CN107066476A - A kind of real-time recommendation method based on article similarity - Google Patents
A kind of real-time recommendation method based on article similarity Download PDFInfo
- Publication number
- CN107066476A CN107066476A CN201611144613.8A CN201611144613A CN107066476A CN 107066476 A CN107066476 A CN 107066476A CN 201611144613 A CN201611144613 A CN 201611144613A CN 107066476 A CN107066476 A CN 107066476A
- Authority
- CN
- China
- Prior art keywords
- user
- article
- preference
- real
- matrix
- 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
- G06F16/90—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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of real-time recommendation method based on article similarity, the step of including pretreatment historical use data, the step of calculating T+1 similar matrixes, the step of collecting newest user data, the step of calculating user preference matrix, the step of recommendation list is calculated according to similar matrix and preference matrix.The present invention, which can identify the user for logging in and being not logged in user, to be unified, and propose rational preference value calculating logic, add time factor, take into account two weight factors of number of clicks simultaneously, user is reasonably operated the hobby value for being converted into quantizing to product behavior, the true hobby of user is approached to the full extent, so that the calculating of user preference value is more accurate, what is more important is calculated according to the user behavior data obtained in real time, so as to greatly improve recommendation efficiency and counting accuracy.
Description
Technical field
The invention belongs to technical field of data processing, and in particular to a kind of real-time recommendation method based on item-based.
Background technology
In recent years, with the rapid development of Internet, network has been increasingly becoming people's free choice of goods, inquiring information of goods
The first choice of data.Type of merchandize is various, information refinement, website its structure while more and more selections are provided the user
Also become more complicated, in face of countless merchandise news, user is often got lost in substantial amounts of merchandise news space, nothing
Method smoothly finds the commodity of oneself needs.
To solve this problem, personalized recommendation method arises at the historic moment.Collaborative filtering of the conventional recommendation method based on article
Technology, be previously according to all users history preference data calculate article between similitude, then with user's history
The topN articles that the article of access is similar recommend user.Its concretism is：First pretreatment user access record, according to
Family article preference value calculates article similar matrix, recommends the similar article of its preference article to user.
But it is unsatisfactory using the effect of conventional recommendation method, not only recommend the degree of accuracy not high, and also operation efficiency is low.
To find out its cause, being because conventional method has following several defects：It is primarily due to user preference value and sets relatively simple, causes tradition
User preference matrix can not represent user preferences well, can persistently be amplified which results in the deviation subsequently calculated.Its
It is required for calculating full dose user every time during secondary computing, this has just computed repeatedly user's recommendation row that access behavior does not change actually
Table, causes consumption calculations resource more.In addition, being currently based on collaborative filtering is applied substantially to T+1 recommendations, it is not extended to
Real-time recommendation so that recommendation list renewal speed is too slow, it is difficult to meet actual demand.
The content of the invention
To solve the above problems, the invention discloses a kind of real-time recommendation method based on article similarity, will can step on
Land is unified with the user's mark for being not logged in user, and defines more accurate user preference value function, even more important
It is to be calculated according to the user behavior data obtained in real time, so as to greatly improve recommendation efficiency and counting accuracy.
In order to achieve the above object, the present invention provides following technical scheme：
A kind of real-time recommendation method based on article similarity, comprises the following steps：
Step A, collects recent user behavior data, behavioral data is pre-processed, when user is in Entered state, used
UserId Murmur cryptographic Hash is identified as user；User uses visitor_trace Murmur under non-Entered state
Cryptographic Hash is identified as user, and user's mark is converted into unified signless long data value；
Step B, calculates each user to browsing the preference values of commodity using preference function model, generates basic user article preference
Value matrix, calculates article similar matrix,
The preference function model is as follows：
score = i +
WhereiniFor initial value, xjRepresent the time difference of current time and user's access time；
Step C, filters out the newest behavioral data of current visitor, is processed into real-time preference matrix, conjugate product similar matrix, meter
User's recommendation list is calculated, including：
Step C-1, screens all behavioral datas of newest Guest User, data is pre-processed, when user is logging in shape
State, the Murmur cryptographic Hash using userId is identified as user；User uses visitor_trace under non-Entered state
Murmur cryptographic Hash identified as user, by user mark be uniformly converted into long numerical value；
Step C-2, using each user of preference function model conversion to browsing the preference values of commodity, by all users to some thing
The preference value of product calculates the similarity between article as a vector, generates active user article preference matrix；
Step C-3, the article similar matrix preserved using the step B obtained same day, using matrix multiplication, passes through following formula meter
Calculate：
Article similar matrix * active user article preference matrixs
Step C-4, the result obtained based on step C-3 is calculated median sum of each user under same article, obtains each use
Family filters out user and browsed with after the commodity that placed an order, sort, taken before coming by preference value size to the predicted value of each product
Some commodity be recommendation list.
As an improvement, in order to avoid recommending in hot product, step C-4, adding similarity and being used as penalty factor, predicted value
Calculated and obtained by following formula：
Predicted value=Σ medians/Σ similarities.
Preferably, the step C is developed based on MapReduce.
Preferably, the step C screens all behavioral datas of Guest User in 5 minutes.
Preferably, the step C is developed based on SparkStreaming.
Preferably, time window is set to 5s, data flow update status is detected every 5s, once updating, is then collected more
Recommended after new data operation.
Preferably, the article similar matrix is by calling mahout-itembased interfaces to obtain.
Preferably, the article similar matrix is preserved using sparse matrix.
Preferably, when user only gives a mark in consumer articles preference value matrix to an article, then not calculating the use
The recommendation list at family.
Compared with prior art, the invention has the advantages that and beneficial effect：
1. merging two fields of useId and visitor_trace, and user's mark is scattering into unique signless long data value, it is real
Existing two fields merge unified, integrate log in and non-login user access data, it is ensured that login user and non-login user
Mark is consistent.
2. propose rational preference value calculating logic, add time factor, at the same take into account two weights of number of clicks because
User, is reasonably operated the hobby value for being converted into quantizing to product behavior, the true happiness of user is approached to the full extent by son
It is good so that the calculating of user preference value is more accurate.
3. filtering out the history and current behavior data of current visitor in real time, preference matrix is pre-processed into, with reference to by near
Phase full dose data calculate and preserved obtained article similar matrix, calculate user's recommendation list.Substantially increase recommendation method
Efficiency, the recommendation time can be narrowed down in second level complete.
Brief description of the drawings
Fig. 1 is the inventive method flow chart.
Fig. 2 is the real-time recommendation method flow chart developed based on SparkStreaming.
Embodiment
The technical scheme provided below with reference to specific embodiment the present invention is described in detail, it should be understood that following specific
Embodiment is only illustrative of the invention and is not intended to limit the scope of the invention.
In the conventional technology, different user is distinguished with userId, non-login user not can recognize that, can so missed a large amount of
The access behavioral data of non-login user, causes the missing of basic data, causes the reliability decrease of subsequent arithmetic.It is of the invention first
This problem is first solved, the present invention combines unique mark that userId and visitor_trace integrates every site visitor of generation
Know, when user is in Entered state, preferentially use userId Murmur cryptographic Hash to be identified as user；User is in non-Entered state
Under, then using visitor_trace（Hardware encoding）Murmur cryptographic Hash identified as user, specifically using Murmur breathe out
Uncommon function is corresponding character string（userId、visitor_trace）It is converted into long values.It is every in the present invention to be related to user
Do this processing in the place of mark.The unrecognizable problem of non-login user is so can solve the problem that, and completely remains all visits
Behavioral data is asked, reliability is added during article similarity and recommendation list is calculated.
The improvement of the present invention focuses on redefining user preference value, adds in preference function mathematical modeling
User, is reasonably operated the hobby for being converted into quantizing by two weight factors of angle of incidence factor and number of clicks to product behavior
Value.Preference function model definition is as follows：
Preference value is initialized first, specific user behavior has：The number such as browse, collect, cancel collection, place an order, cancel the order
Kind, based on this, preference value initialization is as shown in table 1 below：
Table 1
Then define time weighting model and number of clicks weight model：
It is believed that the user behavior nearer apart from current time, can more reflect user preferences；Conversely, being got over apart from current time
User behavior remote, is defined weaker to user preference.
Based on above thinking, defining time weighting specific formula for calculation isk * (x represents that current time is visited with user
The time difference of time is asked,=2.718281828,k =1, according to different usage scenarios, kIt can set different
Value, fits preference value of the user to article)
It is believed that user is more frequent to the click of some commodity, represent that user is stronger to the preference degree of the commodity.More than being based on
Thinking, defines number of clicks weight, and user often clicks on a commodity, i.e., add up to add the preference that this time is browsed in preference value
Value.
It is as follows that generalized time weight model and number of clicks weight model obtain real user preferences value：
score = i + （iFor user the last behavior corresponding initial value, x in table 1jRepresent to work as
Preceding time and the time difference of user's access time）
Basic, the real-time recommendation method based on article similarity that the present invention is provided is improved based on more than, as shown in figure 1, including
Following steps：
Step A, collects recent（This example choosing value 15 days）User behavior data（Including but not limited to visitor_trace, uid, are produced
Time, product type are clicked in product ID, concrete operations）, behavioral data is pre-processed, preprocessing process is mainly using existing
Usual manner in technology, unlike, in behavior process of data preprocessing, present invention fusion user's mark（uid、
visitor_trace）, it is scattering into unified signless long data value.
Step B, using each user of preference function model conversion to browsing the preference values of commodity, generates basic user article
Preference value matrix, calculates the similarity between article as a vector to the preference value of some article using user, calls
Mahout-itembased interfaces（Mahout-itembased interfaces are preferred embodiment, it would however also be possible to employ other modes calculate thing
Product similarity, such as calculates similarity based on article essential content）, calculating obtains article similar matrix, and generates T+1 initialization
Recommendation list.Based on the openness of article similar matrix, we preserve article similar matrix using sparse matrix, specific similar
Matrix storage format is：(itemA, itemB, similarity), with the similar degrees of data of Item Number 210054274 such as
Shown in table 2 below：
Table 2
Step C, is recommended method using the batch processing near real-time developed based on MapReduce or is developed based on SparkStreaming
Real-time recommendation method carry out Products Show.
Near real-time recommends method to be developed based on MapReduce, screens newest（In 5 minutes）All behaviors of Guest User
Data, specifically include following steps：
Step C-1, screens all behavioral datas of newest Guest User（Including but not limited to visit_trace, uid, product
Time, product type are clicked in ID, concrete operations）, data are pre-processed, user's mark is uniformly converted into long number
Value；This example achieves the behavioral data of user 10001, and he has browsed two articles on website（101,104）.
Step C-2, using each user of preference function model conversion to browsing the preference values of commodity, by all users to certain
The preference value of individual article calculates the similarity between article as a vector, generates active user article preference value matrix.
Based on the behavioral data obtained in step C-1, the active user article preference value matrix that this step is obtained is as shown in the table：
Table 3
If certain user only gives a mark in active user article preference value matrix to an article, then do not calculate his recommendation row
Table.
Step C-3, the article similar matrix preserved using the step B obtained same day, using matrix multiplication, passes through following formula
Calculate：
Article similar matrix * active user article preference matrixs
101,104 are learnt from the similar matrix preserved has similar relation as shown in the table to other articles：
Table 4
Calculated by article similar matrix * active user articles preference matrix and obtain user to being tied in the middle of this similar article
Really, that is, preference value * correspondence similarities are calculated, result are obtained as shown in the table：
Table 5
Step C-4, the result obtained based on step C-3 is calculated median sum of each user under same article, obtains each use
Predicted value of the family to each commodity.
For example, user 10001 is to the final prediction score value calculating process of article 102：Predicted value=0.805+1.722=
2.527。
Browsed after filtering user with after the commodity that placed an order, sorted by preference value size, take some business come above
Product are recommendation list, and this example takes preceding 36 commodity.Recommendation list is input into Hbase tables, is easy to distributed program to call and is pushed away
Recommend data.
As an improvement, in step C-4, in order to avoid recommending hot product, adding similar degrees of data as penalty factor, leading to
Cross following formula and calculate the predicted value after being optimized：
Predicted value=Σ medians/Σ similarities
For example, user 10001 is to the final prediction score value calculating process of article 102：Predicted value=(0.805+1.722）/
(0.23+0.41) = 3.948。
Consequently recommended merging predicts the outcome as shown in the table：
Table 6
Real-time recommendation method is developed based on SparkStreaming, and its key step is approximate near real-time mode, and difference is,
This method sets time window to be 5s, and data flow update status is detected every 5s, once updating, then the data for collecting renewal are gone forward side by side
Row is recommended.In addition, active user article preference matrix is stored in spark-RDD data structures, the similar square of the article preserved
Battle array is also read into Spark-RDD, is realized matrix multiplication operation using Spark, is obtained user-article marking value.Push away in real time
Recommend method flow diagram as shown in Figure 2., should with efficient and fault-tolerant characteristic using the Spark based on internal memory as enforcement engine
Proposed algorithm is realized with SparkStreaming, the purpose that recommendation list is generated in real time can be reached.
We are tested the inventive method according to the way ox net clickstream data of 15 days, user's row in these data
It is that daily PV numbers are 2000W or so for data.If conventionally calculating recommendation list to full dose data every time, recommend
Program operation duration takes around 90 minutes.And the real-time recommendation method provided using the present invention, newest visitor's behavior is collected, most
Visitor's base sheet in nearly 5 minutes is extracted after the user behavior for obtaining this part visitor in 1.5W or so, do mark it is unitized and
User preference matrix is pre-processed, and the article similar matrix preserved does matrix operation and obtains recommendation list, then recommends the time can
Recommendation list renewal is completed in second level to narrow down to.
Technological means disclosed in the present invention program is not limited only to the technological means disclosed in above-mentioned embodiment, in addition to
Constituted technical scheme is combined by above technical characteristic.It should be pointed out that for those skilled in the art
For, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (9)
1. a kind of real-time recommendation method based on article similarity, it is characterised in that comprise the following steps：
Step A, collects recent user behavior data, behavioral data is pre-processed, when user is in Entered state, used
UserId Murmur cryptographic Hash is identified as user；User uses visitor_trace Murmur under non-Entered state
Cryptographic Hash is identified as user, and user's mark is converted into unified signless long data value；
Step B, calculates each user to browsing the preference values of commodity using preference function model, generates basic user article preference
Value matrix, calculating obtains article similar matrix,
The preference function model is as follows：
score = i +
WhereiniFor initial value, xjRepresent the time difference of current time and user's access time；
Step C, filters out the newest behavioral data of current visitor, is processed into real-time preference matrix, conjugate product similar matrix, meter
User's recommendation list is calculated, including：
Step C-1, screens all behavioral datas of newest Guest User, data is pre-processed, when user is logging in shape
State, the Murmur cryptographic Hash using userId is identified as user；User uses visitor_trace under non-Entered state
Murmur cryptographic Hash identified as user, by user mark be uniformly converted into long numerical value；
Step C-2, using each user of preference function model conversion to browsing the preference values of commodity, by all users to some thing
The preference value of product calculates the similarity between article as a vector, generates active user article preference matrix；
Step C-3, the article similar matrix preserved using the step B obtained same day, using matrix multiplication, passes through following formula meter
Calculate：
Article similar matrix * active user article preference matrixs
Step C-4, the result obtained based on step C-3 is calculated each user to the median sum under same article, obtains each use
Family filters out user and browsed with after the commodity that placed an order, sort, taken before coming by preference value size to the predicted value of each product
Some commodity be recommendation list.
2. the real-time recommendation method according to claim 1 based on article similarity, it is characterised in that in order to avoid recommending
In hot product, step C-4, similarity is added as penalty factor, predicted value is calculated by following formula and obtained：
Predicted value=Σ medians/Σ similarities.
3. the real-time recommendation method according to claim 1 based on item-based, it is characterised in that：The step C bases
In MapReduce exploitations.
4. the real-time recommendation method according to claim 3 based on item-based, it is characterised in that：The step C sieves
Select all behavioral datas of Guest User in 5 minutes.
5. the real-time recommendation method according to claim 1 based on article similarity, it is characterised in that：The step C bases
In SparkStreaming exploitations.
6. the real-time recommendation method according to claim 5 based on article similarity, it is characterised in that：Time window is set
For 5s, data flow update status is detected every 5s, once updating, is then recommended after the data operation for collecting renewal.
7. the real-time recommendation method according to claim 1 based on article similarity, it is characterised in that：The article is similar
Matrix is by calling mahout-itembased interfaces to obtain.
8. the real-time recommendation method according to claim 1 based on article similarity, it is characterised in that：The article is similar
Matrix is preserved using sparse matrix.
9. the real-time recommendation method according to claim 1 based on article similarity, it is characterised in that：When user is in user
When only being given a mark in article preference value matrix to an article, then the recommendation list of the user is not calculated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611144613.8A CN107066476B (en) | 2016-12-13 | 2016-12-13 | Real-time recommendation method based on article similarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611144613.8A CN107066476B (en) | 2016-12-13 | 2016-12-13 | Real-time recommendation method based on article similarity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107066476A true CN107066476A (en) | 2017-08-18 |
| CN107066476B CN107066476B (en) | 2021-07-30 |
Family
ID=59619498
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611144613.8A Active CN107066476B (en) | 2016-12-13 | 2016-12-13 | Real-time recommendation method based on article similarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107066476B (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198019A (en) * | 2017-12-27 | 2018-06-22 | 网易无尾熊（杭州）科技有限公司 | Item recommendation method and device, storage medium, electronic equipment |
| CN108256018A (en) * | 2018-01-09 | 2018-07-06 | 北京奇艺世纪科技有限公司 | A kind of item recommendation method and device |
| CN108269167A (en) * | 2017-12-22 | 2018-07-10 | 合肥天源迪科信息技术有限公司 | A kind of e-commerce operation management method based on big data processing |
| CN108510373A (en) * | 2018-04-12 | 2018-09-07 | 京东方科技集团股份有限公司 | Paintings recommend method, paintings recommendation apparatus, equipment and storage medium |
| CN108647227A (en) * | 2018-03-26 | 2018-10-12 | 北京奇艺世纪科技有限公司 | A kind of recommendation method and device |
| CN108647996A (en) * | 2018-04-11 | 2018-10-12 | 中山大学 | A kind of personalized recommendation method and system based on Spark |
| CN109190033A (en) * | 2018-08-23 | 2019-01-11 | 微梦创科网络科技（中国）有限公司 | A kind of user's friend recommendation method and system |
| CN109657434A (en) * | 2018-12-18 | 2019-04-19 | 浙江口碑网络技术有限公司 | Application access method and device |
| CN109697629A (en) * | 2018-11-15 | 2019-04-30 | 平安科技（深圳）有限公司 | Product data method for pushing and device, storage medium, computer equipment |
| CN109903104A (en) * | 2017-12-08 | 2019-06-18 | 北京京东尚科信息技术有限公司 | Method and device for data analysis |
| CN110473043A (en) * | 2018-05-11 | 2019-11-19 | 北京京东尚科信息技术有限公司 | A kind of item recommendation method and device based on user behavior |
| CN110659417A (en) * | 2019-09-12 | 2020-01-07 | 广东浪潮大数据研究有限公司 | Information pushing method and system, electronic equipment and storage medium |
| CN110688577A (en) * | 2019-09-26 | 2020-01-14 | 招商局金融科技有限公司 | Intelligent product recommendation method and device and computer readable storage medium |
| CN110689410A (en) * | 2019-09-29 | 2020-01-14 | 京东数字科技控股有限公司 | Data processing method, device, equipment and storage medium |
| CN110738538A (en) * | 2018-07-18 | 2020-01-31 | 北京京东尚科信息技术有限公司 | Method and device for identifying similar articles |
| CN111191602A (en) * | 2019-12-31 | 2020-05-22 | 深圳云天励飞技术有限公司 | Pedestrian similarity obtaining method and device, terminal equipment and readable storage medium |
| CN112765466A (en) * | 2021-01-15 | 2021-05-07 | 电子科技大学 | Equipment personalized recommendation method based on user history |
| CN112950242A (en) * | 2019-12-10 | 2021-06-11 | 深圳云天励飞技术有限公司 | Information pushing method and device and electronic equipment |
| CN113190763A (en) * | 2021-06-28 | 2021-07-30 | 南京信息工程大学 | Information recommendation method and system |
| CN113127750B (en) * | 2021-05-19 | 2024-04-02 | 同盾科技有限公司 | Information list generation method and device, storage medium and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070033098A1 (en) * | 2005-08-05 | 2007-02-08 | International Business Machines Corporation | Method, system and storage medium for creating sales recommendations |
| CN103824213A (en) * | 2014-03-13 | 2014-05-28 | 北京理工大学 | Individualized recommendation method based on user preferences and commodity properties |
| CN103995839A (en) * | 2014-04-30 | 2014-08-20 | 兴天通讯技术（天津）有限公司 | Commodity recommendation optimizing method and system based on collaborative filtering |
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| CN106022865A (en) * | 2016-05-10 | 2016-10-12 | 江苏大学 | Goods recommendation method based on scores and user behaviors |
-
2016
- 2016-12-13 CN CN201611144613.8A patent/CN107066476B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070033098A1 (en) * | 2005-08-05 | 2007-02-08 | International Business Machines Corporation | Method, system and storage medium for creating sales recommendations |
| CN103824213A (en) * | 2014-03-13 | 2014-05-28 | 北京理工大学 | Individualized recommendation method based on user preferences and commodity properties |
| CN103995839A (en) * | 2014-04-30 | 2014-08-20 | 兴天通讯技术（天津）有限公司 | Commodity recommendation optimizing method and system based on collaborative filtering |
| CN105469263A (en) * | 2014-09-24 | 2016-04-06 | 阿里巴巴集团控股有限公司 | Commodity recommendation method and device |
| CN106022865A (en) * | 2016-05-10 | 2016-10-12 | 江苏大学 | Goods recommendation method based on scores and user behaviors |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109903104A (en) * | 2017-12-08 | 2019-06-18 | 北京京东尚科信息技术有限公司 | Method and device for data analysis |
| CN108269167A (en) * | 2017-12-22 | 2018-07-10 | 合肥天源迪科信息技术有限公司 | A kind of e-commerce operation management method based on big data processing |
| CN108198019A (en) * | 2017-12-27 | 2018-06-22 | 网易无尾熊（杭州）科技有限公司 | Item recommendation method and device, storage medium, electronic equipment |
| CN108256018A (en) * | 2018-01-09 | 2018-07-06 | 北京奇艺世纪科技有限公司 | A kind of item recommendation method and device |
| CN108647227A (en) * | 2018-03-26 | 2018-10-12 | 北京奇艺世纪科技有限公司 | A kind of recommendation method and device |
| CN108647996B (en) * | 2018-04-11 | 2022-04-19 | 中山大学 | Spark-based personalized recommendation method and system |
| CN108647996A (en) * | 2018-04-11 | 2018-10-12 | 中山大学 | A kind of personalized recommendation method and system based on Spark |
| CN108510373A (en) * | 2018-04-12 | 2018-09-07 | 京东方科技集团股份有限公司 | Paintings recommend method, paintings recommendation apparatus, equipment and storage medium |
| WO2019196881A1 (en) * | 2018-04-12 | 2019-10-17 | 京东方科技集团股份有限公司 | Recommendation method, recommendation apparatus, recommendation device, recommendation system, and storage medium |
| CN110473043A (en) * | 2018-05-11 | 2019-11-19 | 北京京东尚科信息技术有限公司 | A kind of item recommendation method and device based on user behavior |
| CN110738538B (en) * | 2018-07-18 | 2024-05-24 | 北京京东尚科信息技术有限公司 | Method and device for identifying similar objects |
| CN110738538A (en) * | 2018-07-18 | 2020-01-31 | 北京京东尚科信息技术有限公司 | Method and device for identifying similar articles |
| CN109190033B (en) * | 2018-08-23 | 2022-01-11 | 微梦创科网络科技（中国）有限公司 | User friend recommendation method and system |
| CN109190033A (en) * | 2018-08-23 | 2019-01-11 | 微梦创科网络科技（中国）有限公司 | A kind of user's friend recommendation method and system |
| CN109697629A (en) * | 2018-11-15 | 2019-04-30 | 平安科技（深圳）有限公司 | Product data method for pushing and device, storage medium, computer equipment |
| WO2020098156A1 (en) * | 2018-11-15 | 2020-05-22 | 平安科技（深圳）有限公司 | Product data pushing method and apparatus, and non-volatile computer-readable storage medium |
| CN109657434A (en) * | 2018-12-18 | 2019-04-19 | 浙江口碑网络技术有限公司 | Application access method and device |
| CN110659417A (en) * | 2019-09-12 | 2020-01-07 | 广东浪潮大数据研究有限公司 | Information pushing method and system, electronic equipment and storage medium |
| CN110688577A (en) * | 2019-09-26 | 2020-01-14 | 招商局金融科技有限公司 | Intelligent product recommendation method and device and computer readable storage medium |
| CN110689410A (en) * | 2019-09-29 | 2020-01-14 | 京东数字科技控股有限公司 | Data processing method, device, equipment and storage medium |
| CN110689410B (en) * | 2019-09-29 | 2021-09-03 | 京东数字科技控股有限公司 | Data processing method, device, equipment and storage medium |
| CN112950242A (en) * | 2019-12-10 | 2021-06-11 | 深圳云天励飞技术有限公司 | Information pushing method and device and electronic equipment |
| CN112950242B (en) * | 2019-12-10 | 2024-08-23 | 深圳云天励飞技术有限公司 | Information pushing method and device and electronic equipment |
| CN111191602A (en) * | 2019-12-31 | 2020-05-22 | 深圳云天励飞技术有限公司 | Pedestrian similarity obtaining method and device, terminal equipment and readable storage medium |
| CN112765466A (en) * | 2021-01-15 | 2021-05-07 | 电子科技大学 | Equipment personalized recommendation method based on user history |
| CN112765466B (en) * | 2021-01-15 | 2022-05-03 | 电子科技大学 | Equipment personalized recommendation method based on user history |
| CN113127750B (en) * | 2021-05-19 | 2024-04-02 | 同盾科技有限公司 | Information list generation method and device, storage medium and electronic equipment |
| CN113190763A (en) * | 2021-06-28 | 2021-07-30 | 南京信息工程大学 | Information recommendation method and system |
| CN113190763B (en) * | 2021-06-28 | 2021-09-24 | 南京信息工程大学 | Information recommendation method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107066476B (en) | 2021-07-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107066476A (en) | A kind of real-time recommendation method based on article similarity | |
| CN103714084B (en) | The method and apparatus of recommendation information | |
| CN109559208A (en) | A kind of information recommendation method, server and computer-readable medium | |
| CN109508419B (en) | Knowledge learning-based recommendation method and system | |
| CN104899273B (en) | A kind of Web Personalization method based on topic and relative entropy | |
| CN102831234B (en) | Personalized news recommendation device and method based on news content and theme feature | |
| CN102279851B (en) | Intelligent navigation method, device and system | |
| CN109190043A (en) | Recommended method and device, storage medium, electronic equipment and recommender system | |
| CN103729446B (en) | A kind of processing method, processing unit and the server of user's operation data | |
| US8990208B2 (en) | Information management and networking | |
| CN106997358A (en) | Information recommendation method and device | |
| CN107103000A (en) | It is a kind of based on correlation rule and the integrated recommended technology of Bayesian network | |
| CN107016026A (en) | A kind of user tag determination, information-pushing method and equipment | |
| CN103353880B (en) | A kind of utilization distinctiveness ratio cluster and the data digging method for associating | |
| CN103345698A (en) | Personalized recommendation method based on cloud processing mode and applied in e-business environment | |
| CN104899229A (en) | Swarm intelligence based behavior clustering system | |
| CN104679835A (en) | Book recommending method based on multi-view hash | |
| CN107895038A (en) | A kind of link prediction relation recommends method and device | |
| CN103744904B (en) | A kind of method and device that information is provided | |
| CN104063429A (en) | Predicting method for user behavior in e-commerce | |
| Li | Accurate digital marketing communication based on intelligent data analysis | |
| CN105389330B (en) | Across the community open source resources of one kind match correlating method | |
| CN104008204A (en) | Dynamic multi-dimensional context awareness film recommending system and achieving method thereof | |
| CN116362823A (en) | Recommendation model training method, recommendation method and recommendation device for behavior sparse scene | |
| CN110910201B (en) | Information recommendation control method and device, computer equipment and storage medium |
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 |