are constructed by calculating the information gain ratio of the local similarity index mentioned above, selecting the attribute with the largest information gain ratio as the root node, and recursively building the branches of the tree using the same method until the samples in all branch nodes select the same result. The generated multiple trees are combined into a random forest, and the random forest is used to analyze S. p* The classification is performed, and the classification result depends on the number of votes cast by the tree classifier. Assume S... p* The prediction result based on the random forest classifier is w3.

[0120] Logistic regression is designed for binary classification problems, and link prediction is a typical binary classification problem. Let the conditional probability P{Y=1|x)=p be the probability of a link occurring based on the nine ratings. Then the logistic regression model can be expressed as:

[0121]

[0122] Where g(x) = β0 + β1x1 + β2x2 + ... + β9x9, β0 is the intercept term, and β = (β1, β2, ..., β9) are the regression coefficients of the independent variables. The probability of the dependent variable Y = 1 (i.e., the probability of the link existing) is estimated based on π(x). Let S... T* Given m observations with values ​​{y1, y2, ..., y... m Therefore, the likelihood function for the m observations is:

[0123]

[0124] Taking the natural logarithm of both sides of the above equation yields the log-likelihood function. Further differentiation provides the intercept term and regression coefficients of the model. Substituting these parameters into equation (3) establishes the logistic regression prediction model. The test set S... p* When applied to a well-established logistic regression prediction model, a probability greater than 0.5 is considered to indicate that Y = 1, meaning a link exists. Assume S... p* The prediction result based on the random forest classifier is w4.

[0125] The basic idea of ​​discriminant analysis algorithm is to use the training sample set S T* The centroid coordinates of each category are obtained, and then the test set S is... p* Calculate their distance from the centroid of each category, and then assign them to the category closest to them.

[0126] Let S T* Two populations, Y1 (Y=0) and Y2 (Y=1), have expected vectors u1 and u2, respectively, and covariance matrices Σ1 and Σ2. Define a test set S. p* The distances from X to Y1 and Y2 in the equation are:

[0127]

[0128] The calculated distances d(X,Y1) and d(X,Y2) can be used to determine whether X is a link or not, according to the following criteria.

[0129]

[0130] Assume S p* The prediction result based on the random forest classifier is w5.

[0131] In one embodiment, the probability of future social relationships between users and the probability of future non-social relationships are calculated based on the initial prediction results and similarity, respectively, including: using a pre-trained Bayesian classifier, and calculating the probability of future social relationships between users and the probability of future non-social relationships based on the initial prediction results and similarity, respectively.

[0132] Specifically, the task of the meta-classifier is to reasonably combine the output set and correct the classification errors of the base classifier. To further improve prediction accuracy and predict links from multiple perspectives, a Bayesian classifier is added in this embodiment to calculate the probability of link (Yes) and no link (No). When P(Yes) > P(No), the link is predicted to exist; otherwise, no link is predicted.

[0133] The server merges the initial prediction results of each base classifier with the similarity calculated above, and inputs the merged data into the meta-classifier trained based on the Bayesian classifier to obtain the target prediction result, namely the probability that users will have social relationships in the future and the probability that they will not have social relationships in the future.

[0134] In one embodiment, the training method of the base classifier includes: obtaining a sample network carrying annotation results; calculating the sample user similarity between sample users based on the sample network; and training based on the sample user similarity and annotation results to obtain each base classifier.

[0135] In one embodiment, the training method of the meta-classifier includes: inputting the sample user similarity into the trained base classifier to obtain the initial prediction result of the sample; and training the Bayesian classifier based on the initial prediction result of each sample, the sample user similarity, and the annotation result.

[0136] Specifically, this embodiment mainly describes the training process of the base classifier and meta classifier, which mainly includes:

[0137] First, a predetermined number of links from the entire social network to be processed is randomly selected, for example, 80%, as the training set S. T The remaining portion, or 20%, is used as the test set S. p Suppose that the set of standardized link scores and the true link results from the training data calculated by the local similarity link prediction algorithm is S, and its mathematical expression is:

[0138]

[0139] Among them, attribute X ij Y represents the score of the i-th link on the j-th local similarity metric. i This indicates the corresponding link result, i.e., Yes or No, with 1 representing Yes and 0 representing No.

[0140] Then, take a portion of S from S T* As the training set for five base classifiers—KNN, SVM, Random Forest, Logistic Regression, and Discriminant Analysis—the remaining S... p* As a test set, for S p* The link in the middle will yield five prediction results.

[0141] Let T = {w1, w2, ..., w5, Y} be the training sample set obtained by five base classifiers: KNN, SVM, random forest, logistic regression, and discriminant analysis. Here, w1 represents the prediction result of KNN on the link (1 or 0), and w2, w3, w4, and w5 respectively represent the prediction results of SVM, random forest, logistic regression, and discriminant analysis on the link.

[0142] The training sample set is divided into two classes, denoted as Y = {Y1, Y2}. Then, for each class Y... i The prior probability is P(Y) i ), i = 1, 2, its value is Y i The number of samples in each class is divided by the total number of samples in the training set, n. Based on the training set, the server calculates w for each class. j In class Y i The probability of occurrence P(w) j |Y i For a new sample d, it belongs to Y. i The conditional probability of class is P(d|Y) i Y i The posterior probability of class Y is P(Y). i |d)

[0143]

[0144] Since P(d) is a constant for all classes, it can be ignored. Therefore, equation (5-6) simplifies to P(Y). i |d)∝ P(d|Y i )P(Y i ), where d is composed of the prediction results of 5 base classifiers, i.e., d = (w1, w2, ..., w5), then we get

[0145]

[0146] In the formula P(w j |Y i ) indicates w j In class Y i The probability of occurrence. Using the method above, we can calculate the two posterior probabilities P(Yes) and P(No) of the test sample, thus enabling bidirectional prediction of the link.

[0147] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0148] Based on the same inventive concept, this application also provides a target user determination device and a product recommendation device for implementing the target user determination method and product recommendation method described above. The solution provided by this device is similar to the implementation scheme described in the above method. Therefore, the specific limitations in one or more embodiments of the target user determination device and product recommendation device provided below can be found in the limitations of the target user determination method and product recommendation method described above, and will not be repeated here.

[0149] In one embodiment, such as Figure 5 As shown, a target user determination device is provided, comprising: a first network acquisition module 501, a first similarity calculation module 502, a first prediction module 503, a second prediction module 504, and a social relationship determination module 505, wherein:

[0150] The first network acquisition module 501 is used to acquire the social network to be processed, which includes users and the social relationships between users.

[0151] The first similarity calculation module 502 is used to calculate the similarity between different users in the social network to be processed;

[0152] The first prediction module 503 is used to determine the initial prediction result of whether there is a social relationship between users in the future based on similarity.

[0153] The second prediction module 504 is used to calculate the probability that users will have a social relationship in the future and the probability that they will not have a social relationship in the future based on the initial prediction results and similarity.

[0154] The social relationship determination module 505 is used to determine whether a social relationship exists between users in the future based on the probability that a social relationship will exist between the users in the future and the probability that a social relationship will not exist between the users in the future.

[0155] In one embodiment, the first similarity calculation module 502 includes:

[0156] The associated user acquisition unit is used to acquire the associated users of each user in the social network to be processed.

[0157] The similarity calculation unit is used to calculate the similarity between different users in the social network to be processed based on the associated users.

[0158] In one embodiment, the similarity calculation unit is further configured to calculate the similarity between different users in the social network to be processed based on associated users and at least one similarity calculation rule.

[0159] In one embodiment, the above-described apparatus further includes:

[0160] The evolution module is used to evolve the social network to be processed based on whether there are future social relationships between users, so as to obtain the target social network;

[0161] The target user determination module is used to determine the target user for each user based on the social network to be processed and the target social network.

[0162] The push module is used to push the corresponding target user to each user.

[0163] In one embodiment, the target user determination module includes:

[0164] The first social breadth calculation unit is used to calculate the first social breadth of each user in the social network to be processed.

[0165] The second social breadth calculation unit is used to calculate the second social breadth of each user in the target social network.

[0166] The classification unit is used to classify each user according to the size of their social breadth based on a first social breadth and a second social breadth.

[0167] The target user determination unit is used to determine the target user corresponding to each user from the categories.

[0168] In one embodiment, the first prediction module 503 is used to calculate an initial prediction result based on similarity to determine whether there is a future social relationship between users using at least one pre-trained base classifier.

[0169] In one embodiment, the second prediction module 504 is used to calculate the probability that users will have a future social relationship and the probability that they will not have a future social relationship, respectively, based on a pre-trained Bayesian classifier and the initial prediction result and similarity. In one embodiment, the above apparatus further includes:

[0170] The sample network acquisition module is used to acquire sample networks, which carry annotation results.

[0171] The sample similarity index calculation module is used to calculate the sample user similarity between sample users based on the sample network.

[0172] The first training module is used to train various base classifiers based on the sample user similarity and annotation results.

[0173] In one embodiment, the above-described apparatus further includes:

[0174] The third prediction module is used to input the sample user similarity into the trained base classifier to obtain the initial prediction result of the sample;

[0175] The second training module is used to train a Bayesian classifier based on the initial prediction results of each sample, the sample user similarity, and the annotation results.

[0176] Each module in the aforementioned social relationship prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0177] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When executed by the processor, the computer program implements a social relationship prediction method.

[0178] Those skilled in the art will understand that Figure 5The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0179] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to perform the following steps: acquiring a social network to be processed, the social network including users and social relationships between users; calculating the similarity between different users in the social network to be processed; determining an initial prediction result based on the similarity to determine whether there will be a social relationship between users in the future; calculating the probability that there will be a social relationship between users in the future and the probability that there will not be a social relationship between users in the future based on the initial prediction result and the similarity, respectively.

[0180] In one embodiment, the calculation of similarity between different users in a social network to be processed, implemented by the processor executing a computer program, includes: obtaining associated users of each user in the social network to be processed; and calculating the similarity between different users in the social network to be processed based on the associated users.

[0181] In one embodiment, the similarity calculation between different users in a social network to be processed based on associated users, implemented by the processor when executing a computer program, includes: calculating the similarity between different users in a social network to be processed based on associated users and at least one similarity calculation rule.

[0182] In one embodiment, after the processor executes a computer program to determine whether a social relationship exists between users in the future based on the probability that a social relationship will exist between users in the future and the probability that a social relationship will not exist in the future, the process includes: evolving the social network to be processed based on whether a social relationship exists between users in the future to obtain a target social network; determining the target user corresponding to each user based on the social network to be processed and the target social network; and pushing the corresponding target user to each user.

[0183] In one embodiment, the process of determining the target user corresponding to each user based on the social network to be processed and the target social network when the processor executes the computer program includes: calculating a first social breadth for each user in the social network to be processed; calculating a second social breadth for each user in the target social network; classifying each user according to the size of the social breadth based on the first social breadth and the second social breadth; and determining the target user corresponding to each user from the classification.

[0184] In one embodiment, the initial prediction result of determining whether there is a future social relationship between users based on similarity, implemented by the processor when executing a computer program, includes: calculating the initial prediction result of whether there is a future social relationship between users based on similarity using at least one pre-trained base classifier.

[0185] In one embodiment, the process of the processor executing a computer program to calculate the probability of future social relationships between users and the probability of future non-social relationships based on the initial prediction results and similarity includes: using a pre-trained Bayesian classifier to calculate the probability of future social relationships between users and the probability of future non-social relationships based on the initial prediction results and similarity.

[0186] In one embodiment, the training method of the base classifiers involved when the processor executes the computer program includes: acquiring a sample network carrying labeled results; calculating the sample user similarity between sample users based on the sample network; and training based on the sample user similarity and the labeled results to obtain each base classifier.

[0187] In one embodiment, the training method of the meta-classifier involved when the processor executes the computer program includes: inputting the sample user similarity into the trained base classifier to obtain the initial prediction result of the sample; and training the Bayesian classifier based on the initial prediction result of each sample, the sample user similarity, and the annotation result.

[0188] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program performs the following steps: acquiring a social network to be processed, the social network including users and social relationships between users; calculating the similarity between different users in the social network to be processed; determining an initial prediction result based on the similarity to determine whether there will be a social relationship between users in the future; calculating the probability that there will be a social relationship between users in the future and the probability that there will not be a social relationship between users in the future based on the initial prediction result and the similarity.

[0189] In one embodiment, the calculation of similarity between different users in a social network to be processed, when executed by a processor, includes: obtaining associated users of each user in the social network to be processed; and calculating the similarity between different users in the social network to be processed based on the associated users.

[0190] In one embodiment, the computer program is implemented when executed by a processor.

[0191] Calculating the similarity between different users in the social network to be processed based on associated users includes: calculating the similarity between different users in the social network to be processed based on associated users and at least one similarity calculation rule.

[0192] In one embodiment, after the processor executes a computer program to determine whether a social relationship exists between users in the future based on the probability that a social relationship will exist between users in the future and the probability that a social relationship will not exist in the future, the process includes: evolving the social network to be processed based on whether a social relationship exists between users in the future to obtain a target social network; determining the target user corresponding to each user based on the social network to be processed and the target social network; and pushing the corresponding target user to each user.

[0193] In one embodiment, the process of determining the target user corresponding to each user based on the social network to be processed and the target social network when the processor executes the computer program includes: calculating a first social breadth for each user in the social network to be processed; calculating a second social breadth for each user in the target social network; classifying each user according to the size of the social breadth based on the first social breadth and the second social breadth; and determining the target user corresponding to each user from the classification.

[0194] In one embodiment, the initial prediction result of determining whether there is a future social relationship between users based on similarity when the computer program is executed by the processor includes: calculating the initial prediction result of whether there is a future social relationship between users based on similarity using at least one pre-trained base classifier.

[0195] In one embodiment, the computer program, when executed by a processor, calculates the probability of future social relationships between users and the probability of future non-social relationships based on initial prediction results and similarity, respectively, including: using a pre-trained Bayesian classifier, and calculating the probability of future social relationships between users and the probability of future non-social relationships based on initial prediction results and similarity, respectively.

[0196] In one embodiment, the training method of the base classifiers involved when the computer program is executed by the processor includes: acquiring a sample network carrying annotation results; calculating the sample user similarity between sample users based on the sample network; and training based on the sample user similarity and annotation results to obtain each base classifier.

[0197] In one embodiment, the training method of the meta-classifier involved when the computer program is executed by the processor includes: inputting the sample user similarity into the trained base classifier to obtain the initial prediction result of the sample; and training based on the initial prediction result of each sample, the sample user similarity, and the labeling result to obtain a Bayesian classifier.

[0198] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps: acquiring a social network to be processed, the social network including users and social relationships between users; calculating the similarity between different users in the social network to be processed; determining an initial prediction result based on the similarity to determine whether there will be a social relationship between users in the future; calculating the probability that there will be a social relationship between users in the future and the probability that there will not be a social relationship between users in the future based on the initial prediction result and the similarity.

[0199] In one embodiment, the calculation of similarity between different users in a social network to be processed, when executed by a processor, includes: obtaining associated users of each user in the social network to be processed; and calculating the similarity between different users in the social network to be processed based on the associated users.

[0200] In one embodiment, the computer program is implemented when executed by a processor.

[0201] Calculating the similarity between different users in the social network to be processed based on associated users includes: calculating the similarity between different users in the social network to be processed based on associated users and at least one similarity calculation rule.

[0202] In one embodiment, after the processor executes a computer program to determine whether a social relationship exists between users in the future based on the probability that a social relationship will exist between users in the future and the probability that a social relationship will not exist in the future, the process includes: evolving the social network to be processed based on whether a social relationship exists between users in the future to obtain a target social network; determining the target user corresponding to each user based on the social network to be processed and the target social network; and pushing the corresponding target user to each user.

[0203] In one embodiment, the process of determining the target user corresponding to each user based on the social network to be processed and the target social network when the processor executes the computer program includes: calculating a first social breadth for each user in the social network to be processed; calculating a second social breadth for each user in the target social network; classifying each user according to the size of the social breadth based on the first social breadth and the second social breadth; and determining the target user corresponding to each user from the classification.

[0204] In one embodiment, the initial prediction result of determining whether there is a future social relationship between users based on similarity when the computer program is executed by the processor includes: calculating the initial prediction result of whether there is a future social relationship between users based on similarity using at least one pre-trained base classifier.

[0205] In one embodiment, the computer program, when executed by a processor, calculates the probability of future social relationships between users and the probability of future non-social relationships based on initial prediction results and similarity, respectively, including: using a pre-trained Bayesian classifier, and calculating the probability of future social relationships between users and the probability of future non-social relationships based on initial prediction results and similarity, respectively.

[0206] In one embodiment, the training method of the base classifiers involved when the computer program is executed by the processor includes: acquiring a sample network carrying annotation results; calculating the sample user similarity between sample users based on the sample network; and training based on the sample user similarity and annotation results to obtain each base classifier.

[0207] In one embodiment, the training method of the meta-classifier involved when the computer program is executed by the processor includes: inputting the sample user similarity into the trained base classifier to obtain the initial prediction result of the sample; and training based on the initial prediction result of each sample, the sample user similarity, and the labeling result to obtain a Bayesian classifier.

[0208] It should be noted that the method and apparatus for determining the target user of this disclosure can be used in the financial field, or in any field other than the financial field. The application field of the method and apparatus for determining the target user of this disclosure is not limited.

[0209] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0210] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0211] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0212] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting social relationships, characterized in that, The social relationship prediction method includes: Obtain the social network to be processed, which includes users and the social relationships between users; Calculate the similarity between different users in the social network to be processed; Using at least one pre-trained base classifier, an initial prediction result is determined based on the similarity to determine whether there is a future social relationship between the users; The probability of future social relationships between users and the probability of future non-social relationships are calculated using a pre-trained Bayesian classifier, based on the initial prediction results and the similarity. Based on the probability that social relationships will exist between users in the future and the probability that social relationships will not exist between users in the future, determine whether social relationships will exist between users in the future; The training methods for the base classifier include: Obtain a sample network, wherein the sample network carries the annotation results; The similarity between sample users is calculated based on the sample network. Each base classifier is obtained by training based on the sample user similarity and the annotation results; The training methods for the Bayesian classifier include: The sample user similarity is input into the trained base classifier to obtain the initial prediction result of the sample; A Bayesian classifier is obtained by training based on the initial prediction results of each sample, the sample user similarity, and the annotation results.

2. The social relationship prediction method according to claim 1, characterized in that, The calculation of similarity between different users in the social network to be processed includes: Obtain the associated users of each user in the social network to be processed; The similarity between different users in the social network to be processed is calculated based on the associated users.

3. The social relationship prediction method according to claim 2, characterized in that, The step of calculating the similarity between different users in the social network to be processed based on the associated users includes: The similarity between different users in the social network to be processed is calculated based on the associated users and at least one similarity calculation rule.

4. The social relationship prediction method according to claim 1, characterized in that, After determining whether a future social relationship exists between users based on the probability that such a relationship will exist and the probability that such a relationship will not exist, the process includes: The social network to be processed is evolved based on whether there are future social relationships between users to obtain the target social network; Based on the social network to be processed and the target social network, determine the target user corresponding to each user; The corresponding target user is pushed to each user.

5. The social relationship prediction method according to claim 4, characterized in that, The step of determining the target user corresponding to each user based on the social network to be processed and the target social network includes: Calculate the first social breadth of each user in the social network to be processed; Calculate the second social breadth of each user in the target social network; Users are categorized according to the size of their social breadth based on the first social breadth and the second social breadth. The target user corresponding to each user is determined from the classification.

6. A social relationship prediction device, characterized in that, The social relationship prediction device includes: The first network acquisition module is used to acquire the social network to be processed, which includes users and the social relationships between users. The first similarity calculation module is used to calculate the similarity between different users in the social network to be processed; The first prediction module is used to determine an initial prediction result of whether there is a future social relationship between the users based on the similarity using at least one pre-trained base classifier. The second prediction module is used to calculate the probability that users will have a social relationship in the future and the probability that they will not have a social relationship in the future based on the initial prediction results and the similarity, using a pre-trained Bayesian classifier. The target user setting module is used to determine whether there will be a social relationship between users in the future based on the probability that there will be a social relationship between them in the future and the probability that there will be no social relationship between them in the future. A sample network acquisition module is used to acquire a sample network, wherein the sample network carries annotation results; The sample similarity index calculation module is used to calculate the sample user similarity between sample users based on the sample network. The first training module is used to train each base classifier based on the sample user similarity and the annotation results. The third prediction module is used to input the sample user similarity into the trained base classifier to obtain the initial prediction result of the sample; The second training module is used to train a Bayesian classifier based on the initial prediction results of each sample, the sample user similarity, and the annotation results.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images