CN107220328B - Social network-based weak relation and strong relation video recommendation method - Google Patents
Social network-based weak relation and strong relation video recommendation method Download PDFInfo
- Publication number
- CN107220328B CN107220328B CN201710371550.8A CN201710371550A CN107220328B CN 107220328 B CN107220328 B CN 107220328B CN 201710371550 A CN201710371550 A CN 201710371550A CN 107220328 B CN107220328 B CN 107220328B
- Authority
- CN
- China
- Prior art keywords
- user
- relation
- video
- weak
- strong
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- 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
The invention discloses a video recommendation method based on weak relation and strong relation of social network, comprising the following steps: 1) calculating a first estimated click probability of the user on the video based on the influence of the weak-relation friend; 2) calculating a second estimated click probability based on the behavior characteristics of strong-relation friends with different intimacy degrees to the video; 3) linearly fusing the first and second estimated click probabilities to obtain an estimated click probability; 4) constructing a loss function based on the behavior observation value of the user to the video and the estimated click probability; 5) optimizing the characteristic parameters in the steps 1) -4) by adopting a random gradient algorithm to minimize a loss function, and finally obtaining the click probability of the user on the video; 6) and recommending the N videos with the highest click probability to the user. The video recommendation method combines the weak relation friends and the strong relation friends related to the user to perform video recommendation, and acts the behavior characteristics of the friends on the prediction model, so that the accuracy of video recommendation can be improved.
Description
Technical Field
The invention relates to the technical field of information, in particular to a video recommendation method.
Background
Video recommendation has been a popular research topic in academic research, and is closely related to user click rate and video popularity. With the rise of social software Facebook, Twitter and microblog, the influence of social network characteristics on the popularization of video transmission becomes a new research hotspot. In the traditional research work, the future popularity (the number of browsed videos) of the videos is predicted according to the historical browsing total number of the videos, the popularity of the videos is considered to be linearly increased along with the time, so that the future popularity of the videos can be predicted by training a linear regression model according to the increment of the popularity of the videos in different time periods compared with the popularity of the videos in the previous time period as a characteristic. In addition, the method also comprises a model of online learning based on multi-stage time sequence prediction video popularity, utilizes the characteristics of dynamic propagation of the video in a social network, and predicts the popularity of the video in each divided propagation time period.
The published videos in the social network may contain tags, whether the videos are published by large V authenticated people, text length and other features, and the potential interest value of the user in the new videos is estimated according to videos clicked by the user in history. However, the traditional model does not take into account the mutual information between users, such as mention, like, forward, comment, reply. Deciding whether a user clicks on a video in a social network depends largely on the location of the video contributors in the social network, regardless of the content of the video. Video contributors refer to user behavior on the video, such as publishing, forwarding, praise, collection, etc. And people are easier to accept the items recommended by friends. In the traditional work, an LBM (lattice boltzmann Method) model based on friendship clustering is proposed, and the model firstly adopts a KMeans algorithm to classify friends of a user into close friends, familiar friends and unfamiliar friends according to the interaction frequency between the user and two-way friends, and then adopts an LBM algorithm to train the weight of user behavior characteristics of the friends in different friendships. And finally, accumulating the behavior weights of the friends in different clusters to obtain the probability that the video is clicked by the user at last. However, in this model, only a strong relationship of bidirectional attention, which is a friendship, is considered, but a weak relationship of unidirectional attention is not considered.
The social relationships in the social network mainly include a strong relationship network and a weak relationship network. In the strong relationship network, the user may be a real friend relationship, such as family, classmates, colleagues, etc., and the strong relationship mainly represents the intimacy of the user to friends. The weak relation is a virtual relation which is not known among ordinary users, and can better reflect the preference interest of the users. Social networks such as microblog and Twitter are mainly based on the weak relationship of users. At present, the combination of two social relationships, namely a strong relationship and a weak relationship, is not considered in the work of predicting the video popularity and the user click rate based on the social network.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a video recommendation method based on the weak relation and the strong relation of the social network, and video recommendation is further carried out based on all friends by fusing the preference of the weak relation friends and the intimacy of the strong relation friends.
The technical scheme is as follows: in order to achieve the above object, the method for recommending videos based on weak relationships and strong relationships of social networks of the present invention comprises the following steps:
1) calculating a first estimated click probability of the user on the video based on the influence of the weak-relation friend of the user, the correlation degree of the user and the weak-relation friend and the behavior characteristics of the weak-relation friend on the video;
2) calculating a second estimated click probability of the user to the video based on behavior characteristics of strong-relation friends with different intimacy to the video;
3) linearly fusing the first estimated click probability and the second estimated click probability to obtain the estimated click probability of the user to the video;
4) constructing a loss function based on the behavior observation value of the user to the video and the estimated click probability of the user to the video;
5) optimizing the characteristic parameters in the steps 1) -4) by adopting a random gradient algorithm to minimize the value of the loss function, and stopping optimization when the value of the loss function meets a preset condition to finally obtain the click probability of the user on the video;
6) and recommending the N videos with the highest click probability to the user by using a TopN algorithm.
For the weak relation friend, the influence (w) of the weak relation friend is calculated by a RePageRank algorithm, and the calculation formula is
Wherein the content of the first and second substances,
d is the probability of the user walking, namely the probability of the user clicking other links, and is initialized through a Gaussian function;
b represents a fan set concerning the weak relation friends;
influence (u) is the influence of user u, which is initialized by a Gaussian function;
out (u) is the number of friends the user u has focused on;
re (u) is the forwarding rate of the user and is obtained according to the ratio of the number of times the video sent by the user u is forwarded to the total number of videos sent by the user u.
Calculating a first estimated click probability of the user on the video t in the step 1) by using the following formula according to the influence obtained above
Wherein the content of the first and second substances,
f represents the behavior characteristics of the weak relation friend to the video t;
f represents a behavior feature set of the weak-relation friend to the video t, wherein the behavior feature set comprises actions of publishing, mentioning, praise, commenting, replying, forwarding, collecting and the like;
bf,wweights for the behavior features f of the weak-relation friends w, which are initialized by a gaussian function;
It f,wis an indication function which indicates whether the weak-relation friend w has the past behavior on the video t, if so, It f,w1, otherwise It f,w＝0；
ru,wAnd obtaining the relevance between the user and the weak relation friend according to the ratio of the total number of the videos of the weak relation friend clicked by the user to the total number of the videos of the user.
For the strong-relation friends, firstly, based on interaction frequency binary groups (f1, f2) of the user and the strong-relation friends, the strong-relation friends are divided into friend clusters with different intimacy degrees by using a KMeans algorithm, and then second estimated click probability of the user on the video t is calculated through behavior characteristics f of the strong-relation friends with different intimacy degrees on the video t
Wherein K is the friend cluster of the strong relation friend, bf,sIs the weight of the behavior feature f of the strong relationship friend in the cluster K,
is an indication function which indicates whether strong-relation friends in the cluster K have over-interactive behaviors with the video t, if so, the strong-relation friends in the cluster K have over-interactive behaviors with the video t
Otherwise
To obtain
And
and then, linearly fusing the videos, and considering the click preference of the user on the videos and the average click rate of the user for clicking the videos during fusion so as to enable the finally obtained estimated click probability to be more accurate. In addition, in order to further improve the hit rate, a loss function is constructed based on the actual action observed value of the user on the videos and the estimated click probability, the loss function is corrected, the characteristic parameters are optimized by using a random Gradient SGD (storage Gradient decision) algorithm, the accurate click probability is finally obtained, and the N videos with the highest click probability are recommended to the user.
Has the advantages that: the video recommendation method combines the weak relation friends and the strong relation friends related to the user to perform video recommendation, comprehensively considers the influence of the weak relation friends, the intimacy of the strong relation friends, the weight of the interactive behavior characteristics of the strong relation friends and the weak relation friends and the cumulative sum of the weight of the behavior characteristics to obtain the click probability of the user on the video, and therefore the accuracy of the video recommendation can be improved.
Drawings
FIG. 1 is a graph of weak and strong relationships in a social network;
FIG. 2 is a general flow diagram of the method of the present invention;
FIG. 3 is a flow chart of the present invention for calculating the influence of a user's weak-relationship friends;
fig. 4 is a flowchart of clustering friends with strong relationship of the user according to the present invention.
Detailed Description
The technical scheme of the invention is further explained by combining the attached drawings.
Two types of social relationships exist for users in a social network, as shown in FIG. 1, where a unidirectional arrow represents a weak relationship of one-way interest and a bidirectional arrow represents a strong relationship of one-way interest to each other. In a social network, a weak relationship is a virtual relationship which is not known among general users, and the attention can better reflect the preference interest of the users. The strong relationship is a real relationship, that is, the users are real friendships, such as family, classmates, colleagues, etc., and the strong relationship mainly reflects the intimacy and trust degree of the users to friends.
FIG. 2 shows a flow chart of the method of the present invention, wherein, for the weak-relation friends of the social network, a new algorithm RePageRank for calculating the influence of the user is designed according to the fan number and the forwarding rate of the user; for strong-relation friends of the social network, different intimacy degrees are divided according to the interaction frequency of the user and the friends based on a KMeans algorithm; then, combining weak relation friends and strong relation friends of the user, and predicting the click probability of the video t according to behavior characteristics (publication, praise, forwarding, comment and the like) of the friends on the video t; then, according to the fact that the sum of squares of the observation value and the predicted value of the user on the video t is used as a loss function, a random gradient SGD algorithm is adopted to minimize the loss function, and characteristic parameters are optimized; and finally, recommending the N video sets with the maximum click probability to the user by adopting a TopN algorithm. In this embodiment, N is 10. The specific process is described in detail below.
First, the influence of the weak-relation friend is calculated, as shown in fig. 3, the steps required for calculating the influence of the weak-relation friend are as follows:
3A) the influence (u) of the user u is initialized by a gaussian function.
3B) And obtaining the forwarding rate Re (u) of the user according to the ratio of the number of times of forwarding the video sent by the user u to the total number of the videos sent by the user u.
3C) And mapping the user forwarding rate Re (u) to be between 0 and 1 by adopting a Min-Max normalization method.
3D) And calculating the influence of the weak relation friends by combining the forwarding rate and the number of the fans, wherein the calculation formula is as follows:
wherein d is the probability of the user walking, that is, the probability of the user clicking other links, and is set to 0.85 in this embodiment; b represents a fan set concerning the weak relation friends; out (u) is the number of buddies that user u is interested in.
3E) And repeating the steps 3B) -3D) until the influence of the weak relation friend is converged.
Secondly, for friends with strong relationship, the friends with strong relationship of the user are real "friend" relationships, such as family, classmates, colleagues, etc., but the "friend" relationships also have close relationships with different degrees, so that friends with different close densities, such as close friends, familiar friends and unfamiliar friends, can be classified according to the interaction frequency degree of the user and the friends with strong relationship. As shown in fig. 4, the steps of dividing the friend clusters with 3 different affinities are as follows:
4A) according to the proportion of the total number of the interaction behaviors (publishing, commenting and forwarding) between the user u and the strong-relation friends s to the total number of the video tweets of the user, the interaction frequency f1 of the user u and the strong-relation friends s is calculated, and similarly, the interaction frequency f2 of the strong-relation friends s and the user u is calculated.
4B) The interaction frequencies of the user u and the strong relationship friends s form an interaction frequency binary (f1, f 2).
4C) And 3 interactive duplets are randomly generated by using a random function in java language as 3 cluster centers.
4D) And dividing the strong relation friends into 3 different clusters according to different interaction frequency binary groups by using a KMeans algorithm.
4E) The center of 3 clusters was calculated using an averaging method.
4F) Repeating steps 4D) -4E) until the cluster center converges.
Whether the function is converged can be judged in two ways, one way is to compare whether the difference between the two calculation results before and after the function is within an allowable error range, and if the difference is within the allowable error range, the calculated values of the two calculation steps tend to be consistent, the function can be considered to be converged; another is to define the number of repeated executions based on experience, and when the calculation is repeatedly executed for a specified number of times, the function is considered to be converged.
Next, the click probability of the video is predicted based on the behavior characteristics (publication, praise, forward, comment, etc.) of the weak relation friend and the strong relation friend. Training according to the historical information of the user u to obtain the weight of the friend interaction behavior characteristic, and finally accumulating according to the weight of the friend interaction behavior characteristic to obtain the click probability of the user on a new video t, wherein the steps are as follows:
1) the influence (w) of the weak-relationship friend of user u is calculated according to the flow of fig. 3.
2) And selecting the first 10 most influential weak relation friends of the user u to form a weak relation friend set.
3) Calculating the correlation r of the user u and the weak relation friend w according to the ratio of the total number of the videos t of the weak relation friend w clicked by the user u to the total number of the videos of the user uu,w。
4) According to the interactive behavior characteristics f (like praise, comment and forwarding) of the weak relation friend to the video t, accumulating and summing to obtain a first estimated click probability of the user to the video t
F represents the behavior characteristics of the weak relation friend to the video t, F represents the behavior characteristic set of the weak relation friend to the video t, bf,wWeights for the behavioral characteristics f of weakly related friends, initialized by a Gaussian function, It f,wIs an indication function which indicates whether the weak-relation friend has the past behavior on the video t, if so, It f,w1, otherwise It f,w＝0。
5) According to the flow of fig. 4, the strong-relationship friends of the user u are divided into friend clusters K of close friends, familiar friends, and unfamiliar friends, and the cluster size is 3.
6) Calculating a second estimated click probability of the user to the video t through the user behavior characteristics f of strong-relation friends in different clusters K to the video t
Wherein b isf,sIs the weight of the behavior feature f of the strong relationship friend in the cluster K,
is an indication function which indicates whether strong-relation friends in the cluster K have over-interactive behaviors with the video t, if so, the strong-relation friends in the cluster K have over-interactive behaviors with the video t
Otherwise
7) And calculating the average click rate a of the user u to the videos of all the friends.
8) Initializing click bias parameter b of user u by Gaussian function0，b0Indicating the preference of user u to click on the video.
9) Linearly fusing average click rate a and bias parameter b0Obtaining a first estimated click probability according to the friends with weak relation
And a second estimated click probability obtained according to the strong relation friend
Obtaining the estimated click probability of the user to the video t
10) Calculating the observed value r of the user u to the video ttIf the user has past behavior on the video t, then rt1, otherwise rt＝0。
11) Will observe the value rtAnd an estimate
As a function of the loss
Where T represents a collection of videos that the user has been behaving.
12) To prevent overfitting that leads to learning in optimizing the loss function, the L2 norm of the parameters is added and the final loss function can be expressed as
13) Optimization of learning parameter b by adopting random gradient SGD algorithm0、bf,wAnd bf,sThe value of the loss function is minimized.
14) The click probability of the end user u to a new video t can be obtained according to the accumulated sum of the interaction behavior characteristic weights of the friends with the weak relationship of the user
And the cumulative sum of the interactive behavior characteristic weights of the strong relationship friend clusters
To calculate both and the offset value b of the user u0And the linear combination of the average value a obtains the click probability p of the user u to the video tt＝a+b0+pt 1+pt 2。
15) And finally recommending the 10 videos with the highest click probability to the user.
The method of the invention predicts the click rate based on the social network. And according to the interaction behavior characteristics of the weak-relation friends and the strong-relation friends in the social network, linearly fusing and predicting the video click rate of the user.
Although the preferred embodiments of the present invention have been described in detail, the present invention is not limited to the details of the embodiments, and various equivalent modifications can be made within the technical spirit of the present invention, and the scope of the present invention is also within the scope of the present invention.
Claims (8)
1. A video recommendation method based on weak relations and strong relations of a social network is characterized by comprising the following steps:
1) calculating a first estimated click probability of the user on the video based on the influence of the weak-relation friend of the user, the correlation degree of the user and the weak-relation friend and the behavior characteristics of the weak-relation friend on the video;
2) calculating a second estimated click probability of the user to the video based on behavior characteristics of strong-relation friends with different intimacy to the video;
3) linearly fusing the first estimated click probability and the second estimated click probability to obtain the estimated click probability of the user on the video;
4) constructing a loss function based on the behavior observation value of the user to the video and the estimated click probability of the user to the video;
5) optimizing the characteristic parameters in the steps 1) -4) by adopting a random gradient algorithm to minimize the value of the loss function, and stopping optimization when the value of the loss function meets a preset condition to finally obtain the click probability of the user on the video;
6) and recommending the N videos with the highest click probability to the user by using a TopN algorithm.
2. The social network-based weak relation and strong relation video recommendation method according to claim 1, wherein the step 1) specifically comprises the following steps:
11) calculating the influence (w) of the weak relation friend of the user;
12) selecting n weak relation friends with the largest influence to form a weak relation friend set Wn；
13) Obtaining the correlation degree r of the user and the weak relation friend according to the ratio of the total number of the videos of the weak relation friend clicked by the user to the total number of the videos of the useru,w；
14) Calculate the first of the user to the video t using the following formulaPredicting click probability
Wherein F represents the behavior characteristics of the weak-relation friend to the video t, F represents the behavior characteristic set of the weak-relation friend to the video t, and bf,wWeights for the behavior characteristics f of weakly related friends w, initialized by a Gaussian function, It f,wIs an indication function which indicates whether the weak-relation friend w has the past behavior on the video t, if so, It f,w1, otherwise It f,w＝0。
3. The social network-based weak relation and strong relation video recommendation method according to claim 2, wherein the step 11) specifically comprises the following steps:
11a) initializing the influence (u) of the user u through a Gaussian function;
11b) obtaining the forwarding rate Re (u) of the user according to the ratio of the number of times of forwarding the video sent by the user u to the total number of the videos sent by the user u;
11c) mapping the user forwarding rate Re (u) to be between 0 and 1 by adopting a Min-Max normalization method;
11d) the influence of the weak relationship friend is calculated according to the following formula:
wherein d is the probability of the user walking, namely the probability of the user clicking other links, which is initialized by a Gaussian function, B represents the fan set concerning the weak-relation friends, and out (u) is the number of friends concerned by the user u;
11e) and repeating the steps 11b) -11d) until the influence of the weak relation friend is converged.
4. The social network-based weak relation and strong relation video recommendation method according to claim 1, wherein the step 2) specifically comprises the following steps:
21) obtaining a first interaction frequency f1 of the user u and the strong relation friends s according to the ratio of the total number of the interaction behaviors of the user u to the videos of the strong relation friends s to the total number of the videos of the user u;
22) obtaining a second interaction frequency f2 of the strong relation friends s and the user u according to the ratio of the total number of the interaction behaviors of the strong relation friends s to the video of the user u to the total number of the videos of the strong relation friends s;
23) the first interaction frequency f1 and the second interaction frequency f2 form an interaction frequency binary group (f1, f2), and strong relation friends are divided into friend clusters K with different intimacy degrees by using a KMeans algorithm according to the interaction frequency binary group (f1, f 2);
24) calculating second estimated click probability of the user to the video t through behavior characteristics f of strong-relation friends with different intimacy degrees to the video t
Wherein, bf,sIs the weight of the behavioral characteristic f of the strong-relation friends s in the cluster K, which is initialized by a gaussian function,
is an indication function, which indicates whether the strong-relation friend in the cluster K has an interaction behavior f to the video t, if so, the strong-relation friend in the cluster K has an interaction behavior f to the video t
Otherwise
5. The social network-based weak relation and strong relation video recommendation method according to claim 1, wherein the step 3) is linear fusion
And
the method specifically comprises the following steps:
31) calculating the average click rate a of the user u to the videos of all the friends;
32) initializing click bias parameter b of user u by Gaussian function0，b0Representing the preference of the user u to click on the video;
33) linearly fusing average click rate a and bias parameter b0Obtaining a first estimated click probability according to the friends with weak relation
And a second estimated click probability obtained according to the strong relation friend
Obtaining the estimated click probability of the user to the video t
6. The social network-based weak relation and strong relation video recommendation method according to claim 1, wherein the specific method for constructing the loss function in the step 4) comprises the following steps:
61) defining a loss function
Where T represents a collection of videos in which the user has been behaving, rtRepresenting the observed value of the user u to the video t, and if the user has excessive behavior to the video t, rt1, otherwise rt＝0；
62) Modifying the loss function by adding the L2 norm of the parameter
7. According to the claimsSolving 1 the method for recommending videos based on weak relations and strong relations of social networks, wherein the characteristic parameters optimized in the step 5) comprise b0、bf,wAnd bf,sWherein b is0Indicates the preference of user u to click on the video, bf,wWeight of behavior feature f for Weak relationship friend w, bf,sIs the weight of the behavior feature f of the strong relationship friend s.
8. The social network based weak relationship and strong relationship video recommendation method according to any one of claims 1-7, wherein the behavior comprises posting, mentioning, praise, commenting, replying, forwarding and favorites.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710371550.8A CN107220328B (en) | 2017-05-23 | 2017-05-23 | Social network-based weak relation and strong relation video recommendation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710371550.8A CN107220328B (en) | 2017-05-23 | 2017-05-23 | Social network-based weak relation and strong relation video recommendation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107220328A CN107220328A (en) | 2017-09-29 |
| CN107220328B true CN107220328B (en) | 2020-05-19 |
Family
ID=59944949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710371550.8A Active CN107220328B (en) | 2017-05-23 | 2017-05-23 | Social network-based weak relation and strong relation video recommendation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107220328B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107911719B (en) * | 2017-10-30 | 2019-11-08 | 中国科学院自动化研究所 | Video Dynamic recommendation device |
| CN108108419B (en) * | 2017-12-15 | 2021-01-26 | 百度在线网络技术（北京）有限公司 | Information recommendation method, device, equipment and medium |
| CN110968771A (en) * | 2018-09-29 | 2020-04-07 | 北京淘友天下技术有限公司 | Position recommendation cold start method and system based on friendship |
| CN109670077B (en) * | 2018-11-01 | 2021-07-13 | 北京达佳互联信息技术有限公司 | Video recommendation method and device and computer-readable storage medium |
| CN110377521B (en) * | 2019-07-23 | 2022-11-18 | 郑州阿帕斯科技有限公司 | Target object verification method and device |
| CN112328865B (en) * | 2019-07-31 | 2024-01-05 | 北京达佳互联信息技术有限公司 | Information processing and recommending method, device, equipment and storage medium |
| CN110647916B (en) * | 2019-08-23 | 2022-10-28 | 苏宁云计算有限公司 | Pornographic picture identification method and device based on convolutional neural network |
| CN110516153B (en) * | 2019-08-28 | 2023-05-30 | 腾讯科技（深圳）有限公司 | Intelligent video pushing method and device, storage medium and electronic device |
| CN113360777B (en) * | 2021-08-06 | 2021-12-07 | 北京达佳互联信息技术有限公司 | Content recommendation model training method, content recommendation method and related equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106168953A (en) * | 2016-06-02 | 2016-11-30 | 中国人民解放军国防科学技术大学 | Blog article towards weak relation social networks recommends method |
-
2017
- 2017-05-23 CN CN201710371550.8A patent/CN107220328B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106168953A (en) * | 2016-06-02 | 2016-11-30 | 中国人民解放军国防科学技术大学 | Blog article towards weak relation social networks recommends method |
Non-Patent Citations (5)
| Title |
|---|
| On Video Recommendation over Social Network;Xiaojian Zhao等;《Advances in Multimedia Modeling》;20121231;全文 * |
| 基于用户行为特征的微博转发预测研究;刘玮等;《计算机学报》;20161031;第39卷(第10期);全文 * |
| 基于社交网络的推荐算法应用研究;孟婷婷;《中国优秀硕士学位论文全文数据库信息科技辑》;20160315;全文 * |
| 基于社交网络的视频推荐技术研究;孙莉莉;《中国优秀硕士学位论文全文数据库信息科技辑》;20160215;全文 * |
| 面向微博的Pagerank 算法的改进与应用;原野等;《计算机应用与软件》;20170331;第34卷(第3期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107220328A (en) | 2017-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107220328B (en) | Social network-based weak relation and strong relation video recommendation method | |
| Xia et al. | Reciprocal recommendation system for online dating | |
| Yuan et al. | COM: a generative model for group recommendation | |
| Sewell et al. | Latent space models for dynamic networks | |
| Song et al. | Online learning in large-scale contextual recommender systems | |
| US9122989B1 (en) | Analyzing website content or attributes and predicting popularity | |
| US10291559B1 (en) | Systems and method for communications routing based on electronic communication data | |
| CN108776844B (en) | Social network user behavior prediction method based on context perception tensor decomposition | |
| US8909546B2 (en) | Privacy-centric ad models that leverage social graphs | |
| CN107330798B (en) | Method for identifying user identity between social networks based on seed node propagation | |
| Pan et al. | Diffusion-aware personalized social update recommendation | |
| US10785181B2 (en) | Sharing content to multiple public and private targets in a social network | |
| US20140059141A1 (en) | Electronic messaging system utilizing social classification rules | |
| US9946799B2 (en) | Federated search page construction based on machine learning | |
| WO2013012872A1 (en) | Method for social search | |
| US20170124472A1 (en) | Activity sensing online preference assay | |
| Koroleva et al. | Reducing information overload: Design and evaluation of filtering & ranking algorithms for social networking sites | |
| Lin et al. | Steering information diffusion dynamically against user attention limitation | |
| CN106991496B (en) | User behavior hierarchical association prediction method oriented to mobile social environment | |
| Zhou et al. | Continual information cascade learning | |
| CN107729569B (en) | Social relationship prediction method fusing network structure and text information | |
| Xiong et al. | An estimation model for social relationship strength based on users' profiles, co-occurrence and interaction activities | |
| US10180990B2 (en) | Activity sensing online preference assay | |
| Jamil et al. | A collaborative names recommendation in the Twitter environment based on location | |
| Samir et al. | Enhancing Multi-factor Friend Recommendation in Location-based Social Networks |
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 |