CN106649714B - TopN recommendation system and method for data nonuniformity and data sparsity - Google Patents
TopN recommendation system and method for data nonuniformity and data sparsity Download PDFInfo
- Publication number
- CN106649714B CN106649714B CN201611189954.7A CN201611189954A CN106649714B CN 106649714 B CN106649714 B CN 106649714B CN 201611189954 A CN201611189954 A CN 201611189954A CN 106649714 B CN106649714 B CN 106649714B
- Authority
- CN
- China
- Prior art keywords
- user
- data
- time
- item
- model
- 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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Abstract
The invention requests to protect a topN recommendation system and method aiming at data nonuniformity and data sparseness, and belongs to the field of data mining and information retrieval. On the basis of the relationship between a user and social network information friends, considering implicit relations between scores and the personal activeness of the user, the popularity of articles and the interest degree of the user on the articles, adding three-aspect time effect influence through a time discretization and time slicing method, meanwhile, aiming at the problems of uneven distribution of user score data and data sparsity, constructing a GM (1, N) prediction model based on a gray theory, and mining the explicit dynamic relationship between the scores and the three aspects. And inputting data into the prediction model to predict the item rating of the user, recommending the item which the user is interested in to the user according to the user rating condition, and realizing topN recommendation.
Description
Technical Field
The invention belongs to the field of data mining and information retrieval, relates to personalized recommendation of a recommendation system, and provides a personalized recommendation scheme for solving the problems of uneven data and sparse data based on a gray system for a social network.
Background
With the rapid development of information technology, various information on the network is rapidly increased, and the information overload problem is followed. In order to solve the problems caused by the massive information, numerous scientists and engineers continuously provide new solutions, such as strengthening a search engine, optimizing a recommendation system and the like.
In recent years, with the increasing scale of the internet, recommendation systems are becoming more and more popular. The development of the personalized recommendation technology plays an important role in improving user experience and improving service quality. The directions of recommendation system research can be divided into two categories: prediction and topN recommendations. The former is a score estimate, and the unknown user score is predicted from the known user historical score. The latter is to give a user a personalized recommendation list comprising N items. And existing recommendations can be classified into a content-based recommendation and a collaborative filtering-based recommendation. Content-based recommendations recommend items to a user that may be of interest by calculating a degree of match between known user preferences and content characterized by the item attributes. And searching for a potential interest object of the current user by referring to the interest characteristics of the neighboring users by using the historical behaviors of the users based on the collaborative filtering algorithm, and further carrying out personalized recommendation. Because the content-based system is difficult to distinguish the quality and style of the resource content, cannot discover new interesting resources for the user, and can only discover resources similar to the existing interests of the user, the system based on collaborative filtering has been widely researched and applied in the industry.
The collaborative filtering based recommendation system can find new interesting information for users, but has two problems which are difficult to solve, one is sparseness problem caused by the rapid increase of the number of users and the number of items, and the other is expandability problem of the system performance reduction along with the increase of users and resources. Under the condition that user scoring data is extremely sparse, the recommendation quality of the traditional recommendation system is sharply reduced. In addition, user interest and articles in the recommendation system are time-efficient, and the accuracy of recommendation can be greatly improved by modeling time effect into a model of the recommendation system. However, due to uncertainty of user behavior, the scoring time of the user for the project is also uneven, which causes certain difficulty in time series analysis of user scoring. Therefore, solving the problems of data non-uniformity and sparseness is very important for the quality of topN recommendation.
Due to the problems of non-uniform data and sparse data of the recommendation system, a good recommendation effect cannot be achieved based on the traditional collaborative filtering algorithm. The research object of the grey system theoretical model is an uncertain system with characteristics of small data and poor information, a dynamic model of a multivariable system can be described, and the dominant relation between system variables and related factor variables can be mined. The grey prediction model can construct a system model by using a small amount of data and can obtain good prediction accuracy, so that the problem of data sparsity can be relieved to a certain extent by using the grey prediction model. However, how to construct a scoring prediction model based on the gray system theory and how to process the uneven time series is the difficulty of the research.
Disclosure of Invention
The problems to be solved by the invention are as follows: the method aims at the problems of data nonuniformity and data sparseness caused by uncertain user scores and rapid increase of user number and project data. The invention provides a topN recommendation scheme based on an improved grey theoretical model. According to the scheme, the dominant relations between the user scores and the user liveness, the item popularity and the user interest in the items are explored from the correlation attributes of the users and the items, the timeliness factors of the user liveness and the item popularity are considered, and the social network information of the users is considered, so that the prediction accuracy and the recommendation quality are improved. A recommendation system and method aiming at data nonuniformity and data sparseness are provided, and the accuracy of prediction and the quality of recommendation are improved. The technical scheme of the invention is as follows:
a recommendation system for uneven data and sparse data comprises a data source acquisition module, an attribute analysis module, a model construction module and a prediction recommendation module, wherein the data source acquisition module is used for acquiring user information, article information and user social network friend information; the attribute analysis module extracts relevant attribute information from the user information set, the concern list set and the project information set which are acquired by the data source acquisition module, separates the relevant attribute through time slicing and time, and then defines three relevant function indexes of user activity, article popularity and user interest in the articles by using the relevant attribute with the time information; meanwhile, the user scoring observation sequence is separated by using a time attenuation function, and a grey system theoretical model is improved; the model construction module is used for mining the explicit relation between the user score and the index to be evaluated by utilizing the improved grey system theoretical model and mastering the user score rule; and the prediction recommendation module is used for inputting the relevant data of any time t into the prediction model to predict the scores of the user on the articles at the time and recommending the N articles with the highest scores to the user.
Further, the construction model module excavates the explicit relation between the user score and the index to be evaluated, and the explicit relation comprises the following three indexes: firstly, whether a user scores an item or not, the time for the user to score the item, and the category of the item; secondly, label information of the user and friends thereof and the use times of each label; third, a set of user scores and a set of users who score items.
Further, the three related function indexes of the user activity, the item popularity and the user interest in the item represent the related relation between the three indexes and time according to a formula:an activity index function representing the activity of user u on item i at time T, whereinTo count the total number T of the user scoring the project within the time Ttotal,To count the total number T of the item scores of the categories to which the item i belongsiWherein if user u overacts on item j, then ruj1, otherwise ruj0; according to the formula:denotes the popularity indicator function of item i at time T, wherein α1The attenuation parameter is t, and the user scores the item for time; according to the formulaRepresenting the user' S interest in the item at time T, where S (u, K) contains K users with similar interests to user u, where we replace the K users with social networking information user friends, where α2Is an attenuation parameter.
Further, the user's interest in the item at time TAnd calculating the similarity between the users, wherein N (u, v) represents the user u and the label set of the user v, and N (b, u) represents the number of times of using the label b by the user u.
Further, the improved grey system theoretical model comprises: a data pre-processing part for pre-processing the data,a sequence equal interval part and a model building part; the data preprocessing part mainly removes the part of data with unreasonable scoring time or excessive aggregation; the sequence equal interval part passes the non-equal interval time sequence through the time attenuation functionα therein3The model building part builds a score prediction model by using the content of a gray GM (1, N) model, and then estimates model parameters by using a least square estimation method to obtain a state equation of the GM (1, N) model.
Further, three index values at time t are calculated by using three index functions, and the three index values are input into a time response formula to obtain a score to be predicted. And ranking the predicted scores from large to small, and recommending the N items with the highest scores to the user.
The recommendation method aiming at data nonuniformity and data sparseness according to the system comprises the following steps:
acquiring user information, article information and social network friend information;
extracting relevant attribute information from a user information set, an attention list set and a project information set in an acquired data source; adding timeliness to the attributes through a time slicing and time dispersion method, and then defining three related function indexes of user activity, article popularity and user interest to the articles by using related attributes with time information; meanwhile, the user scoring observation sequence is equally spaced by using a time attenuation function, and a grey system theoretical model is improved;
utilizing the improved gray GM (1, N) model to mine the dominant relation between the user score and the three-aspect indexes and master the user score rule;
the item scoring of the user at any moment can be predicted by inputting the relevant data of the moment t into the prediction model, and the N items with the highest scoring are recommended to the user.
The invention has the following advantages and beneficial effects:
the invention provides a user score prediction model based on three indexes influencing user score, and topN recommendation is carried out based on the user score prediction model. The three indexes in the method are user activity, item popularity and user interest in the items respectively. The method has the key points that relevant attributes are extracted, relevant function indexes are defined, the influence of timeliness factors of the indexes is quantized, a time attenuation function is introduced to improve a grey system theoretical model, a new application thought is provided for the grey system theoretical model, and a new dynamic scoring prediction model is provided. By the method, the defects of the traditional recommendation system are overcome, and the recommendation accuracy is improved.
Considering that the relevant factors influencing the user score are mainly expressed in three aspects of user activity, item popularity and user interest in the items, the method provided by the invention is based on the time discretization and time slicing method aiming at the characteristics of the three aspects, and the influence of the timeliness factors of the three aspects is added, so that the relevant indexes influencing the user score are defined. When calculating the interest of a user to an article, the social network data is directly utilized in consideration of higher complexity and longer time consumption when searching for users with similar interests.
Meanwhile, aiming at the problem of uneven time distribution of user scoring data, a time attenuation function is introduced, a gray GM (1, N) model is improved, the application range of the model is expanded, and the user interest can be dynamically monitored. And then aiming at the problem of sparsity of user data, based on implicit connection between the scores and the three indexes, constructing an improved grey GM (1, N) prediction model by using ideas and methods of a grey system theory to predict the scores of the user, mining an explicit relation between the scores and the three indexes, and simultaneously generating a real-time topN recommendation set.
Drawings
FIG. 1 is a general flow diagram of the preferred embodiment of the present invention
FIG. 2 is a flow chart for constructing an improved gray prediction model.
Detailed Description
The technical solutions in the embodiments of the present invention will be described in detail and clearly with reference to the accompanying drawings. The described embodiments are only some of the embodiments of the present invention.
The technical solution of the present invention for solving the above technical problems is,
fig. 1 shows a general flowchart of the present invention, which includes four modules, namely a data acquisition module, an attribute analysis module, a model construction module, and a prediction recommendation module. The detailed implementation process of the invention comprises the following four steps:
s1: a data source is acquired. The data source may be obtained by downloading directly from existing Web-based research recommendation systems or by using the public API of a sophisticated social platform. And (4) data preprocessing is carried out to obtain a user information set, an attention list set and a project information set.
S11: raw data is acquired. And acquiring data disclosed to the public through a public API of the social network, and supplementing the data by combining methods such as a web crawler and the like.
S111: and acquiring the rating of a certain user on the project, the rating time, all tags used for labeling the project, the use times of each tag and the basic information of the user.
S112: and acquiring the friend list of the user and the information of the friends.
S12: simple data cleaning. The data is analyzed through simple data cleaning, and the data cleaning comprises repeated data, invalid data and the like.
S13: simple data preprocessing. And processing the original data into a user information set, an attention list set and an item information set.
S2: and extracting the relevant attributes. And respectively extracting the attributes of the information of the user information set, the user scoring item set, the attention list information set and the attention user scoring item set according to an attribute extraction method. Then the influence between the training data time slicing, specific quantitative index and time is carried out. Defining and calculating the user activity, the item popularity and the user interest correlation function indexes of the items, and quantifying the influence of the three indexes and time.
S21: and extracting the activity related attributes of the user. Whether the user scores an item rujTime of scoring t, category C to which item i belongsiWhether the user is to the objectProduct score rujUser u rates the set of items N (u). According to the formulaTo count the total number T of the user scoring the project within the time TtotalAccording to the formulaTo count the total number T of the item scores of the categories to which the item i belongsiWherein if user u overacts on item j, then ruj1, otherwise ruj=0。
S22: and extracting the article popularity related attributes. A set of users n (i) who rate item i, and a rating time t of the user.
S23: and extracting the attribute related to the interest degree of the user in the article. A friend list set of a user, a friend scoring item set, Similarity (u, v) between the user and a friend, and whether the user generates over-scoring behavior r on the itemuiIf user u overacts on item j, then rui1, otherwise ruiScore time t of 0ui。
The Similarity (u, v) between the user and the friend can be determined by the following formula:
wherein, N (u, v) represents the label set of the user u and the user v, and N (b, u) represents the number of times of using the label b by the user u.
S24: after the relevant attributes of the three aspects are extracted, index functions of the three aspects are obtained, and the index functions quantify the influence of the three aspects and time. The formula is as follows:
① user activity indicator function:
wherein, mului(T) is represented byThe value of the user activity of the user at time T,represents the total number T of the user's scores to the item within the time Ttotal,Total number of item scores T representing categories to which item i belongsi。
② item popularity indicator function:
wherein mipi(T) indicates the popularity of item i at time T, α1The attenuation factor can be set according to the popularity variation.
③ index function of the degree of interest of a user in an item:
wherein, muiui(T) represents the degree of interest of user u in item i at time T, Simlarity (u, v) represents the similarity of users u and v, α2Is the attenuation factor.
S3: an improved gray GM (1, N) model was constructed. Fig. 2 shows a flow chart of the improved model construction. The user scoring sequence is equally spaced in time by using a time attenuation function, and then a prediction model of the user scoring is constructed according to the thought and the method of a grey system theory. The model shows that the user scoring feature sequences are mainly related to three related sequences. Time series of user activity, time series of item popularity, and data series of user interest in the item changing with time. The gray system GM (1, N) model expresses an explicit relationship of user scores to these three aspects.
S31: the scoring sequence is equally spaced. Due to the fact that the grading observation sequence is not uniformly distributed in time caused by uncertainty of user behaviors, the original grading sequence is processed by introducing a time attenuation function, and the processed grading sequence has the characteristic of equal time intervals. Whereas the grey system theoretical model is mostly based on a sequence of equal time intervals. Therefore, the scoring sequence equidistant process expands the application range of the grey theoretical model and provides a new application idea for the grey system theoretical series model.
S311: the sequence averaging interval is calculated. Let X(0)={x(0)(t1),x(0)(t2),...,x(0)(tn) Is the original scoring sequence, where Δ ti=ti-ti-1Not equal to const is a non-equidistant sequence. Let Δ k be the average interval of the sequences, the calculation formula is:
s312: a time series of equal intervals. Is X'(0)={x(0)(k1),x(0)(k2),...,x(0)(kn) Is the scoring sequence after the equal interval processing, wherein ki-ki-1=Δk,k1=t1. The formula of the equal interval processing is x(0)(ki)=x(0)(ti)*f(|ti-kiI ≠ 1, where f (| t)i-ki|) is a time decay function, and the mathematical expression is:
s32: the grey theory GM (1, N) model was constructed. The grey system prediction method is to mine, discover and master the user score and the explicit rules of the user activity, the item popularity and the evolution of the user interest degree of the item through the processing of the original data and the establishment of the grey model. The grey GM (1, N) model is built as follows.
S321: constructing a GM (1, N) basic model. Setting the processed equal time interval sequence as a characteristic observation sequence of the system: x'(0)={x(0)(k1),x(0)(k2),...,x(0)(kn)}. Taking the value sequence of the time discretization user activity with time information, the article popularity and the article interest degree of the user as a related factor sequence:
X‘(1)is X‘(0)The 1-AGO sequence of (a),the other sequences are similar. Z(1)Is X‘(1)Wherein z is(1)(ki)=βx(1)(ki-1)+(1-β)x(1)(ki). The resulting GM (1, N) model is then as follows:
the matrix equation is expressed as Y-B × deltaT. Wherein, Delta is [ lambda, gamma ]1,γ2,γ3],
The least estimation value of the parameter column delta can be obtained by utilizing a least square algorithm
S4: prediction and recommendation processes. The input of the model is the values of the three indexes at the T moment, and the output is the score of the user at the T moment. The input data is input into the prediction model, and the prediction scores of the unscored items can be calculated. And analyzing the user interest according to the predicted scores to generate a recommendation set with the length of N.
Aiming at the problems of non-uniform data, sparse data and the like in a recommendation system, the invention introduces and improves a grey system theory related model, and further realizes a novel social network topN recommendation scheme on the basis. Firstly, relevant factors influencing the user score are found out through analyzing the scoring behavior of the user, and the influence of three indexes and time is quantified by using a time dispersion and time slicing method from three indexes of the individual attribute of the user, the popularity of the item and the interest degree of the user in the item; then aiming at the problem of non-uniformity of the scoring observation sequence in time, introducing a time attenuation function, improving a theoretical GM (1, N) model of a gray system, and expanding the application range of the gray model; and then, an improved grey system theoretical model is utilized to mine the explicit relation between the user score and the three indexes, a score prediction algorithm is constructed to generate a score, and the accuracy of recommendation of the target user interest list is improved.
The above examples are to be construed as merely illustrative and not limitative of the remainder of the disclosure. After reading the description of the invention, the skilled person can make various changes or modifications to the invention, and these equivalent changes and modifications also fall into the scope of the invention defined by the claims.
Claims (6)
1. A topN recommendation system aiming at data nonuniformity and data sparseness is characterized by comprising a data source acquisition module, an attribute analysis module, a model construction module and a prediction recommendation module, wherein the data source acquisition module is used for acquiring user information, article information and user social network friend information; the attribute analysis module extracts relevant attribute information from the user information set, the concern list set and the project information set which are acquired by the data source acquisition module, separates the relevant attribute through time slicing and time, and then defines three relevant function indexes of user activity, article popularity and user interest in the articles by using the relevant attribute with the time information; meanwhile, the user scoring observation sequence is separated by using a time attenuation function, and a grey system theoretical model is improved; the model construction module is used for mining the explicit relation between the user score and the index to be evaluated by utilizing the improved grey system theoretical model and mastering the user score rule; the prediction recommendation module is used for inputting the relevant data of any time t into the prediction model to predict the scores of the user to the articles at the time and recommending the N articles with the highest scores to the user;
the three related function indexes of the user activity, the item popularity and the user interest in the items represent the related relation between the three indexes and time, and according to a formula:an activity index function representing the activity of user u on item i at time T, whereinT < T to count the total number T of the user scoring the item within the time Ttotal,T < T to count the total number T of the item scores of the categories to which the item i belongsiWherein if user u overacts on item j, then ruj1, otherwise ruj0; according to the formula:denotes the popularity indicator function of item i at time T, wherein α1The attenuation parameter is t, and the user scores the item for time; according to the formulaRepresenting the user' S interest in the item at time T, where S (u, K) contains K users with similar interests to user u, and replacing the K users with similar interests with social networking information user friends, where α2Is an attenuation parameter.
2. The topN recommendation system for uneven data and sparse data as claimed in claim 1, wherein the model construction module mines an explicit relationship between the user score and an index to be evaluated, and comprises the following three indexes: firstly, whether a user scores an item or not, the time for the user to score the item, and the category of the item; secondly, label information of the user and friends thereof and the use times of each label; third, a set of user scores and a set of users who score items.
3. The topN recommendation system for data non-uniformity and data sparseness of claim 1, wherein said user's interest in items at T timeAnd calculating the similarity between the users, wherein N (u, v) represents the user u and the label set of the user v, and N (b, u) represents the number of times of using the label b by the user u.
4. The topN recommendation system for data non-uniformity and data sparsity according to claim 1, wherein said improved gray system theoretical model comprises: a data preprocessing part, a sequence equal interval part and a model building part; the data preprocessing part mainly removes the part of data with unreasonable scoring time or excessive aggregation; the sequence equal interval part passes the non-equal interval time sequence through the time attenuation functionα therein3The model building part builds a score prediction model by using the content of a gray GM (1, N) model, and then estimates model parameters by using a least square estimation method to obtain a state equation of the GM (1, N) model.
5. The topN recommendation system for uneven data and sparse data according to claim 4, wherein three index values at time t are calculated by using three index functions, the three index values are input into a time response formula to obtain scores to be predicted, the predicted scores are ranked from large to small, and the N items with the highest scores are recommended to the user.
6. The recommendation method for data nonuniformity and data sparseness of the system according to claim 1, comprising the steps of:
acquiring user information, article information and social network friend information;
extracting relevant attribute information from a user information set, an attention list set and a project information set in an acquired data source; adding timeliness to the attributes through a time slicing and time dispersion method, and then defining three related function indexes of user activity, article popularity and user interest to the articles by using related attributes with time information; meanwhile, the user scoring observation sequence is equally spaced by using a time attenuation function, and a grey system theoretical model is improved;
utilizing the improved gray GM (1, N) model to mine the dominant relation between the user score and the three-aspect indexes and master the user score rule;
the item scoring of the user at any moment can be predicted by inputting the relevant data of the moment t into the prediction model, and the N items with the highest scoring are recommended to the user.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611189954.7A CN106649714B (en) | 2016-12-21 | 2016-12-21 | TopN recommendation system and method for data nonuniformity and data sparsity |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611189954.7A CN106649714B (en) | 2016-12-21 | 2016-12-21 | TopN recommendation system and method for data nonuniformity and data sparsity |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106649714A CN106649714A (en) | 2017-05-10 |
CN106649714B true CN106649714B (en) | 2020-08-04 |
Family
ID=58834382
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611189954.7A Active CN106649714B (en) | 2016-12-21 | 2016-12-21 | TopN recommendation system and method for data nonuniformity and data sparsity |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106649714B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109063979A (en) * | 2018-07-13 | 2018-12-21 | 四川享宇金信金融服务外包有限公司 | A kind of air control model building method |
CN108876193A (en) * | 2018-07-13 | 2018-11-23 | 四川享宇金信金融服务外包有限公司 | A kind of air control model building method based on credit score |
CN111125428B (en) * | 2019-12-17 | 2021-11-05 | 东北大学 | Time-dependent movie recommendation method based on score prediction function fitting structure |
CN112699174B (en) * | 2021-03-23 | 2021-08-17 | 中建电子商务有限责任公司 | Big data building product drive chain diagram generation method |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7552078B2 (en) * | 2005-02-28 | 2009-06-23 | International Business Machines Corporation | Enterprise portfolio analysis using finite state Markov decision process |
US20070244747A1 (en) * | 2006-04-14 | 2007-10-18 | Nikovski Daniel N | Method and system for recommending products to consumers by induction of decision trees |
CN101814068A (en) * | 2009-02-24 | 2010-08-25 | 日电(中国)有限公司 | Rating prediction based project recommending method for time-sequence control and system thereof |
CN101551825A (en) * | 2009-05-15 | 2009-10-07 | 中国科学技术大学 | Personalized film recommendation system and method based on attribute description |
CN103116588B (en) * | 2011-11-17 | 2017-07-04 | 深圳市世纪光速信息技术有限公司 | A kind of personalized recommendation method and system |
US9152948B2 (en) * | 2012-02-20 | 2015-10-06 | Yahoo! Inc. | Method and system for providing a structured topic drift for a displayed set of user comments on an article |
CN103400286B (en) * | 2013-08-02 | 2016-12-28 | 世纪禾光科技发展(北京)有限公司 | A kind of commending system and method carrying out article characteristics mark based on user behavior |
CN105205081B (en) * | 2014-06-27 | 2019-11-05 | 华为技术有限公司 | Item recommendation method and device |
CN104317900A (en) * | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
CN104408108B (en) * | 2014-11-18 | 2018-08-14 | 重庆邮电大学 | Much-talked-about topic group influence power analysis system based on gray system theory and method |
CN105205128B (en) * | 2015-09-14 | 2018-08-28 | 清华大学 | Sequential based on scoring feature recommends method and recommendation apparatus |
CN105809554B (en) * | 2016-02-07 | 2020-03-17 | 重庆邮电大学 | Prediction method for user participating in hot topics in social network |
CN106126607B (en) * | 2016-06-21 | 2019-12-31 | 重庆邮电大学 | User relationship analysis method facing social network |
-
2016
- 2016-12-21 CN CN201611189954.7A patent/CN106649714B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN106649714A (en) | 2017-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sigaki et al. | History of art paintings through the lens of entropy and complexity | |
CN107808278B (en) | Github open source project recommendation method based on sparse self-encoder | |
CN106649657B (en) | Social network oriented tensor decomposition based context awareness recommendation system and method | |
CN106649714B (en) | TopN recommendation system and method for data nonuniformity and data sparsity | |
CN104077357B (en) | Collaborative filtering combined recommendation method based on user | |
CN107562947B (en) | Method for establishing dynamic instant recommendation service model under mobile space-time perception | |
CN107944035B (en) | Image recommendation method integrating visual features and user scores | |
CN103136337A (en) | Distributed knowledge data mining device and mining method used for complex network | |
CN103678431A (en) | Recommendation method based on standard labels and item grades | |
CN103559262A (en) | Community-based author and academic paper recommending system and recommending method | |
CN110647678B (en) | Recommendation method based on user character label | |
CN108804577A (en) | A kind of predictor method of information label interest-degree | |
CN110932962B (en) | Network structure de-anonymization system and method based on matrix decomposition | |
Huang et al. | Information fusion oriented heterogeneous social network for friend recommendation via community detection | |
Chen et al. | A fuzzy matrix factor recommendation method with forgetting function and user features | |
CN108664558A (en) | A kind of Web TV personalized ventilation system method towards large-scale consumer | |
CN106650972B (en) | Recommendation system score prediction method based on cloud model and oriented to social network | |
CN116304332A (en) | Knowledge graph mixed recommendation method | |
CN109711651B (en) | Hidden factor model wecker task recommendation method with correction vector | |
Rusyaeva et al. | Integrative Approach to Creation of Information Systems and Entropy Analysis of Linguistic Information | |
Zhao et al. | Preference-aware Group Task Assignment in Spatial Crowdsourcing: Effectiveness and Efficiency | |
Negaresh et al. | Gender identification of mobile phone users based on internet usage pattern | |
CN106682204B (en) | Semantic extraction method based on crowdsourcing | |
Qin et al. | Recommender resources based on acquiring user's requirement and exploring user's preference with Word2Vec model in web service | |
Fang et al. | A process evaluation method for crossover service recommendation |
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 |