CN104239390A - Audio recommending method on basis of improved collaborative filtering algorithm - Google Patents
Audio recommending method on basis of improved collaborative filtering algorithm Download PDFInfo
- Publication number
- CN104239390A CN104239390A CN201410257826.6A CN201410257826A CN104239390A CN 104239390 A CN104239390 A CN 104239390A CN 201410257826 A CN201410257826 A CN 201410257826A CN 104239390 A CN104239390 A CN 104239390A
- Authority
- CN
- China
- Prior art keywords
- project
- scoring item
- scoring
- highest
- targeted customer
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/63—Querying
- G06F16/635—Filtering based on additional data, e.g. user or group profiles
- G06F16/637—Administration of user profiles, e.g. generation, initialization, adaptation or distribution
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an audio recommending method on the basis of an improved collaborative filtering algorithm. The audio recommending method specifically includes steps of S1, looking up N neighboring users with highest similarity to a target user; S2, selecting unscored projects c, and calculating project relevancy Rc between the unscored projects c and the highest-score project of the target user within an effective time; S3, calculating estimated scores Pc of the unscored projects c of the target user according to the scores of the unscored projects c of the neighboring users and the project relevancy Rc; S4, repeating the step S2 and S3 until all the unscored projects are calculated to obtain the estimated scores; S5, sequencing the unscored projects according to the estimated scores, and recommending M unscored projects with highest estimated scores to the users. Project relevancy and user similarity are combined, the projects with low relevancy are eliminated, recommending accuracy is improved, and the audio recommending method is applicable to network music playing software or network music players.
Description
Technical field
The present invention relates to user requirements analysis and recommend field, especially relate to a kind of audio frequency recommend method based on modified collaborative filtering for network audio play system.
Background technology
Flourish along with Internet technology, we have been in the information big bang epoch, from the information of numerous and complicated, how to find the information useful to oneself to become extremely important and urgent, from split catalog development search engine, had significant lifting searching in the efficiency of knowledge, you can according to the contents of any one keyword search and this keyword match.But this has individual prerequisite, that is exactly that you know what to be looked for.Facts have proved, often many times we and do not know information definition one key word accurately how to want to oneself.An effective method of head it off utilizes commending system.
The principle of commending system is, machine is by mutual automatic learning that is each and people, and utilize some mathematical models that interbehavior is quantified as indices, thus set up a virtual character, what learn is more, the definition of character is also more accurate, presses the next possible behavior of current space-time environment conjecture user, provide rational recommendation finally by search engine.
At present, commending system is found everywhere, external Amazon, Netflix, domestic Taobao, Jingdone district, bean cotyledon FM, dried shrimp etc., all to a certain extent to we providing a lot of suggestion, simplify that we search information, do shopping, listen to the music, the complexity of the information of searching such as to see a film.Common commending system algorithm mainly contains: content-based recommendation, collaborative filtering recommending, based on correlation rule recommend, based on effectiveness recommend, knowledge based recommend and combined recommendation.
The relative merits of each proposed algorithm are as following table:
State Intellectual Property Office of the People's Republic of China disclosed the patent documentation of application publication number CN103559622A for 02 month on the 05th in 2014, title is the collaborative filtering recommending method of feature based, it comprises the following steps: step one, feature according to article, original article-user's scoring matrix is projected in different article characteristics, obtains feature-user's scoring matrix of multiple polymerization; Step 2, add up its variance of giving a mark in each feature for each user, and portray the taste degree of user to this feature with this variance yields, variance is larger, shows that the value of user to this feature has and is partial to more by force; Step 3, based on each feature-user's scoring matrix, prediction user to the marking value of certain new article; Step 4, utilize the variance that each feature of step 2 is given a mark, the marking predicted value that step 3 calculates is weighted on average, obtains the final marking predicted value of user to these article; Step 5, based on final marking predicted value, carry out article recommendation.This scheme does not consider given a mark article and the correlativity of not giving a mark between article, recommends precision not high enough, easily occurs the situation of by mistake recommending.
Summary of the invention
The present invention mainly solves can not consider existing for prior art and to have given a mark correlativity between project and non-marking project, the technical matters of recommending precision high not, provide a kind of introducing given a mark the degree of correlation between project and non-marking project, get rid of the low project of the degree of correlation, effectively improve the audio frequency recommend method based on modified collaborative filtering recommending precision.
The present invention is directed to that above-mentioned technical matters mainly solved by following technical proposals: a kind of audio frequency recommend method based on modified collaborative filtering, comprises following flow process:
S1, search the N number of neighbor user the highest with targeted customer's similarity;
S2, a selection non-scoring item c, calculate the project dependency Rc between non-scoring item c and the highest scoring item of targeted customer within effective time; Non-scoring item is defined as the project that targeted customer did not mark;
S3, according to neighbor user, targeted customer is calculated to the expectation score value Pc of non-scoring item c to the score value of non-scoring item c and project dependency Rc;
S4, repetition step S2 and S3, until all non-scoring items are all calculated expectation score value;
S5, by all non-scoring items on the estimation score value sort, recommend to estimate the highest M of a score value non-scoring item to user.
As preferably, in described step S1, search and be specially for N number of neighbor user that similarity is the highest with target:
S11, a selection neighbor user, search all items that targeted customer marked with this neighbor user; If the number of entry found is n;
S12, calculate similarity r between targeted customer and this neighbor user, computing formula is
In formula, x
ifor targeted customer is to the scoring of i-th project,
for targeted customer is to the average score value of the n a found project, y
ineighbor user is to the scoring of i-th project for this reason,
neighbor user is to the average score value of the n a found project for this reason;
S13, repetition step S11 and S12, until all neighbor users all calculate complete;
S14, to all neighbor users according to sequencing of similarity, select the highest N number of neighbor user of similarity.
As preferably, in step S2, the project dependency Rc calculated between non-scoring item c and the highest scoring item of targeted customer within effective time is specially:
S21, read the value of non-each label of scoring item c;
The value of each label of the highest scoring item described in S22, reading;
Correlativity Rc described in S23, calculating between the highest scoring item and this non-scoring item c, computing formula is as follows:
In formula, m is total number of labels, T
ifor the value of i-th label of the highest scoring item, C
ithe value of i-th label of non-scoring item c for this reason; ∩ is same or computing;
The value of each label is 1 or 0.
As preferably, in described step S3, according to neighbor user, the expectation score value Pc of targeted customer to non-scoring item c is calculated to the score value of non-scoring item and project dependency Rc and is specially and is calculated as follows:
In formula,
for targeted customer is to the average score value self having commented project,
be u neighbor user to the average score value self having commented project, r
ufor the similarity between targeted customer and u neighbor user,
be the score value of u neighbor user to non-scoring item c.
As preferably, it is characterized in that, described N is not less than 5.
As preferably, in described step S2, primary election effective time value is 15 days to 30 days, if user does not carry out scoring operation within effective time, then the project that in 10 scorings selecting user nearest, the highest scoring is corresponding is as the highest scoring item.
Project in this programme is audio file (song), each project has oneself label, for representing the characteristic that this project possesses, label can comprise song, opera, popular, classic, network, cheerful and light-hearted, graceful etc., and keeper is it and sets label after set up item.Item label is more, user self scoring number of times more by scoring number of times recommends precision also can be higher more at most.
New user needs first to after listened at least two project marking, and system could recommend suitable non-scoring item for it, and the degree of accuracy that the project commented is recommended more at most also can be higher.
When two projects have identical label, assert that their degree of correlation is 1 (the highest), when two projects do not have any identical label, assert that their degree of correlation is 0 (minimum).
This programme can be applied to online music playout software or online music playback equipment.Can also introduce in practical application that project is collected again, the number of times of subscription etc. to revise expectation score value, thus improves and recommend precision.
The substantial effect that the present invention brings is, project dependency and user's similarity is combined, eliminates the project that correlativity is low, improve and recommend precision, enable user obtain better experience.
Accompanying drawing explanation
Fig. 1 is a kind of process flow diagram of the present invention.
Embodiment
Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.
Embodiment: a kind of audio frequency recommend method based on modified collaborative filtering of the present embodiment, as shown in Figure 1, comprises following flow process:
S1, search the N number of neighbor user the highest with targeted customer's similarity; The general value about 10 of N;
S2, a selection non-scoring item c, calculate the project dependency Rc between non-scoring item c and the highest scoring item of targeted customer within effective time; Non-scoring item is defined as the project that targeted customer did not mark; Primary election effective time value is 15 days to 30 days, if user does not carry out scoring operation within effective time, then the project that in 10 scorings selecting user nearest, the highest scoring is corresponding is as the highest scoring item;
S3, according to neighbor user, targeted customer is calculated to the expectation score value Pc of non-scoring item c to the score value of non-scoring item c and project dependency Rc;
S4, repetition step S2 and S3, until all non-scoring items are all calculated expectation score value;
S5, by all non-scoring items on the estimation score value sort, recommend to estimate the highest M of a score value non-scoring item to user.The occurrence of M can be determined as required, is generally 5.10.
In step S1, search and be specially for N number of neighbor user that similarity is the highest with target:
S11, a selection neighbor user, search all items that targeted customer marked with this neighbor user; If the number of entry found is n;
S12, calculate similarity r between targeted customer and this neighbor user, computing formula is
In formula, x
ifor targeted customer is to the scoring of i-th project,
for targeted customer is to the average score value of the n a found project, y
ineighbor user is to the scoring of i-th project for this reason,
neighbor user is to the average score value of the n a found project for this reason;
S13, repetition step S11 and S12, until all neighbor users all calculate complete;
S14, to all neighbor users according to sequencing of similarity, select the highest N number of neighbor user of similarity.
In step S2, the project dependency Rc calculated between non-scoring item c and the highest scoring item of targeted customer within effective time is specially:
S21, read the value of non-each label of scoring item c;
The value of each label of the highest scoring item described in S22, reading;
Correlativity Rc described in S23, calculating between the highest scoring item and this non-scoring item c, computing formula is as follows:
In formula, m is total number of labels, T
ifor the value of i-th label of the highest scoring item, C
ithe value of i-th label of non-scoring item c for this reason; ∩ is same or computing;
The value of each label is 1 or 0.
As being as following table to the non-scoring item of certain targeted customer:
Can obtain by calculating, the correlativity respectively between non-scoring item and the highest scoring item is:
Song A:0.571
Song B:0.857
Song C:0.286
Song D:0
So, to the recommendation order of targeted customer be: song B, song A, song C, song D.
In order to improve the degree of accuracy of recommendation further, part labels can be got rid of label as first-selection, if " song " and " opera " is two fields differed greatly, for user, if these two labels are not identical, then can directly delete from recommendation list.
Time song label is set up item database, directly set.
In step S3, according to neighbor user, the expectation score value Pc of targeted customer to non-scoring item c is calculated to the score value of non-scoring item and project dependency Rc and is specially and is calculated as follows:
In formula,
for targeted customer is to the average score value self having commented project,
be u neighbor user to the average score value self having commented project, r
ufor the similarity between targeted customer and u neighbor user,
be the score value of u neighbor user to non-scoring item c.
Revise expectation score value by project dependency, higher recommendation degree of accuracy can be obtained.
Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.
Although more employ the terms such as similarity, correlativity, expectation score value, label herein, do not get rid of the possibility using other term.These terms are used to be only used to describe and explain essence of the present invention more easily; The restriction that they are construed to any one additional is all contrary with spirit of the present invention.
Claims (6)
1. based on an audio frequency recommend method for modified collaborative filtering, it is characterized in that, comprise following flow process:
S1, search the N number of neighbor user the highest with targeted customer's similarity;
S2, a selection non-scoring item c, calculate the project dependency Rc between non-scoring item c and the highest scoring item of targeted customer within effective time; Non-scoring item is defined as the project that targeted customer did not mark;
S3, according to neighbor user, targeted customer is calculated to the expectation score value Pc of non-scoring item c to the score value of non-scoring item c and project dependency Rc;
S4, repetition step S2 and S3, until all non-scoring items are all calculated expectation score value;
S5, by all non-scoring items on the estimation score value sort, recommend to estimate the highest M of a score value non-scoring item to user.
2. a kind of audio frequency recommend method based on modified collaborative filtering according to claim 1, is characterized in that, in described step S1, searches and is specially for N number of neighbor user that similarity is the highest with target:
S11, a selection neighbor user, search all items that targeted customer marked with this neighbor user; If the number of entry found is n;
S12, calculate similarity r between targeted customer and this neighbor user, computing formula is
In formula, x
ifor targeted customer is to the scoring of i-th project,
for targeted customer is to the average score value of the n a found project, y
ineighbor user is to the scoring of i-th project for this reason,
neighbor user is to the average score value of the n a found project for this reason;
S13, repetition step S11 and S12, until all neighbor users all calculate complete;
S14, to all neighbor users according to sequencing of similarity, select the highest N number of neighbor user of similarity.
3. a kind of audio frequency recommend method based on modified collaborative filtering according to claim 2, is characterized in that, in step S2, the project dependency Rc calculated between non-scoring item c and the highest scoring item of targeted customer within effective time is specially:
S21, read the value of non-each label of scoring item c;
The value of each label of the highest scoring item described in S22, reading;
Correlativity Rc described in S23, calculating between the highest scoring item and this non-scoring item c, computing formula is as follows:
In formula, m is total number of labels, T
ifor the value of i-th label of the highest scoring item, C
ithe value of i-th label of non-scoring item c for this reason; ∩ is same or computing;
The value of each label is 1 or 0.
4. a kind of audio frequency recommend method based on modified collaborative filtering according to claim 3, it is characterized in that, in described step S3, according to neighbor user, the expectation score value Pc of targeted customer to non-scoring item c is calculated to the score value of non-scoring item and project dependency Rc and is specially and is calculated as follows:
In formula,
for targeted customer is to the average score value self having commented project,
be u neighbor user to the average score value self having commented project, r
ufor the similarity between targeted customer and u neighbor user,
be the score value of u neighbor user to non-scoring item c.
5. a kind of audio frequency recommend method based on modified collaborative filtering according to claim 1 or 2 or 3 or 4, it is characterized in that, described N is not less than 5.
6. a kind of audio frequency recommend method based on modified collaborative filtering according to claim 5, it is characterized in that, in described step S2, primary election effective time value is 15 days to 30 days, if user does not carry out scoring operation within effective time, then the project that in 10 scorings selecting user nearest, the highest scoring is corresponding is as the highest scoring item.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410257826.6A CN104239390B (en) | 2014-06-11 | 2014-06-11 | A kind of audio based on modified collaborative filtering recommends method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410257826.6A CN104239390B (en) | 2014-06-11 | 2014-06-11 | A kind of audio based on modified collaborative filtering recommends method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104239390A true CN104239390A (en) | 2014-12-24 |
| CN104239390B CN104239390B (en) | 2017-12-29 |
Family
ID=52227459
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410257826.6A Active CN104239390B (en) | 2014-06-11 | 2014-06-11 | A kind of audio based on modified collaborative filtering recommends method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104239390B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105721279A (en) * | 2016-01-15 | 2016-06-29 | 中国联合网络通信有限公司广东省分公司 | Relationship circle excavation method and system of telecommunication network users |
| CN106156333A (en) * | 2016-07-06 | 2016-11-23 | 合肥工业大学 | A kind of improvement list class collaborative filtering method of mosaic society information |
| CN108052598A (en) * | 2017-12-12 | 2018-05-18 | 何旭连 | A kind of intelligent music commending system based on label |
| CN108205533A (en) * | 2016-12-16 | 2018-06-26 | 北京酷我科技有限公司 | A kind of song recommendations method and system based on user collaborative |
| WO2019085120A1 (en) * | 2017-11-01 | 2019-05-09 | 平安科技（深圳）有限公司 | Collaborative filtering recommendation method, electronic device, and computer readable storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101751448A (en) * | 2009-07-22 | 2010-06-23 | 中国科学院自动化研究所 | Commendation method of personalized resource information based on scene information |
| CN102208086A (en) * | 2010-03-31 | 2011-10-05 | 北京邮电大学 | Field-oriented personalized intelligent recommendation system and implementation method |
| CN102495864A (en) * | 2011-11-25 | 2012-06-13 | 清华大学 | Collaborative filtering recommending method and system based on grading |
| CN102654860A (en) * | 2011-03-01 | 2012-09-05 | 北京彩云在线技术开发有限公司 | Personalized music recommendation method and system |
| CN102982466A (en) * | 2012-07-17 | 2013-03-20 | 华东师范大学 | Graded forecasting method based on user liveness |
| CN103383702A (en) * | 2013-07-17 | 2013-11-06 | 中国科学院深圳先进技术研究院 | Method and system for recommending personalized news based on ranking of votes of users |
| CN103793476A (en) * | 2014-01-08 | 2014-05-14 | 西安电子科技大学 | Network community based collaborative filtering recommendation method |
-
2014
- 2014-06-11 CN CN201410257826.6A patent/CN104239390B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101751448A (en) * | 2009-07-22 | 2010-06-23 | 中国科学院自动化研究所 | Commendation method of personalized resource information based on scene information |
| CN102208086A (en) * | 2010-03-31 | 2011-10-05 | 北京邮电大学 | Field-oriented personalized intelligent recommendation system and implementation method |
| CN102654860A (en) * | 2011-03-01 | 2012-09-05 | 北京彩云在线技术开发有限公司 | Personalized music recommendation method and system |
| CN102495864A (en) * | 2011-11-25 | 2012-06-13 | 清华大学 | Collaborative filtering recommending method and system based on grading |
| CN102982466A (en) * | 2012-07-17 | 2013-03-20 | 华东师范大学 | Graded forecasting method based on user liveness |
| CN103383702A (en) * | 2013-07-17 | 2013-11-06 | 中国科学院深圳先进技术研究院 | Method and system for recommending personalized news based on ranking of votes of users |
| CN103793476A (en) * | 2014-01-08 | 2014-05-14 | 西安电子科技大学 | Network community based collaborative filtering recommendation method |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105721279A (en) * | 2016-01-15 | 2016-06-29 | 中国联合网络通信有限公司广东省分公司 | Relationship circle excavation method and system of telecommunication network users |
| CN106156333A (en) * | 2016-07-06 | 2016-11-23 | 合肥工业大学 | A kind of improvement list class collaborative filtering method of mosaic society information |
| CN106156333B (en) * | 2016-07-06 | 2017-09-22 | 合肥工业大学 | A kind of improvement list class collaborative filtering method of mosaic society's information |
| CN108205533A (en) * | 2016-12-16 | 2018-06-26 | 北京酷我科技有限公司 | A kind of song recommendations method and system based on user collaborative |
| WO2019085120A1 (en) * | 2017-11-01 | 2019-05-09 | 平安科技（深圳）有限公司 | Collaborative filtering recommendation method, electronic device, and computer readable storage medium |
| CN108052598A (en) * | 2017-12-12 | 2018-05-18 | 何旭连 | A kind of intelligent music commending system based on label |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104239390B (en) | 2017-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102349072B (en) | Identifying query aspects | |
| CN104428767B (en) | For identifying the mthods, systems and devices of related entities | |
| Ali et al. | Movie recommendation system using genome tags and content-based filtering | |
| CN103116657B (en) | A kind of individuation search method of network teaching resource | |
| US8478779B2 (en) | Disambiguating a search query based on a difference between composite domain-confidence factors | |
| CN103136257B (en) | Information providing method and device thereof | |
| CN104239390A (en) | Audio recommending method on basis of improved collaborative filtering algorithm | |
| US20170154116A1 (en) | Method and system for recommending contents based on social network | |
| CN103718178B (en) | Utilize the feature extracted from structured document to improve search relevance | |
| CN104641371B (en) | Based on the object retrieval of context in social networking system | |
| CN105787767A (en) | Method and system for obtaining advertisement click-through rate pre-estimation model | |
| CN104751354A (en) | Advertisement cluster screening method | |
| US20180139296A1 (en) | Method of producing browsing attributes of users, and non-transitory computer-readable storage medium | |
| US9390166B2 (en) | Specific online resource identification and extraction | |
| Moya et al. | Integrating web feed opinions into a corporate data warehouse | |
| Ortiz-Cordova et al. | External to internal search: Associating searching on search engines with searching on sites | |
| Xu | Computer-aided design of personalized recommendation in teaching system | |
| Lin | A case-based reasoning system for interior design using a new cosine similarity retrieval algorithm | |
| Magnani et al. | Neural product retrieval at walmart. com | |
| US20060276996A1 (en) | Fast tracking system and method for generalized LARS/LASSO | |
| CN108984752B (en) | Intelligent recommendation method for professional books in library | |
| Pang et al. | Query expansion and query fuzzy with large-scale click-through data for microblog retrieval | |
| Liang et al. | A hybrid recommender systems based on weighted tags | |
| Li et al. | Long-form information retrieval for enterprise matchmaking | |
| Kim et al. | Evaluating an associative browsing model for personal information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 310052, Zhejiang, Hangzhou province Binjiang District Long River Street No. 399, No. 3, four, five floor Applicant after: Hangzhou grandwell Polytron Technologies Inc Address before: 310052, No. 3, building 399, building four, Qiu Yi Road, Binjiang District Changhe Road, Hangzhou, Zhejiang Applicant before: Hangzhou Linker Digital Technology Co., Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |