CN112784171B - Movie recommendation method based on context typicality - Google Patents
Movie recommendation method based on context typicality Download PDFInfo
- Publication number
- CN112784171B CN112784171B CN202110083203.1A CN202110083203A CN112784171B CN 112784171 B CN112784171 B CN 112784171B CN 202110083203 A CN202110083203 A CN 202110083203A CN 112784171 B CN112784171 B CN 112784171B
- Authority
- CN
- China
- Prior art keywords
- context
- user
- movie
- typicality
- representing
- 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
- G06F16/9536—Search customisation based on social or collaborative filtering
Abstract
The invention belongs to the field of computer science and technology, and particularly relates to a movie recommendation method based on context typicality, which comprises the following steps: acquiring user data, and constructing a context typicality matrix according to the user data and the context typicality of each user to different movie types; searching neighbors in the context typicality matrix, and calculating scores of neighbor users according to context attribute values of the neighbors; calculating the importance of the context attribute to each movie type by adopting knowledge granularity according to the grading information of the user data; calculating a multi-context prediction score according to the importance and the single context; recommending the film according to the multi-context prediction score; compared with the traditional collaborative filtering recommendation system, the method has the advantages that the traditional collaborative filtering is replaced by the typical collaborative filtering, and the sparsity of data is effectively relieved.
Description
Technical Field
The invention belongs to the field of computer science and technology, and particularly relates to a movie recommendation method based on context typicality.
Background
With the increasing data of the internet, the conventional Information Retrieval (IR) has been unable to meet the user's demand. In order to be able to determine the interests of a user and to present information of interest to the user, recommendation Systems (recommendation Systems) have become the most promising way for the new era of the internet.
In the field of recommendation systems, only depending on a user-item binary relationship, the recorded data has a sparsity problem in a scoring process of the system, and the depicted preference relationship is not reasonable. Traditional collaborative filtering algorithms measure a user using item scores, capturing user preferences in the low dimension of the item scores. When data is sparse, there are not enough common scoring items among users, which may cause inaccuracy in measuring similarity of users, so that the information recommended to the users by the system is inaccurate.
Disclosure of Invention
In order to solve the above problems in the prior art, the present invention provides a movie recommendation method based on contextual representativeness, the method comprising:
s1: acquiring user data, and collecting the user data to obtain a data set; calculating contextual typicality of the data according to information in the data; the user data comprises movie information, rating information and context attribute values;
s2: constructing a context typicality matrix according to the user data and the context typicality of each user to different movie types;
s3: searching neighbors in the context typicality matrix, and calculating scores of neighbor users according to context attribute values of the neighbors;
s4: predicting the score of the target user on the movie under the context attribute value by adopting the scores of the neighbor users;
s5: calculating the importance of the context attribute to each movie type by adopting knowledge granularity according to the grading information of the user data;
s6: calculating a multi-context prediction score according to the importance and the single context;
s7: the movie is recommended based on the multi-context prediction scores.
Preferably, the specific process of constructing the context user-typicality vector comprises:
step 1: calculating the user typicality of the user to the movie type under the context attribute value;
and 2, step: converting the contextual user-typicality of the user for each movie type into a contextual user-typicality vector; the expression of the context user-typicality vector is:
and 3, step 3: all context user-typicality vectors are constructed as a context-typicality matrix.
Further, the formula for user-typicality of a user to movie types under the context attribute values is:
further, the context typicality matrix is:
preferably, the searching of the context typical matrix comprises calculating cosine similarity between the user data in the context typical matrix and the target user data, and judging the magnitude of all cosine similarity, when the user data with the largest cosine similarity is the neighbor of the context typical matrix; the cosine similarity formula is:
preferably, the formula for predicting scores according to the single context of the neighbor users of the target user for different movies is as follows:
preferably, the formula for calculating the importance of the context attribute to each movie type by using the knowledge granularity is as follows:
preferably, the formula of the multi-context prediction score is:
the invention has the advantages that:
1) Compared with the traditional manual recommendation, the movie recommendation system provided by the invention has higher accuracy and efficiency;
2) Compared with the traditional collaborative filtering recommendation system, the method has the advantages that the typical collaborative filtering is used for replacing the traditional collaborative filtering, so that the data sparsity is effectively relieved;
3) The preference degree of the user to the items under the context is measured by the knowledge granularity, so that the preference degree can be described more accurately;
4) And the context is introduced, the user preference is measured in the context, the actual scene is more suitable, and the recommendation accuracy is improved.
Drawings
FIG. 1 is an overall flow diagram of the present invention;
FIG. 2 is a context-based exemplary matrix according to an embodiment of the present invention.
Detailed Description
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.
A movie recommendation method based on context typicality, as shown in fig. 1, the method comprising:
s1: acquiring user data, and collecting the user data to obtain a data set; calculating contextual typicality of the data from information in the data; the user data comprises movie information, rating information and context attribute values;
s2: constructing a context typicality matrix according to the user data and the context typicality of each user for different movie types;
s3: searching neighbors in the context typicality matrix, and calculating scores of neighbor users according to context attribute values of the neighbors;
s4: predicting the score of the target user on the movie under the context attribute value by adopting the scores of the neighbor users;
s5: calculating the importance of the context attribute to each movie type by adopting knowledge granularity according to the grading information of the user data;
s6: calculating a multi-context prediction score according to the importance and the single context;
s7: the movie is recommended according to the multi-context predictive score.
After the user data are obtained, the user data are collected to obtain a user data set; the user data includes movie information, rating information, and context attribute values. The data set of the user is U, U x Representing the xth user. In a data setComprises an item set O, a context attribute set C and attribute values in the context attribute set, wherein C h,k A kth attribute value representing an h context attribute. In the raw data set, historical scores for individual movies by individual users are also included.
After the user data is acquired, the types of the movies are classified into different types of groups T, T according to the types of the movies l Is the ith project group.
And generating a decision information table according to the classification result of the final movie, wherein the content in the decision information table is shown in table 1.
TABLE 1 decision information Table in context recommendation System
There is a lot of context information in the recommendation system, but some of them have a large influence on the decision of the user, and some have little influence on the decision. To determine which contexts are important for different types, the present invention bases its knowledge granularity G (d) y,t |C h,k ) To measure the importance of the context to different item types:
wherein d is y,t Indicating a certain film type t, C to which a film y belongs h,k A k-th attribute value representing an h-th context attribute,
set of scores, R, representing types of movies in the same single context attribute i Represents one score and n represents the nth.
Context typicals are built to measure user preferences.
The specific process of constructing the context typicality matrix comprises the following steps:
(1) Calculating the user typicality of the user to the movie type under the context attribute value, wherein the formula is as follows:
wherein
Indicating how well user x scores movie type t,
representing the frequency of scoring of movie type t by user x.
Wherein R is Max Is the maximum rating of the score.
Wherein the content of the first and second substances,
indicates the number of times user x scored the movie type under the context attribute value, indicates
The number of times user x scores all movie types under the context attribute value.
(2) Converting the contextual user-typicality of the user for each movie type into a contextual user-typicality vector; the expression of the context user-typicality vector is:
a context-typicality matrix is then constructed from the context-user-typicality vectors.
As shown in fig. 2, the expression of the context typicality matrix is:
wherein m represents the number of users, n represents the number of user groups,
representing user u x Context attribute value C h,k The user representative vector of (a).
Searching adjacent cosine similarity between user data in the context typical matrix and target user data, and judging the cosine similarity, wherein the adjacent cosine similarity comprises the cosine similarity between the user data in the context typical matrix and the target user data, and when the user data with the largest cosine similarity is adjacent to the context typical matrix; the cosine similarity formula is:
wherein | · | | represents the modulus of the vector,
a context representative vector of the representation.
After the similarity between the target user and other users is calculated, a prediction score is obtained according to the context typicality of the other users to the movie and the score of the project in the current context. The formula is as follows:
wherein d is y,t Represents the movie type t, C to which movie y belongs h,k A k-th attribute value representing an h-th context attribute,
set of scores, R, representing types of movies in the same single context attribute x,y Represents the rating of movie y by user x, n represents the nth rating,
value C representing the context attribute of a neighbor user i h,k The history of movie y is scored.
For target user u in multi-context c x The score prediction for the ith item is calculated as:
wherein x represents the x-th user, C represents the context attribute set where the user is, y represents the y-th movie, C h,k K-th attribute value, d, representing h-th context attribute y,t Indicates the movie type, G (d), to which movie y belongs y,t |C h,k ) Represents a context attribute value C h,k For type d of film y,t The degree of importance of (a) is,
representing the single context prediction score of user x for movie y.
The formula for calculating the importance of the context attribute to each movie type by using the knowledge granularity formula is as follows:
wherein, d y,t Represents the movie type t, C to which movie y belongs h,k A k-th attribute value representing an h-th context attribute,
set of scores, R, representing types of movies in the same single context attribute x,y Representing the rating of movie y by user x and n representing the nth rating.
The formula for the multi-context prediction score is:
where x represents the x-th user, C represents the set of context attributes where the user is located, y represents the y-th movie, C h,k K-th attribute value, d, representing h-th context attribute y,t Represents the movie type t, G (d) to which movie y belongs y,t |C h,k ) Represents a context attribute value C h,k For type d of film y,t The importance of (a) to (b),
representing the single context prediction score for movie y for user x.
The above-mentioned embodiments, which further illustrate the objects, technical solutions and advantages of the present invention, should be understood that the above-mentioned embodiments are only preferred embodiments of the present invention, and should not be construed as limiting the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A method for recommending movies based on contextual representativeness, the method comprising:
s1: acquiring user data, and collecting the user data to obtain a data set; calculating contextual typicality of the data from information in the data; the user data comprises movie information, rating information and context attribute values; the formula for calculating the context typicality of the data is:
wherein the content of the first and second substances,
indicates the degree of rating, C, of user x for movie type t h,k A kth attribute value representing an h context attribute, r represents a degree of scoring,
representing the scoring frequency of the user x on the movie type t, and f represents the scoring frequency;
s2: constructing a context typicality matrix according to the user data and the context typicality of each user to different movie types;
s3: searching neighbors in the context typicality matrix, and calculating scores of neighbor users according to context attribute values of the neighbors;
s4: predicting the score of the target user on the movie under the context attribute value by adopting the scores of the neighbor users;
s5: calculating the importance of the context attribute to each movie type by adopting knowledge granularity according to the grading information of the user data;
s6: calculating a multi-context prediction score according to the importance and the score of the target user on the movie under the context attribute value;
s7: the movie is recommended according to the multi-context predictive score.
2. The method of claim 1, wherein the step of constructing the context-user-typicality vector comprises:
step 1: calculating the user typicality of the user to the movie type under the context attribute value;
step 2: converting the contextual user-typicality of the user for each movie type into a contextual user-typicality vector; the expression of the contextual user-typicality vector is:
and step 3: constructing all context user representativeness vectors into a context representativeness matrix;
wherein the content of the first and second substances,
representing a contextual user-typicality vector, C h,k A kth attribute value representing an h context attribute,
representing user x in context attribute value C h,k User typicality for the nth movie type.
3. The method of claim 2, wherein the context representativeness matrix is:
wherein m represents the number of users, n represents the number of user groups,
value C of a contextual attribute representing user x h,k The user representative vector of (a).
4. The method of claim 1, wherein the step of searching for neighbors in the context canonical matrix comprises calculating cosine similarity between the user data in the context canonical matrix and the target user data, determining the magnitude of all cosine similarities, and when the user data with the largest cosine similarity is a neighbor of the context canonical matrix; the cosine similarity formula is:
where | l | · | | represents the modulus of the vector,
a context-representative vector of representations.
5. The method of claim 1, wherein the formula for the predictive scoring of a single context of different types of movies by neighboring users is as follows:
wherein
Represents a context attribute value C h,k User-typicality of next user i to type t to which movie y belongs,
indicating a neighbor user i context attribute value C h,k The history of movie y is scored.
6. A method as claimed in claim 1, wherein the formula for calculating the importance of the context attribute to each movie type by using the knowledge granularity is as follows:
wherein d is y,t Represents the movie type t, C to which movie y belongs h,k K-th attribute representing h-th context attributeProperty value, [ r ]] dy,t Set of scores, R, representing types of movies in the same single context attribute x,y Representing the rating of movie y by user x and n representing the nth rating.
7. The method of claim 1, wherein the multi-context prediction score is formulated as:
where x represents the x-th user, C represents the set of context attributes where the user is located, y represents the y-th movie, C h,k K-th attribute value, d, representing h-th context attribute y,t Represents the movie type t, G (d) to which movie y belongs y,t |C h,k ) Represents a context attribute value C h,k For type d of film y,t The degree of importance of (a) is,
representing the single context prediction score of user x for movie y.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110083203.1A CN112784171B (en) | 2021-01-21 | 2021-01-21 | Movie recommendation method based on context typicality |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110083203.1A CN112784171B (en) | 2021-01-21 | 2021-01-21 | Movie recommendation method based on context typicality |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112784171A CN112784171A (en) | 2021-05-11 |
| CN112784171B true CN112784171B (en) | 2022-11-04 |
Family
ID=75758271
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110083203.1A Active CN112784171B (en) | 2021-01-21 | 2021-01-21 | Movie recommendation method based on context typicality |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112784171B (en) |
Family Cites Families (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6655963B1 (en) * | 2000-07-31 | 2003-12-02 | Microsoft Corporation | Methods and apparatus for predicting and selectively collecting preferences based on personality diagnosis |
| US20030126606A1 (en) * | 2001-12-27 | 2003-07-03 | Koninklijke Philips Esectronics N.V. | Hierarchical decision fusion of recommender scores |
| CN102915307B (en) * | 2011-08-02 | 2018-01-02 | 腾讯科技（深圳）有限公司 | The method, apparatus and information processing system of a kind of recommendation of personalized information |
| CN102508846A (en) * | 2011-09-26 | 2012-06-20 | 深圳中兴网信科技有限公司 | Internet based method and system for recommending media courseware |
| CN103593410B (en) * | 2013-10-22 | 2017-04-12 | 上海交通大学 | System for search recommendation by means of replacing conceptual terms |
| CN104008193B (en) * | 2014-06-12 | 2017-04-05 | 安徽融数信息科技有限责任公司 | A kind of information recommendation method based on group of typical user discovery technique |
| CN104881499A (en) * | 2015-06-19 | 2015-09-02 | 中国石油大学(华东) | Collaborative filtering recommendation method based on attribute rating scaling |
| CN104899321A (en) * | 2015-06-19 | 2015-09-09 | 中国石油大学(华东) | Collaborative filtering recommendation method based on item attribute score mean value |
| CN105868334B (en) * | 2016-03-28 | 2020-10-30 | 云南财经大学 | Feature incremental type-based personalized movie recommendation method and system |
| CN106295675B (en) * | 2016-07-25 | 2019-09-27 | 南京邮电大学 | A kind of Activity recognition method based on smart phone of high accuracy |
| CN106326483A (en) * | 2016-08-31 | 2017-01-11 | 华南理工大学 | Collaborative recommendation method with user context information aggregation |
| CN106484876A (en) * | 2016-10-13 | 2017-03-08 | 中山大学 | A kind of based on typical degree and the collaborative filtering recommending method of trust network |
| US10430214B2 (en) * | 2016-12-30 | 2019-10-01 | Google Llc | Dynamically generating custom application onboarding tutorials |
| CN106802956B (en) * | 2017-01-19 | 2020-06-05 | 山东大学 | Movie recommendation method based on weighted heterogeneous information network |
| CN106991122B (en) * | 2017-02-27 | 2021-02-02 | 四川大学 | Movie recommendation method based on particle swarm optimization |
| CN107909498B (en) * | 2017-10-26 | 2020-07-28 | 厦门理工学院 | Recommendation method based on area below maximized receiver operation characteristic curve |
| CN108256093B (en) * | 2018-01-29 | 2020-06-19 | 华南理工大学 | Collaborative filtering recommendation algorithm based on multiple interests and interest changes of users |
| CN108804683B (en) * | 2018-06-13 | 2021-11-23 | 重庆理工大学 | Movie recommendation method combining matrix decomposition and collaborative filtering algorithm |
| CN109241412B (en) * | 2018-08-17 | 2020-12-11 | 深圳先进技术研究院 | Recommendation method and system based on network representation learning and electronic equipment |
| CN111723302A (en) * | 2020-06-09 | 2020-09-29 | 云南大学 | Recommendation method based on collaborative dual-model deep representation learning |
-
2021
- 2021-01-21 CN CN202110083203.1A patent/CN112784171B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112784171A (en) | 2021-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107220365B (en) | Accurate recommendation system and method based on collaborative filtering and association rule parallel processing | |
| JP2010504593A (en) | Extracting dominant colors from an image using a classification technique | |
| CN110032679B (en) | Dynamic news recommendation method based on hierarchical attention network | |
| CN111460961B (en) | Static video abstraction method for CDVS-based similarity graph clustering | |
| CN109033132B (en) | Method and device for calculating text and subject correlation by using knowledge graph | |
| JP2011198364A (en) | Method of adding label to medium document and system using the same | |
| CN111460251A (en) | Data content personalized push cold start method, device, equipment and storage medium | |
| CN109816015B (en) | Recommendation method and system based on material data | |
| CN109977299B (en) | Recommendation algorithm fusing project popularity and expert coefficient | |
| CN108874916A (en) | A kind of stacked combination collaborative filtering recommending method | |
| CN111859004A (en) | Retrieval image acquisition method, device, equipment and readable storage medium | |
| CN109086830B (en) | Typical correlation analysis near-duplicate video detection method based on sample punishment | |
| CN111177559A (en) | Text travel service recommendation method and device, electronic equipment and storage medium | |
| CN111858972A (en) | Movie recommendation method based on family knowledge graph | |
| CN113434778B (en) | Recommendation method based on regularization framework and attention mechanism | |
| CN110928986A (en) | Legal evidence sorting and recommending method, device, equipment and storage medium | |
| CN113010705B (en) | Label prediction method, device, equipment and storage medium | |
| JP6368677B2 (en) | Mapping learning method, information compression method, apparatus, and program | |
| CN116541607B (en) | Intelligent recommendation method based on commodity retrieval data analysis | |
| TW201243627A (en) | Multi-label text categorization based on fuzzy similarity and k nearest neighbors | |
| CN112784171B (en) | Movie recommendation method based on context typicality | |
| CN111259223B (en) | News recommendation and text classification method based on emotion analysis model | |
| CN115687676B (en) | Information retrieval method, terminal and computer-readable storage medium | |
| Kesidis et al. | Efficient cut-off threshold estimation for word spotting applications | |
| CN115712780A (en) | Information pushing method and device based on cloud computing and big data |
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 |