CN109190033B - User friend recommendation method and system - Google Patents
User friend recommendation method and system Download PDFInfo
- Publication number
- CN109190033B CN109190033B CN201810967953.3A CN201810967953A CN109190033B CN 109190033 B CN109190033 B CN 109190033B CN 201810967953 A CN201810967953 A CN 201810967953A CN 109190033 B CN109190033 B CN 109190033B
- Authority
- CN
- China
- Prior art keywords
- user
- bridge
- recommendation
- relationship
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
A method and a system for recommending friends of users are provided, wherein the method comprises the steps of constructing a secondary network relationship between a target user and a recommendation candidate set user through user behaviors, integrating the relationship between the target user and the recommendation candidate set user, and recommending a user set which meets the interest preference of the target user according to the user needs. And a behavior time factor is introduced to dynamically adjust the recommendation strategy, and a friend recommendation method based on user interest is established by taking the intermediate node as a recommended bridge based on the second-degree relation of the user. The relationship intimacy among the users has a time dimension, so that the evaluation data is more real, the intimacy among the users can be reflected more objectively and truly, the accuracy of the recommendation result is improved, the users who are most likely to be interested in the recommendation result are recommended for the users, and finally, the accuracy and the effectiveness of the recommendation result are proved through experimental verification.
Description
Technical Field
The present disclosure relates to the field of data processing technologies, and in particular, to a method and a system for recommending friends of a user.
Background
Microblog is taken as a social network platform with the largest user base number and the fastest growth at present, and microblog recommendation becomes one of core functions for improving user experience and meeting user requirements. The method for recommending the microblog materials by adding the friends as important activities in the microblog platform is characterized in that how to recommend the materials meeting the interest requirements of the user by applying the microblog recommending function is carried out in a most appropriate strategy mode and selecting the most appropriate time for recommendation, so that the target user is helped to reduce the experience cost, and the most interesting information can be found efficiently and quickly. By investigating and summarizing recommendation functions of a microblog platform, as shown in fig. 1, the microblog platform has the following 5 recommendation modes: (1) attribute information-based recommendations, (2) influence-based recommendations, (3) social relationship-based recommendations, (4) content information-based recommendations, (5) group-based recommendations, wherein:
(1) recommending based on the attribute information, recommending the recommendation candidates which have the same attribute and are located in the same community network to the target user according to the registration information of the user on the microblog, such as companies, schools and the like. The microblog platform is usually shown in the form of "XX university schoolmates" or "you are from XX company", etc.
(2) Based on the recommendation of influence, the microblog users sometimes have unclear expression of interest and demand, and based on the curious psychology of each user on the hot messages, the microblog users recommend candidates which have high influence and are hot in a period of time to target users. The recommendation presentation reason is in the form of "24-hour hot search", "wind and cloud character", and the like.
(3) And recommending the candidate resource which is most closely related to the user and has the highest interaction frequency to the target user according to the complex network constructed by the behaviors generated by the microblog users based on the recommendation of the social relationship. The recommendation and display reasons of the method are in the forms of friends, attention of concerns and the like. The idea expressed by the former is that the user may also be interested in information that is interested by friends with close relationships, and the strategy conveyed by the "attention focused" is that the user also has a certain interest in the content of the circle of interest to which the key node that the user is interested belongs. Both of these approaches are also very representative of the type of user relationship in the social network.
(4) Based on recommendation of content information, the blog content of a user is analyzed through technical means of word frequency statistics and semantic analysis, keyword extraction is carried out, a vector representation model representing the interest of the user is constructed by combining interest labels owned by the user, and candidates thrown into the interest of a target user are recommended to the target user in an interest-XX category mode by applying a label mapping classification system.
(5) Based on the grouping recommendation, according to the friend grouping information of the user, the friend information which is publicly recommended by a certain celebrity is pertinently recommended in the field group to which the celebrity belongs. The idea to be expressed is that information that the user thinks is valuable and worth being shared to the user who is interested in the information is also interested in. The search and the browse are carried out in the group, which has more pertinence and successfully improves the user experience.
Although the five recommended methods have various characteristics, the five recommended methods have certain defects. Although the former two methods can well solve the problem of cold start in the recommendation system, the user classification is too coarse-grained and lacks obvious pertinence; the recommendation based on the interest only extracts key words from the blog content and the tag information of the user to perform interest tag matching, does not explore the relationship among the hidden users in the dynamic social network, and has a single effect.
Disclosure of Invention
The invention provides a method and a system for recommending friends of a user, aiming at solving the technical problems, the invention adopts the technical scheme that the conception is as follows:
a method for recommending friends of a user, the method comprising: establishing a secondary network relationship between a target user and a recommended candidate set user through user behaviors, wherein the secondary network relationship between the target user and the recommended candidate set user comprises a primary network relationship between the target user and a bridge user and a primary network relationship between the bridge user and the recommended candidate set user;
obtaining an affinity relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user, and obtaining a similarity relationship value between the bridge user and the recommended candidate set user according to the primary network relationship between the bridge user and the recommended candidate set user;
obtaining a recommendation value of the recommendation candidate set user according to the intimacy degree relationship value between the target user and the bridge user and the similarity degree relationship value between the bridge user and the recommendation candidate set user;
and selecting a user set which accords with the interest preference of the target user according to the recommendation value of the recommendation candidate set user.
The obtaining of the affinity relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user comprises:
obtaining an intimacy degree relation value between the target user and the bridge user according to the intimacy degree between the target user and the bridge user;
the intimacy degree of the target user and the bridge user comprises the interaction strength of the target user and the bridge user and the behavior time relation of the target user and the bridge user.
The interaction strength of the target user and the bridge user comprises the interaction quantity and the interaction frequency between the target user and the bridge user within a set time period.
And the behavior time relationship between the target user and the bridge user is determined by a function containing a behavior time factor.
And the similarity relation value between the bridge user and the recommendation candidate set user is obtained by taking the number of common friends of the bridge user and the recommendation candidate set user as a measurement index.
The system comprises:
the data acquisition unit is used for constructing a secondary network relationship between a target user and a recommended candidate set user through user behaviors, wherein the secondary network relationship between the target user and the recommended candidate set user comprises a primary network relationship between the target user and a bridge user and a primary network relationship between the bridge user and the recommended candidate set user;
the data analysis unit is used for obtaining the recommendation value of each recommendation candidate set user according to the secondary network relation constructed by the data acquisition unit; the method specifically comprises the following steps:
the first data analysis module is used for obtaining an intimacy relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user;
the second data analysis module is used for obtaining a similarity relation value between the bridge user and the recommended candidate set user according to the primary network relation between the bridge user and the recommended candidate set user; the third data analysis module is used for obtaining a recommendation value of the recommendation candidate set user according to the intimacy degree relationship value obtained by the first data analysis module and the similarity degree relationship value obtained by the second data analysis module;
and the data feedback unit is used for selecting a user set which accords with the interest preference of the target user according to the recommendation value of the recommendation candidate set user obtained by the third data analysis module.
The first data analysis module is specifically configured to obtain an intimacy degree relationship value between the target user and the bridge user according to the intimacy degree between the target user and the bridge user;
the intimacy degree of the target user and the bridge user comprises the interaction strength of the target user and the bridge user and the behavior time relation of the target user and the bridge user.
The interaction strength between the target user and the bridge user in the first data analysis module comprises the interaction quantity and the interaction frequency between the target user and the bridge user within a set time period.
The behavior time relationship between the target user and the bridge user in the first data analysis module is determined by a function containing a behavior time factor.
And the similarity relation value between the bridge user and the recommended candidate set user in the second data analysis module is obtained by taking the number of common friends of the bridge user and the recommended candidate set user as a measurement index.
After adopting the technical scheme, compared with the prior art, the invention has the following beneficial effects:
firstly, determining a user set with similarity and intimacy reaching a certain quantitative value with a target node user from recommended candidate set users through a bridge user from user behaviors by the recommendation method; secondly, a user behavior time factor is added into the evaluation data as an evaluation factor for evaluating the intimacy degree relationship between the target node user and the bridge user, and an intimacy degree formula between the target node user and the bridge user is comprehensively obtained by combining the characteristics of the change rule of behaviors among users along with the time factor and the interaction strength, so that the intimacy degree relationship among the users has a time dimension, the evaluation data is more real, the intimacy relationship among the users can be objectively and truly reflected, the accuracy of a recommendation result is increased, the users who are most likely to be interested in the users are recommended for the users, and finally, the accuracy and the effectiveness of the recommendation result are proved by experimental verification.
Drawings
Wherein, the symbols in the figure are explained as follows:
FIG. 1 is a summary diagram of the present invention's recommendation method for the prior art;
fig. 2 is a network diagram of a user recommendation secondary relationship in embodiment 1 of the present invention;
fig. 3 is an Ebings forgetting graph according to embodiment 1 of the present invention;
fig. 4 is a flow chart of friend recommendation based on a user interest model according to embodiment 1 of the present invention;
fig. 5 is a statistical chart of the number of times of forwarding, commenting and mentioning of 80 verification users and users concerned therewith according to embodiment 1 of the present invention;
FIG. 6 is a graph comparing the performance of the recommendation method of example 1 of the present invention;
FIG. 7 is a flowchart of a method according to embodiment 1 of the present invention;
fig. 8 is a schematic structural diagram of a recommendation system in embodiment 3 of the present invention.
Detailed Description
Aiming at the defects of the existing microblog recommendation strategy and combining the characteristics of few words and non-standard microblog texts, the interest characteristics of users are mined from a social relationship network, a time factor is introduced to dynamically adjust the recommendation strategy, and a microblog friend recommendation method and a microblog friend recommendation system based on the user interest are established by taking an intermediate node as a recommended bridge based on the second-degree relationship of the users.
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The first embodiment is as follows:
first, the invention explains some special words such as bridge intimacy, user similarity, user recommendation value and the like one by one, and then the embodiment is explained:
first, bridge affinity is an important index for measuring the interaction and relationship strength between a bridge and a target user. If the target user is more closely related to the friends in the interest list of the target user, more interaction between the target user and the friends is generated.
According to the characteristics that a strong two-way relation represents a user social attribute and a weak one-way relation represents a user interest attribute in a microblog network, three actions of forwarding, commenting and mentioning between a target node and a bridge user are selected as behavior characteristics for analyzing the interaction strength between the users, and the interaction strength between a user i and a bridge user j is defined by a formula (1):
wherein ni, j is the total number of interactions (forwarding, commenting, mentioning) between the target node i and the bridge user j, and ni, u represents the total number of interactions between the target user and all bridge users.
As an adjustment factor, in combination with considering the strong and weak relationship types in the social network, setting the strong relationship when the users are in two-way attention
And when the relationship type of the user is weak relationship of one-way attention, setting
The method only considers the behavior factors between the users as the interaction strength between the two users. The closer the interaction behavior of the user is to the current time, the more accurate the intensity of the interaction frequency of the two users at the moment can be judged, the interaction behavior generated by the user for a long time cannot represent the current interest of the user, but the interaction intensity between the users is continuously attenuated along with the continuous change of the time factor. The attenuation law conforms to the distribution characteristics of the Eobiss forgetting curve (shown in FIG. 3).
From this figure, the forgetting curve follows the distribution characteristics of the power function. Therefore, the time factor is introduced into friend recommendation, and a behavior time factor function between the target node and the bridge user is set as follows:
where θ is a parameter of the rate of change of the user behavior with the time factor, and θ is set to 1 in this document, i.e., it is assumed that the speed of change with the time factor is the same for all users for different behavior categories, and t is the samekRepresents the time the kth interaction behavior of the user is from the present, in days, and n represents the total number of interactions (forwarding and commenting) of user i with user j.
Combining the change rule of the behavior among the users along with the time factor and the characteristics of the interaction strength, comprehensively obtaining the intimacy formula between the target node i and the bridge user j as shown in (3):
regarding the user's similarity 702, as the name implies, the user similarity 702 is an index used to measure the degree of similarity between two user features. If the user i pays attention to the user j, the fact that the user i is interested in the interest field of the user j is described. And the user y which has high similarity with the bridge user j and is interested in j is also full of interest, so that the desire to pay attention is generated. The number of fans of a user is the most intuitive index for measuring the similarity degree of two users. Therefore, the similarity between the bridge user j and the recommendation candidate y is defined as shown in formula (4):
wherein, common _ cntj, y is the number of common fans between the user j and the user y, and is used for measuring the similarity degree between the two users.
Regarding the user recommendation value, the calculation methods of user affinity and similarity between different levels of the microblog user are defined above. How to recommend the candidate resources having the second degree relationship with the user to the target user and show the candidate resources through the bridge user with the highest intimacy degree, so that the interest-adding desire of the user is attracted.
As shown in FIG. 2, the user node U0 interacts with the interested user set through the actions of paying attention, forwarding or commenting
And (4) constructing a social relationship by U-1, U2 and … Ui to acquire corresponding information and meet emotional requirements. Such users having a direct relationship with the target node are referred to as bridge users. And constructing a secondary relationship network of the target user based on the characteristic that the target user pays a certain interest to the user concerned by the interested friend, so as to carry out targeted recommendation on the recommended candidate resource for the target user through the bridge user, and mainly comprises the following technical processes:
setting w bridge users (w1, w2, …, wj) in total between the target node i and the recommended candidate y, and defining the recommendation values of the candidate user y recommended to the target user i by combining the calculation methods of the intimacy and the similarity between the users:
and obtaining a recommendation value of the candidate resource by combining the factors, wherein Xi, j is the intimacy score of the target node and the bridge user, and Simj, y is the similarity score between the bridge user and the final recommendation candidate.
A method for recommending friends of microblog users comprises the steps of constructing a secondary network relation between a target user and a recommendation candidate set user through microblog behaviors, synthesizing the relation between users, and recommending to a user set which meets the interest preference of the target user, for example, selecting a TOP-N user set according to the recommendation value.
The affinity relationship between the target node user and the bridge user comprises:
the interaction and relationship strength of the target node and the bridge node comprise the interaction strength of the target node and the bridge node and the behavior time relationship of the target node and the bridge node.
The interaction strength of the target node and the bridge node comprises microblog behavior and action performance factors between the target node and the bridge user; the microblog behavior action performance factor comprises the interaction quantity between the target node and the bridge user and the interaction frequency between the target node and the bridge user.
The similarity relation between the bridge user and the recommended candidate set user comprises the following steps: and the similarity relation between the bridge user and the recommended candidate set user takes the common fan count of the bridge user and the recommended candidate set user as a measurement index.
and recommending a user set (such as a TOP-N user set according to a recommendation value) meeting the interest preference of the target user to the target user according to the recommendation value of the recommendation candidate set user calculated according to the affinity score obtained according to the affinity relationship between the target node user and the bridge user and the similarity score obtained according to the similarity relationship between the bridge user and the recommendation candidate set user, and finishing the recommendation of the user friends.
Example two:
firstly, crawler software is used for acquiring real user information from a microblog platform. Processing and analyzing the original data set to find that the user only generates interactive behaviors with a few interested users in the attention list of the user, counting the interactive times between the target node and the bridge user, and calculating the affinity score of the target node and the bridge user. And filtering the users with the fans less than 50 in the recommended candidate resources according to the characteristic that the users with the fans less than 50 in the second chapter are not easy to pay attention to, and forming a recommended candidate resource set by taking the user with the highest bridge affinity score as a recommendation display reason.
Secondly, on the basis of analyzing the evaluation indexes of the existing personalized recommendation method, the evaluation indexes used for verifying the method are selected, namely the recall ratio, the precision ratio and the F1 mean value, and the expression modes of the four indexes are specifically given.
Finally, design experiments prove the accuracy and effectiveness of the method. Performing TOP-N display on users according to the recommendation value scores of the users in the recommendation candidates, submitting the recommendation results to the users for manual evaluation, judging whether the recommendation results generate attention behaviors to the recommended users, meanwhile, comparing and analyzing the experimental results with the online strategies of the current microblog platform, and evaluating the experimental results through related indexes; in addition, in order to verify the acceptance degree of the recommendation result of the user when the user interest changes, different recommendation list information is generated by setting four groups of different time factors, and the recommendation result is compared and analyzed.
In order to evaluate the accuracy and recommendation effect of the personalized friend recommendation algorithm based on the interest of the microblog user, and verify the acceptance degree of a recommendation result when the interest of the user changes; by collecting a real user data set of a microblog platform, the following two experiments are designed:
experiment one: and (5) checking the accuracy of the recommendation result. And verifying whether the candidate resource information to be recommended obtained by the method can meet the interest of the user. The experiment sets the time factor T to be 30, namely interaction (forwarding, commenting and mentioning) information of a target user and users in an attention list within one month is collected, and an adjusting factor is determined by combining relationship types among the users
And obtaining the intimacy between the target node and the bridge user. And further, the bridge users with high affinity scores are used as the recommendation display reason to calculate the candidate resource set to be recommended. And manually dividing the user into two categories of 'interested' and 'uninteresting', outputting the user in the recommended candidates in a TOP-N mode, and evaluating and verifying the experimental result through the model evaluation index.
In order to verify the accuracy of the results, the model needs to be evaluated by adopting reasonable evaluation indexes. In the application scenes of many current personalized recommendation systems, the focus of attention is no longer how accurate evaluation scores are given by the systems, but whether the recommended articles are interested by users. The classification accuracy is just a measure of the ability of the recommendation system to help the user find the candidate resources that the user really likes. The evaluation and verification of the experimental result are carried out by adopting common indexes of classification accuracy, such as accuracy (Precision), recall (call), F1 mean and Average accuracy (Average Precision).
The formula of accuracy is shown in fig. 6, and is used to measure the recommendation result obtained by the method, which is the probability of interest to the user:
the numerator represents that the friend recommended by the recommendation system is indeed the user marked by the user and interested in the friend, and the denominator represents the number of the friend actually recommended by the system;
and the recall rate expresses the probability that the friends in which the user actually interests are pushed out. The method is used for measuring whether the recommendation result is comprehensive enough and can cover all friends marked as 'interesting' of the user, and the formula is shown as (7):
the numerator represents the number of friends interested in the user in the recommendation list, and the denominator is the number of friends actually interested in the user.
The harmonized mean value F1 is a comprehensive effectiveness index for examining recall ratio and precision ratio as in equation (8):
the required experimental data are all from an actual microblog platform, and crawler software is used for acquiring the data from an API opened by the microblog platform. By using a breadth-first search mode, firstly, starting from a certain target user node U0, collecting the information of the list of interest of the user, namely, all the user sets of interest of the user U0, and respectively counting the total number of interactions (forwarding, commenting and mentioning) between the U0 and each user in the list of interest. Then, the users (U1, U2, …, Ui …) who are concerned by the user U0, that is, bridge users who are the reason for recommendation, count the number of common fans of the users who have a two-degree relationship with their fans, that is, with the user U0, respectively, and calculate the final recommendation value score of the user to be recommended.
The second-degree relation data information of the target node user can be obtained through the data collection mode. The target user set is 80 users who perform user interest analysis. Firstly, obtaining the attention list information of a target user in a mode of manual acquisition (clicking browsing records) and software processing, and using the attention list information as bridge data in a second-degree relationship. And then, obtaining the interactive behaviors generated among the users and the original data of the concern relationship by using a crawler tool, and finally deriving the forwarding and comment of the target user, the total times of referring to each bridge user concerned by the target user and the number of the common fans of the bridge user and the candidate to be recommended finally. And finally, respectively calculating the intimacy and the fan similarity among the users at different levels by using formulas (1-4), and obtaining the final recommendation value score of the candidate resource by using a formula (5).
The method collects the information of the interaction behavior of the experimental user and the concerned friends within one month. Summarizing the number of interactions (forwarding, commenting, mentioning) generated by these users (as shown in fig. 5), it was found that about 59.4% of the interactions between users did not occur. Most bridge users are forwarded, commented and mentioned by experimental users for a few times, and the number of users with the interaction times less than or equal to 5 accounts for about 74.3 percent of the total number. Only about 7.9% of the user interactions are greater than 20, while the total number of interactions generated by this portion of the user population is as high as 2546, which is about 78.6% of the total. Therefore, when the bridge information is selected, the bridge users with high interaction times are mainly used as the recommendation display reason, and meanwhile, the number of users without interactive behaviors in the attention list of the experimental user is filtered.
The data show that the user only interacts with a few interested friends in the attention list of the user, and the user is taken as the recommendation display information of the candidate user, so that the attention efficiency of the user is improved. According to the analysis of the user attention demand in the second chapter, when the number of fans of the user is less than 50, the probability of attention paid by other users is very low. Therefore, when the similarity of the fans of the bridge user and the users in the attention list of the bridge user is calculated, the users with the fan number smaller than 50 in the recommendation candidates are filtered. For the convenience of later analysis and verification, when the final recommendation value of the candidate resource user is calculated, the candidate resource set of the target user is formed by the users to be recommended with the top 20 ranking score.
For convenience in description, and for comparing the candidate resources finally recommended to the target user with the real interests of the user, the experimental user B performing user interest analysis in chapter iii is taken as an example for explanation, and personalized friend recommendation methods of other users are similar to the experimental user B. The attention number of the user B is 108, the user B interacts with 44 users in the attention list of the user B under the condition that the time factor T is 30, and according to the personalized friend recommendation method shown in the text, the steps of selecting the bridge user of the experimental user B and calculating the recommendation value score of the end user are as follows:
firstly, acquiring bridge user behavior data directly associated with an experimental user to form a candidate set of bridge users; the method comprises the steps that an experimental user B is taken as a node, and the total times of interaction (forwarding, commenting and mentioning) between the experimental user B and concerned friends are counted; filtering users with the interaction times of 0 under the condition that the time factor T is 30 to form a candidate set of bridge users;
secondly, calculating the intimacy values of the directly related bridge users and the experimental users and sequencing the intimacy values, wherein the intimacy values comprise the steps of determining an adjusting factor according to the relationship type between a definite target node and the bridge users
According to the current time of the time distance generated by each interactive behavior, calculating the intimacy score of the target node and the bridge user according to a formula (3), and carrying out sequencing output according to the score;
then, acquiring behavior data of recommended candidate users directly associated with the bridge users, and obtaining similarity values of the recommended candidate users and the bridge users according to the similarity between the recommended candidate users and the bridge users, wherein the similarity values comprise that the bridge users are used as nodes, an attention list of each bridge user is filtered, namely the number of fans in the recommended candidates is less than 50, and a similarity score between the recommended candidates and the bridge users is calculated according to a formula (4);
and finally, calculating the recommendation value of the recommendation candidate user according to the score result of the step, and outputting a recommendation result according to the user requirement. The method comprises the steps of counting and combining the number of bridges of each candidate resource according to the steps, calculating the final recommendation value score of the candidate resource of the experimental user B according to a formula (5), and outputting a recommendation result according to the user requirement and TOP-N;
summarizing the interest model of the experimental user B, the information and the score values of the bridge user and the recommended user, as shown in Table 1:
TABLE 1 recommendation candidates and weights thereof for Experimental user B
Summarizing the field categories and the recommendation values of the recommendation candidates of the experimental user B and the user interest information in the third chapter, as shown in Table 2:
TABLE 2 comparison of user interest to value scores of recommendation candidates
| Food | Fashion style | Joke's line | Film | Cartoon | Travel device | Economy of production | |
| Interest weight | 0.329 | 0.228 | 0.181 | 0.087 | 0.074 | 0.067 | 0.034 |
| Value of recommendation | 0.282 | 0.243 | 0.163 | 0.122 | 0.085 | 0.054 | 0.051 |
Performing significance t test on the data in table 2 to obtain t ═ 0, which indicates that the experimental results obtained twice have no significance difference; correlation analysis is carried out on the data, and the Pearson correlation coefficient r is 0.974, which indicates that the two groups of data have high correlation.
Through the analysis, the experimental user B is interested in information in the aspects of gourmet, fashion, joke, movies and the like, and the selected bridge, such as a fashion street photo, a carefully selected joke and a comic home, is the user with the largest number of times of mention, forwarding and comment of the user B, and the selected bridge is taken as a recommendation display reason, so that the diversity of recommendation results is met, and the successful establishment of the relationship is promoted to a great extent. Moreover, the fields to which the users in the candidate resources belong basically cover the interest field categories of the user B, and the correlation between the user B and the user B is found to be high through the verification of the correlation indexes. Therefore, the personalized friend recommendation method based on the user interests has good applicability.
The following tests and analyses were performed on the accuracy of the experimental recommendations:
although there is a great correlation between the interests of the candidate users recommended by the method herein and the experimental user B, the behavior of the target user on the recommendation result is not known. Thus, the indicators presented herein above serve as evaluation criteria to verify whether the target user will produce attentive behavior when the target user is given a recommendation in the form of TOP-N.
The 20 recommended candidates of the experimental user B are handed to the user for manual labeling, and are divided into two types of interest and uninteresting. According to the ranking of the scores of the recommendation values, recommending @3, @5, @7 and @15 users to the target user each time, and counting the results of each evaluation index as shown in table 3:
TABLE 3 statistical table of test results for Experimental user B
| Evaluation index | P@3 | P@5 | P@7 | P@15 |
| Recall ratio of | 0.143 | 0.286 | 0.357 | 0.429 |
| Precision ratio | 0.667 | 0.800 | 0.714 | 0.400 |
| F1 | 0.235 | 0.421 | 0.476 | 0.414 |
All 80 experimental users are treated in the same way, and finally the average values of all the inspection indexes are summarized as shown in the following table:
TABLE 4 statistical table of the overall test results of the experimental users
| Evaluation index | P@3 | P@5 | P@7 | P@15 |
| Recall ratio of | 0.169 | 0.273 | 0.336 | 0.462 |
| Precision ratio | 0.703 | 0.756 | 0.683 | 0.429 |
| F1 | 0.272 | 0.401 | 0.450 | 0.445 |
From the above results, it can be seen that as the number N of recommended friends increases, the precision ratio decreases, while the recall ratio increases, and the application of the F1 average index can be balanced between the two. When N is 7, the index has the largest value. Namely, the effect achieved is the best when 7 friends are recommended to the user each time.
For comparative analysis, and to verify the recommended effectiveness of the methods herein. The existing 3 recommendation methods of the current microblog platform are used as a control group for carrying out experiments, and the data selection process of the control group is as follows:
(1) selecting popular users with the top ranking of 20 from the 24-hour hot search list to form recommendation candidates;
(2) randomly selecting 20 users from information (the same company and the same school) with the same attribute as the target user to form recommendation candidates;
(3) and selecting the top 20 users with the most common friends from the second-degree relationship data of the users to form recommendation candidates by taking the friends of the friends as entry points, namely by a strong relationship type.
The selected data sets of the three recommendation strategies are manually labeled according to the same method, are divided into two types of 'interested' and 'uninteresting', and are output and displayed to the user in a TOP N mode, and the performance comparison conditions of the four different recommendation strategies are shown in FIG. 6.
Experiment two:
the verification analysis above illustrates that the recommendation results derived by the methods herein are largely relevant to the user's interests. And comparative analysis is carried out, and the effect of the method on each evaluation performance index is superior to that of other classical recommendation methods. However, over time, the user interests change. Thus, it is considered whether the user's acceptance of friends in their recommendation list changes under the influence of different time factors.
And adjusting the time window factors to be T1-5, T2-10, T3-15 and T4-30, namely counting the number of interactions between the target user and the bridge user under different time factor windows respectively, calculating the affinity score between the target user and the bridge user, and recommending the similar user of the bridge user with the highest affinity score to the target user. The candidate resource of TOP7 at different time factors is recommended to the experimental user, and table 5 shows the acceptance degree of the experimental user B to the user information in the recommendation list.
TABLE 5 user B acceptance of users in the recommendation list under different time factors
| T1＝5 | T2＝10 | T3＝15 | T4＝30 |
| Food critic (V) | Food critic (V) | Fashion interesting house (check mark) | Food critic (V) |
| Cartoon shop (check square) | Chanting bear (hook) | Eating in Beijing (√) | Cartoon shop |
| Hundred thousand cold jokes (V) | Eating in Beijing (√) | Panyuexin silk ball (V) | Finance news (√ V) |
| Panyuexin silk ball (V) | Fun and fun (V shape) | Chanting bear (hook) | Economic observation newspaper (check square) |
| Chanting bear (hook) | Movie new information (√ V) | Jim (stylist) | Jim (stylist) |
| Fun and fun (V shape) | Finance and economics news | Cartoon shop (check square) | Explosion of humorous body |
| Shanghai Jiang Japanese language (check mark) | Small squad of food (V) | Economic observation newspaper | Thank you |
Note 1): (√ v) indicates that user B is interested in the recommended user information
It can be seen from the table that no matter how the time factor changes, when the microblog user related to the food is recommended to the user, the user expresses a desire to pay attention, that is, the food is a long-term interest of the user. When the time factor T4 is 30, recommending microblog users related to financial information to the users, wherein the users are interested in the microblog users, and then generating attention-adding behaviors; however, when T2 is 10 and T3 is 15, the user does not select the microblog user related to the finance class in the focused recommendation list. I.e., over time, the user's interests may change.
In order to test the recommended effect of the experimental user B under different time factors, the data in table 5 are tested, and the same processing is performed on the other experimental users, and the specific test results are shown in table 6:
TABLE 6 statistical table of test results of experimental users
It can be seen from the table that each test index of the experimental user is continuously reduced along with the change of the time factor, that is, when the interest of the user drifts, the change of the interest of the user needs to be accurately captured, targeted recommendation is generated according to the instant interest characteristics of the user, the attention requirement of the user is met, and meanwhile, the microblog pushing efficiency is greatly improved.
Example three:
a user friend recommendation system, the system comprising: a data acquisition unit 801, a data analysis unit 802, and a data feedback unit 803.
The data obtaining unit 801 is configured to construct a secondary network relationship between the target user and the recommended candidate set user according to the user behavior, where the secondary network relationship between the target user and the recommended candidate set user includes a primary network relationship between the target user and the bridge user and a primary network relationship between the bridge user and the recommended candidate set user.
The data analysis unit 802 is configured to obtain a recommendation value of each user recommending a candidate set according to the secondary network relationship established by the data acquisition unit; the method specifically comprises the following steps: the first data analysis module is used for obtaining an intimacy relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user; the second data analysis module is used for obtaining a similarity relation value between the bridge user and the recommended candidate set user according to the primary network relation between the bridge user and the recommended candidate set user; the third data analysis module is used for obtaining a recommendation value of the recommendation candidate set user according to the intimacy degree relationship value obtained by the first data analysis module and the similarity degree relationship value obtained by the second data analysis module;
the data feedback unit 803 is configured to select a user set that meets the interest and preference of the target user according to the recommended value of the recommended candidate set user obtained by the third data analysis module, and specifically, may select a TOP-N user set according to the recommended value.
The data acquisition unit establishes a secondary network relationship between a target user and a recommended candidate set user through user behaviors, wherein the secondary network relationship between the target user and the recommended candidate set user comprises a primary network relationship between the target user and a bridge user and a primary network relationship between the bridge user and the recommended candidate set user, and specifically refers to comments, forwarding times, forwarding frequency, comment times, comment frequency and the like between the target user and the bridge user and between the bridge user and the recommended candidate set user.
The data analysis unit is used for analyzing the secondary network relation constructed by the data acquisition unit to obtain the recommendation value of each recommendation candidate user; the method comprises the following steps: the first data analysis module obtains an intimacy relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user;
the second data analysis module obtains a similarity relation value between the bridge user and the recommended candidate set user according to the primary network relation between the bridge user and the recommended candidate set user;
and the third data analysis module obtains the recommendation value of the recommendation candidate set user according to the intimacy degree relationship value between the target user and the bridge user obtained by the first data analysis module and the similarity degree relationship value between the bridge user and the recommendation candidate set user obtained by the second data analysis module.
The first data analysis module is specifically used for obtaining an intimacy degree relationship value between the target user and the bridge user according to the intimacy degree between the target user and the bridge user; the intimacy degree of the target user and the bridge user comprises the interaction strength of the target user and the bridge user and the behavior time relation of the target user and the bridge user. The interaction strength of the target user and the bridge user comprises the interaction quantity and the interaction frequency between the target user and the bridge user within a set time period.
And the behavior time relationship between the target user and the bridge user is determined by a function containing a behavior time factor.
And the similarity relation value between the bridge user and the recommendation candidate set user is obtained by taking the number of common friends of the bridge user and the recommendation candidate set user as a measurement index.
The recommendation value of the recommendation candidate set user analyzed by the data analysis unit is determined by the affinity relationship value between the target user and the bridge user and the similarity relationship value between the bridge user and the recommendation candidate set user.
The data feedback unit recommends a user set which meets the interest preference of a target user according to the recommendation value of the recommendation candidate set user analyzed by the data analysis unit and the user requirement, and specifically, a TOP-N user set can be selected according to the recommendation value to complete friend recommendation of the user.
It should be understood that the specific order or hierarchy of steps in the processes disclosed is an example of exemplary approaches. Based upon design preferences, it is understood that the specific order or hierarchy of steps in the processes may be rearranged without departing from the scope of the present disclosure. The accompanying method claims present elements of the various steps in a sample order, and are not intended to be limited to the specific order or hierarchy presented.
In the foregoing detailed description, various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments of the subject matter require more features than are expressly recited in each claim. Rather, as the following claims reflect, invention lies in less than all features of a single disclosed embodiment. Thus, the following claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. To those skilled in the art; various modifications to these embodiments will be readily apparent, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the disclosure. Thus, the present disclosure is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
What has been described above includes examples of one or more embodiments. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the aforementioned embodiments, but one of ordinary skill in the art may recognize that many further combinations and permutations of various embodiments are possible. Accordingly, the embodiments described herein are intended to embrace all such alterations, modifications and variations that fall within the scope of the appended claims. Furthermore, to the extent that the term "includes" is used in either the detailed description or the claims, such term is intended to be inclusive in a manner similar to the term "comprising" as "comprising" is interpreted when employed as a transitional word in a claim. Furthermore, any use of the term "or" in the specification of the claims is intended to mean a "non-exclusive or".
Those of skill in the art will further appreciate that the various illustrative logical blocks, units, and steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate the interchangeability of hardware and software, various illustrative components, elements, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design requirements of the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present embodiments.
The various illustrative logical blocks, or elements, described in connection with the embodiments disclosed herein may be implemented or performed with a general purpose processor, a digital signal processor, an Application Specific Integrated Circuit (ASIC), a field programmable gate array or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a digital signal processor core, or any other similar configuration.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may be stored in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. For example, a storage medium may be coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC, which may be located in a user terminal. In the alternative, the processor and the storage medium may reside in different components in a user terminal.
In one or more exemplary designs, the functions described above in connection with the embodiments of the invention may be implemented in hardware, software, firmware, or any combination of the three. If implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media that facilitate transfer of a computer program from one place to another. Storage media may be any available media that can be accessed by a general purpose or special purpose computer. For example, such computer-readable media can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to carry or store program code in the form of instructions or data structures and which can be read by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Additionally, any connection is properly termed a computer-readable medium, and, thus, is included if the software is transmitted from a website, server, or other remote source via a coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL), or wirelessly, e.g., infrared, radio, and microwave. Such discs (disk) and disks (disc) include compact disks, laser disks, optical disks, DVDs, floppy disks and blu-ray disks where disks usually reproduce data magnetically, while disks usually reproduce data optically with lasers. Combinations of the above may also be included in the computer-readable medium.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are merely exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (8)
1. A method for recommending friends of a user is characterized in that the method comprises the following steps:
establishing a secondary network relationship between a target user and a recommended candidate set user through user behaviors, wherein the secondary network relationship between the target user and the recommended candidate set user comprises a primary network relationship between the target user and a bridge user and a primary network relationship between the bridge user and the recommended candidate set user;
obtaining an affinity relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user, and obtaining a similarity relationship value between the bridge user and the recommended candidate set user according to the primary network relationship between the bridge user and the recommended candidate set user;
obtaining a recommendation value of the recommendation candidate set user according to the intimacy degree relationship value between the target user and the bridge user and the similarity degree relationship value between the bridge user and the recommendation candidate set user;
selecting a user set which accords with the interest preference of the target user according to the recommendation value of the recommendation candidate set user;
the obtaining of the affinity relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user comprises:
obtaining an intimacy degree relation value between the target user and the bridge user according to the intimacy degree between the target user and the bridge user;
the intimacy degree of the target user and the bridge user comprises the interaction strength of the target user and the bridge user and the behavior time relation of the target user and the bridge user.
2. The method of claim 1, wherein the interaction strength of the target user and the bridge user comprises the number of interactions and the frequency of interactions between the target user and the bridge user within a set time period.
3. The method of claim 1, wherein the behavioral time relationship between the target user and the bridge user is determined by a function comprising a behavioral time factor.
4. The method of claim 1, wherein the similarity relationship value between the bridge user and the recommendation candidate set user is obtained by using the number of common friends of the bridge user and the recommendation candidate set user as a measure.
5. A user friend recommendation system, the system comprising:
the data acquisition unit is used for constructing a secondary network relationship between a target user and a recommended candidate set user through user behaviors, wherein the secondary network relationship between the target user and the recommended candidate set user comprises a primary network relationship between the target user and a bridge user and a primary network relationship between the bridge user and the recommended candidate set user;
the data analysis unit is used for obtaining the recommendation value of each recommendation candidate set user according to the secondary network relation constructed by the data acquisition unit; the method specifically comprises the following steps:
the first data analysis module is used for obtaining an intimacy relationship value between the target user and the bridge user according to the primary network relationship between the target user and the bridge user;
the second data analysis module is used for obtaining a similarity relation value between the bridge user and the recommended candidate set user according to the primary network relation between the bridge user and the recommended candidate set user;
the third data analysis module is used for obtaining a recommendation value of the recommendation candidate set user according to the intimacy degree relationship value obtained by the first data analysis module and the similarity degree relationship value obtained by the second data analysis module;
the data feedback unit is used for selecting a user set which accords with the interest preference of the target user according to the recommendation value of the recommendation candidate set user obtained by the third data analysis module;
the first data analysis module is specifically configured to obtain an intimacy degree relationship value between the target user and the bridge user according to the intimacy degree between the target user and the bridge user;
the intimacy degree of the target user and the bridge user comprises the interaction strength of the target user and the bridge user and the behavior time relation of the target user and the bridge user.
6. The system of claim 5, wherein the interaction strength of the target user and the bridge user in the first data analysis module comprises the number of interactions and the interaction frequency between the target user and the bridge user within a set time period.
7. The system of claim 5, wherein the time relationship between the target user and the bridge user in the first data analysis module is determined by a function comprising a time factor of behavior.
8. The system of claim 5, wherein the similarity relationship value between the bridge user and the recommendation candidate set user in the second data analysis module is obtained by using the number of common friends between the bridge user and the recommendation candidate set user as a measure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810967953.3A CN109190033B (en) | 2018-08-23 | 2018-08-23 | User friend recommendation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810967953.3A CN109190033B (en) | 2018-08-23 | 2018-08-23 | User friend recommendation method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109190033A CN109190033A (en) | 2019-01-11 |
| CN109190033B true CN109190033B (en) | 2022-01-11 |
Family
ID=64919263
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810967953.3A Active CN109190033B (en) | 2018-08-23 | 2018-08-23 | User friend recommendation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109190033B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110674413B (en) * | 2019-09-09 | 2022-03-25 | 平安科技（深圳）有限公司 | User relationship mining method, device, equipment and storage medium |
| CN111597454A (en) * | 2020-04-02 | 2020-08-28 | 微梦创科网络科技（中国）有限公司 | Account recommendation method and device |
| CN112380423A (en) * | 2020-10-14 | 2021-02-19 | 广州市百果园网络科技有限公司 | Network object pushing method and device, readable storage medium and electronic equipment |
| CN112559893A (en) * | 2021-01-11 | 2021-03-26 | 引粒网络科技(上海)有限公司 | User recommendation method in internet matching social scene |
| CN113076450B (en) * | 2021-03-15 | 2024-03-22 | 北京明略软件系统有限公司 | Determination method and device for target recommendation list |
| KR20220169206A (en) * | 2021-06-18 | 2022-12-27 | 라인플러스 주식회사 | Method, computer device, and computer program to recommend friend within timeline service |
| CN115357632A (en) * | 2022-07-21 | 2022-11-18 | 广东技术师范大学 | Learner personalized interpretable potential friend recommendation method and device |
| CN115186664B (en) * | 2022-09-13 | 2023-01-13 | 深圳市爱聊科技有限公司 | Method and system for measuring and calculating coincidence degree between subjects based on multiple dimensions |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102637183A (en) * | 2011-02-12 | 2012-08-15 | 北京千橡网景科技发展有限公司 | Method and device for recommending friends to user in social network |
| CN103294733A (en) * | 2012-03-05 | 2013-09-11 | 北京千橡网景科技发展有限公司 | Method and system used for recommending friends to users |
| EP2701103A3 (en) * | 2012-08-24 | 2014-03-26 | Samsung Electronics Co., Ltd | Method of recommending friends, and server and terminal therefor |
| CN106326228A (en) * | 2015-06-17 | 2017-01-11 | 富士通株式会社 | Method and device for evaluating interest tendency of user |
| CN106649781A (en) * | 2016-12-28 | 2017-05-10 | 北京小米移动软件有限公司 | Application recommendation method and device |
| CN107066476A (en) * | 2016-12-13 | 2017-08-18 | 江苏途致信息科技有限公司 | A kind of real-time recommendation method based on article similarity |
-
2018
- 2018-08-23 CN CN201810967953.3A patent/CN109190033B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102637183A (en) * | 2011-02-12 | 2012-08-15 | 北京千橡网景科技发展有限公司 | Method and device for recommending friends to user in social network |
| CN103294733A (en) * | 2012-03-05 | 2013-09-11 | 北京千橡网景科技发展有限公司 | Method and system used for recommending friends to users |
| EP2701103A3 (en) * | 2012-08-24 | 2014-03-26 | Samsung Electronics Co., Ltd | Method of recommending friends, and server and terminal therefor |
| CN106326228A (en) * | 2015-06-17 | 2017-01-11 | 富士通株式会社 | Method and device for evaluating interest tendency of user |
| CN107066476A (en) * | 2016-12-13 | 2017-08-18 | 江苏途致信息科技有限公司 | A kind of real-time recommendation method based on article similarity |
| CN106649781A (en) * | 2016-12-28 | 2017-05-10 | 北京小米移动软件有限公司 | Application recommendation method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109190033A (en) | 2019-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109190033B (en) | User friend recommendation method and system | |
| US10579646B2 (en) | Systems and methods for classifying electronic documents | |
| US11122009B2 (en) | Systems and methods for identifying geographic locations of social media content collected over social networks | |
| CN103246670B (en) | Microblogging sequence, search, methods of exhibiting and system | |
| US11176142B2 (en) | Method of data query based on evaluation and device | |
| KR102473471B1 (en) | Learning and using contextual content retrieval rules for query disambiguation | |
| CN104680453B (en) | Course based on student's attribute recommends method and system | |
| US20130304818A1 (en) | Systems and methods for discovery of related terms for social media content collection over social networks | |
| CN105787025B (en) | Network platform public account classification method and device | |
| US20130297581A1 (en) | Systems and methods for customized filtering and analysis of social media content collected over social networks | |
| US20130297694A1 (en) | Systems and methods for interactive presentation and analysis of social media content collection over social networks | |
| CN111125453B (en) | Opinion leader role identification method in social network based on subgraph isomorphism and storage medium | |
| CN110334356A (en) | Article matter method for determination of amount, article screening technique and corresponding device | |
| Röchert et al. | Opinion-based homogeneity on YouTube: Combining sentiment and social network analysis | |
| Jeon et al. | Hashtag recommendation based on user tweet and hashtag classification on twitter | |
| Okazaki et al. | How to mine brand Tweets: Procedural guidelines and pretest | |
| Bornmann et al. | Alternative article‐level metrics: The use of alternative metrics in research evaluation | |
| CN113934941A (en) | User recommendation system and method based on multi-dimensional information | |
| CN109636682A (en) | A kind of teaching resource auto-collection system | |
| Li et al. | Probabilistic local expert retrieval | |
| Li et al. | Detecting Promotion Campaigns in Community Question Answering. | |
| CN111105117B (en) | User information determining method and device | |
| Heravi et al. | Tweet location detection | |
| CN108520012A (en) | Mobile Internet user comment method for digging based on machine learning | |
| JP7003481B2 (en) | Reinforcing rankings for social media accounts and content |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |