CN104778237A - Individual recommending method and system based on key users - Google Patents
Individual recommending method and system based on key users Download PDFInfo
- Publication number
- CN104778237A CN104778237A CN201510157504.9A CN201510157504A CN104778237A CN 104778237 A CN104778237 A CN 104778237A CN 201510157504 A CN201510157504 A CN 201510157504A CN 104778237 A CN104778237 A CN 104778237A
- Authority
- CN
- China
- Prior art keywords
- user
- product
- key
- neighbor list
- recommendation
- 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.)
- Pending
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an individual recommending method and system based on key users and belongs to the technical field of sensors. The individual recommending method comprises the steps of searching for N neighbors with the highest similarity for each user according to collected user data to obtain a neighbor list of each user, ranking the neighbors according to the similarity from high to low, respectively calculating out the weight of each user based on the user neighbor lists, searching for P users with the maximum weights as the key users, calculating out the recommendation of target users to all products of a platform system based on operation data of the key users to the products and searching for M products with the maximum recommendation as the recommendation results for the target users. In addition, if the higher frequency the user appears in the neighbor list, the higher the rank of the user is, and the greater the weight of the user is. Meanwhile, the invention further discloses an individual recommending system based on the method. The individual recommending method is applied to a network information resource recommending system, can effectively reduce the calculation complexity of recommending calculation and improves the processing efficiency of the system.
Description
Technical field
The invention belongs to net application technology field, be specifically related to a kind of recommend method to network information resource and system.
Background technology
In recent years, along with the growing of infotech and the continuous growth being connected into Internet user's quantity, the Bit data of enormous amount can in internet, at every moment all be produced.These information how are effectively utilized to become the problem of a worldwide concern.Through academia and industry member years of researches and application, for these data surcharges excavation and utilize technology also ripe gradually, topmost is exactly related information commending system in fields such as ecommerce, online information application (as online news, Online Music, online video display).
Most widely used in commending system is collaborative filtering, comprises based on neighbours and the method based on model.Be generally used for studying the Similarity Measure problem between user or product based on the method for neighbours: first the collaborative filtering method based on user searches the similar user of underlying attribute (i.e. neighbours), then the data of Collection and analysis neighbours are with further for targeted customer recommends its interested product.Similar, the collaborative filtering method based on product make use of the advantage of the information grading of like product.Method based on model attempts by the relationship assessment data transformations of user-product be different models (as Bayesian network, factorization or cluster models etc.) and use these models in unknown scene to user's recommended products.
Current, product category involved in commending system is various, and may have thousands of, even more than 1,000,000, and number of users also can be very huge, and when in the face of large-scale data like this, existing recommended technology exists the technological deficiency of inefficiency.
Summary of the invention
Goal of the invention of the present invention is: provide a kind of personalized recommendation method based on key user and the system that reduce computation complexity.
A kind of personalized recommendation method based on key user of the present invention, comprises the following steps:
Step 1: gather user to the service data of product, comprise user to the purchase of product, browse and the operation behavior data such as comment;
Step 2: for each user searches the highest user of N number of similarity as each user's neighbor list, in described user's neighbor list, each user arranges from high to low according to similarity;
Step 3: the weight calculating each user based on each user's neighbor list respectively: user's occurrence number in user's neighbor list is more, and sorting position is more forward, and weight is larger; Search the maximum P of a weight user as key user;
Step 4: targeted customer is calculated to the recommendation degree of each product of plateform system (as the collaborative filtering recommending method based on user to the service data of product based on P key user, or based on material diffusion recommend method or based on the usual recommendation computing method such as random walk recommend method), search the recommendation results of the maximum M of a recommendation degree product as targeted customer.
The user's weight that the present invention is based on set by the present invention is screened plateform system user collection, decrease the number of users at calculated recommendation degree, and then decrease the product of its association, thus effectively reduce the computation complexity recommending to calculate, improve commending system treatment effeciency; Information resources corresponding to key user simultaneously after screening are with a high credibility, thus effectively improve the accuracy and confidence of recommendation results.
Further, the present invention, when calculating the similarity between any two users, can adopt the processing mode of cosine angle, concrete as shown in formula (1):
Wherein s
ijrepresent the similarity of user i and user j, subscript " i " and " j " they are user identifier, and i ≠ j, subscript " a " is product identifiers, and n is the product number of plateform system, a
i αrepresent that user i is to the operational attribute value of product a, if user exists operation behavior to product, then operational attribute value is 1, otherwise is 0, k
irepresent in plateform system, there is the product sum of operation behavior in the product of user i to plateform system, namely
in like manner known
Meanwhile, the invention also discloses a kind of personalized recommendation system based on key user, comprising:
Data processing module: for from the background data base of plateform system, gather and store the service data of user to product;
Key user's processing module: based on the service data of the user inputted from data processing module to product, the step 2 and 3 according to claim 1 or 2 calculates key user and is sent to recommending module; Interactive module: obtain plateform system current accessed user and as targeted customer; The recommendation results simultaneously recommending module sent is shown to the targeted customer of correspondence; Recommending module: the targeted customer sent based on interactive module, calculates the recommendation degree of current goal user to each product of plateform system according to key user, searches the maximum M of a recommendation degree product as the recommendation results of current goal user and sends to interactive module.
In order to reduce calculated amount, data processing module also carries out Screening Treatment to the user stored to the service data of product: if the product number that the product of user i to plateform system exists operation behavior is less than K3*n (n is the product number of plateform system), user then corresponding to deletion user i is to the service data of product, and wherein the span of K3 is 0.1% ~ 1%.
Simultaneously, key user's processing module can be set to processed offline module, the user that can periodically send from data processing module calculates and output key user the service data of product, and its computation period is one week or one month, to reduce the calculated amount of commending system further.
In sum, owing to have employed technique scheme, the invention has the beneficial effects as follows:
The present invention, when to targeted customer's recommended products, only considers the key user that weight is high, and information resources corresponding to these key users are with a high credibility, thus effectively can improve the accuracy and confidence of recommendation results; Simultaneously owing to eliminating the low user of weight, thus effectively can reduce the computation complexity recommending to calculate, promote commending system treatment effeciency.
Accompanying drawing explanation
Fig. 1 is the commending system structural drawing of embodiment;
Fig. 2 is user and product binary relation network diagram;
Fig. 3 is the user-product relation schematic diagram of key user;
Fig. 4 is traditional material diffusion recommend method and the comparison diagram based on the material method of diffusion of key user.
Embodiment
For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail.
See Fig. 1, personalized recommendation system based on key user of the present invention comprises data processing module, key user's processing module, recommending module and interactive module, wherein data processing module can comprise collecting unit, screening unit and storage unit, collecting unit is used for from the background data base of plateform system, gather the service data (hereinafter referred to as user data) of user to product, in order to reduce calculated amount, (n is the product number of plateform system by screening unit, the product number product of plateform system being existed to operation behavior to be less than K3*n, the span of K3 is 0.1% ~ 1%), delete, there is operation behavior product number and be more than or equal to the user of K3*n and store in storage unit in reservation.Key user's processing module reads user data from data processing module and carries out the calculating of key user and export P (P is preset value, can arrange based on the number of users of the accuracy requirement of system and plateform system, usually K2*m can be set to, wherein m represents the number of users of plateform system, the span of K2 is xx% ~ xx%) individual key user carries out the computing of Products Show list to recommending module, recommending module is used for based on all products of P key user's computing platform system the recommendation degree of current goal user, if wherein calculating required corresponding data information is present in data processing module, can directly read from its storage unit, if do not exist, then need to read from the background data base of plateform system, concrete recommendation calculates and existing either type can be adopted to realize, such as based on the collaborative filtering recommending method of user, the recommend method etc. based on material diffusion, when computing, the data source of institute's foundation is no longer all users of existing plateform system, but the key user that the present invention screens, thus reduce the computation complexity of recommendation degree.Front M (empirical value, the demand specifically based on commending system is arranged, and usually can be set to 10 ~ 100) individual recommended products list is sent to interactive module by recommending module.Interactive module comprises targeted customer's detecting unit and display unit, wherein targeted customer's detecting unit is used for the user access information based on plateform system, current accessed user is sent to recommending module as targeted customer, the recommended products list simultaneously sent based on recommending module is shown to current goal user, such as based on product information each in recommended products list, from the background data base of plateform system, read corresponding product image information show to current goal user.
In the present invention, when calculating P key user, first N (predetermined threshold value is searched for each user, accuracy requirement based on commending system is arranged, usually can be set to K1*m, wherein the span of K1 is 1% ~ 5%) the highest user of individual similarity is as each user's neighbor list (top-N neighbor list).For the ease of calculating, the user data that can store based on data processing module builds adjacency matrix A
m × n={ a
i αrepresent the binary relation of each user and each product, wherein a
i αrepresent that user i (i is user identifier) is to product α operational attribute value, if there is operation behavior (buy, comment on or browse) to product α, then a with i in every family
i α=1, otherwise a
i α=0, m and n is respectively number of users and the product number of plateform system, and the data source of user to the operation behavior of product comes from the background data base of plateform system.Above-mentioned binary relation also can represent with two partial graphs, as shown in Figure 2.In Fig. 2, one has 5 users and 5 commodity, and the line between user and commodity represents to there is operation behavior.
Based on adjacency matrix A
m × ntop-N neighbor list is obtained based on the similarity between user, when calculating the similarity between user, the present invention can adopt arbitrary conventional process to realize, also the cosine angle method that the present invention provides can be adopted, namely calculate the similarity between wantonly one or two user respectively based on formula (1), then search N number of most similar neighborhood stored in top-N neighbor list, each neighbours arrange from high to low based on similarity.Table 1 provides the top-N neighbor list (the often row in the table 1 i.e. top-N neighbor list of a corresponding user) of (N=2) of all users of the user data based on Fig. 2:
Table 1
| Most similar neighborhood | |
| User 1 | User 2, user 4 |
| User 2 | User 1, user 4 |
| User 3 | User 5, user 2 |
| User 4 | User 1, user 2 |
| User 5 | User 3, user 2 |
Secondly, the top-N neighbor list based on each user calculates the weight of each user respectively: user's occurrence number in user's neighbor list is more, and sorting position is more forward, and weight is larger.In this embodiment, the calculating of weight is specially:
1) to the top-N neighbor list of each user, arrange the elementary weighted value of each user: the elementary weighted value appearing at the user in current top-N neighbor list is the inverse of the sorting position in top-N neighbor list, the elementary weighted value not appearing at the user of top-N neighbor list is 0.For the top-N neighbor list of the user 1 shown in table 1, its top-N neighbor list is: user 2, user 4, the weighted value then corresponding to each user (user 1-5) of the top-N neighbor list of user 1 is respectively: 0,1/1,1/2,0,0 data line of user 1 correspondence (in the table 2).Table 2 gives each top-N neighbours of correspondence based on table 1 the elementary weighted value example arranged.
Table 2
| User 1 | User 2 | User 3 | User 4 | User 5 | |
| User 1 | 0 | 1 | 0 | 0.5 | 0 |
| User 2 | 1 | 0 | 0 | 0.5 | 0 |
| User 3 | 0 | 0.5 | 0 | 0 | 1 |
| User 4 | 1 | 0.5 | 0 | 0 | 0 |
| User 5 | 0 | 0.5 | 1 | 0 | 0 |
2) the cumulative elementary weighted value of same user in different top-N neighbor list, obtains the weighted value of each user.Table 3 give final each user that the elementary weighted value based on table 2 obtains weighted value and from the sequence of height highland.For user 2, its final weighted value is: 1/1+1/2+1/2+1/2=2.5.
Table 3
| Weighted value | |
| User 2 | 2.5 |
| User 1 | 2 |
| User 3 | 1 |
| User 4 | 1 |
| User 5 | 1 |
Finally, search the maximum P of a weight user as key user, for Fig. 2, if arrange P=2, then corresponding key user is user 2 and user 1.In the system of reality, in order to take into account efficiency and recommend precision, the value of preferred P is 0.2*m.
In the present invention, the computation process of key user can complete by off-line.When to targeted customer's recommended products, recommend computing method only to need to consider the data of key user, thus reduce calculated amount.Fig. 3 illustrates the relational network only comprising key user in Fig. 2, and key user is respectively user 1 and user 2, and user 3 is targeted customer.In this embodiment, with based on user collaborative filtering recommending computing method and based on material diffusion recommendation be calculated as example, the calculating of the recommendation degree based on key user is described.
Traditional needs based on user collaborative filtered recommendation computing method the similarity calculating targeted customer and all the other all users, and therefore calculated amount is comparatively large, and many incoherent users are also taken into account, thus it is lower to result in recommendation results precision.The present invention, when to user's recommended products, only calculates the similarity of targeted customer and key user, and therefore user i (targeted customer) marks with the prediction of product α and is:
wherein U represents that key user of the present invention gathers,
it is normalized parameter.M the product that predicted value is the highest gives user i by recommended.In the example given by Fig. 3, when to user's 3 recommended products, the present invention only calculates the similarity of user 3 and user 1 and user 2, and traditional similarity needing to calculate user 3 and user 1, user 2, user 4 and user 5 based on user collaborative filter algorithm.But user 4 and user 5 may be fictitious users, often affect the precision of other user's recommendation results.
Traditional material diffusion computing method are divided into three steps: (1) distributes product that 1 unit resource selected to targeted customer (specifically refer to there is purchase to product, browse or the product of the operation behavior such as evaluation), as in Fig. 4, distribute 1 element resources to the selected product crossed of user 3, the numeral 1 after the product icon (circle) in corresponding diagram; (2) 1 unit resource is diffused into fifty-fifty the neighbor user node be connected with product; (3) resource obtained is reassigned to its product selected by user again that receive step (2) described resource.The nonoptional product of ideal user is according to the order sequence from big to small of its resource received, and M the most front Products Show that sort is to targeted customer.Suppose
for the initial resource vector of targeted customer, then the final resource vector of product
wherein W is transition matrix:
wherein " a ", " β " (a ≠ β) are product identifiers, k
jfor the product number (product of user j to plateform system exists the product sum of operation behavior) that user j selects, k
βrepresent the user's number selecting product β.If the method to be expanded to the calculating based on key user of the present invention, so in step (2), only have key user to receive resource, and the resource of acquisition is averagely allocated to product.This process is similar to deletes non-key user node in network.Fig. 4 gives traditional material diffusion computing method and the material diffusion computing method based on key user, wherein user 3 is targeted customer, numeral in figure near product icon, user's icon (1/3,32/72 etc.) represent its resource information received.Based in the material diffusion computing method of key user, user 4 and user 5 node are removed, and therefore only have user 1 and user 2 could receive initial resource from user 3.
The above, be only the specific embodiment of the present invention, arbitrary feature disclosed in this specification, unless specifically stated otherwise, all can be replaced by other equivalences or the alternative features with similar object; Step in disclosed all features or all methods or process, except mutually exclusive feature and/or step, all can be combined in any way.
Claims (10)
1. based on a key user's personalized recommendation method, it is characterized in that, comprise the following steps:
Step 1: gather user to the service data of product;
Step 2: for each user searches the highest user of N number of similarity as each user's neighbor list, in described user's neighbor list, each user arranges from high to low according to similarity;
Step 3: the weight calculating each user based on each user's neighbor list respectively: user's occurrence number in user's neighbor list is more, and arrangement position is more forward, and weight is larger; Search the maximum P of a weight user as key user;
Step 4: based on P key user, targeted customer is calculated to the recommendation degree of each product of plateform system to the service data of product, search the recommendation results of the maximum M of a recommendation degree product as targeted customer.
2. the method for claim 1, is characterized in that, in described step 2, and the similarity s between user
ijcomputing formula be:
wherein subscript " i " and " j " they are user identifier, and i ≠ j, subscript " a " is product identifiers, and n is the product number of plateform system, a
i α, a
j αrepresent the operational attribute value of each user to product a, if user exists operation behavior to product, then operational attribute value is 1, otherwise is 0, described parameter
3. method as claimed in claim 1 or 2, it is characterized in that, in described step 3, the weight calculating each user is:
To each user's neighbor list T
i, the elementary weighted value of each user is set: appear at user's neighbor list T
iin the elementary weighted value of user be user's neighbor list T
iin the inverse of sorting position, do not appear at user's neighbor list T
iin the elementary weighted value of user be 0, wherein subscript " i " is user's neighbor list identifier;
The cumulative elementary weighted value of same user in different user neighbor list, obtains the weighted value of each user.
4. method as claimed in claim 1 or 2, is characterized in that, in described step 1, the service data of user to product comprises user to the purchaser record of product and/or browse record and/or review record.
5. method as claimed in claim 1 or 2, it is characterized in that, in described step 4, the computing method of targeted customer to the recommendation degree of each product of plateform system are: based on the collaborative filtering recommending method of user, or based on material diffusion recommend method or based on random walk recommend method.
6. method as claimed in claim 1 or 2, it is characterized in that, in described step 2, the value of described N is: the number of users of K1* plateform system, and wherein the span of K1 is 1% ~ 5%
7. method as claimed in claim 1 or 2, it is characterized in that, in described step 3, the value of described P is: the number of users of K2* plateform system, and wherein the span of K2 is 10% ~ 20%
8. based on a key user's personalized recommendation system, it is characterized in that, comprising:
Data processing module: for from the background data base of plateform system, gather and store the service data of user to product;
Key user's processing module: based on the service data of the user inputted from data processing module to product, the step 2 and 3 according to claim 1 or 2 calculates key user and is sent to recommending module;
Interactive module: obtain plateform system current accessed user and as targeted customer; The recommendation results simultaneously recommending module sent is shown to the targeted customer of correspondence;
Recommending module: the targeted customer sent based on interactive module, calculates the recommendation degree of current goal user to each product of plateform system according to key user, searches the maximum M of a recommendation degree product as the recommendation results of current goal user and sends to interactive module.
9. system as claimed in claim 8, it is characterized in that, described data processing module also carries out Screening Treatment to described service data: if the product number that the product of user i to plateform system exists operation behavior is less than the product number of K3* plateform system, user then corresponding to deletion user i is to the service data of product, and wherein the span of K3 is 0.1% ~ 1%
10. system as claimed in claim 8 or 9, it is characterized in that, described key user's processing module is processed offline module, and its computation period is one week or one month.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510157504.9A CN104778237A (en) | 2015-04-03 | 2015-04-03 | Individual recommending method and system based on key users |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510157504.9A CN104778237A (en) | 2015-04-03 | 2015-04-03 | Individual recommending method and system based on key users |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104778237A true CN104778237A (en) | 2015-07-15 |
Family
ID=53619701
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510157504.9A Pending CN104778237A (en) | 2015-04-03 | 2015-04-03 | Individual recommending method and system based on key users |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104778237A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426392A (en) * | 2015-10-28 | 2016-03-23 | 浪潮软件集团有限公司 | Collaborative filtering recommendation method and system |
| CN106815217A (en) * | 2015-11-30 | 2017-06-09 | 北京云莱坞文化传媒有限公司 | Story recommends method and story recommendation apparatus |
| CN106952111A (en) * | 2017-02-27 | 2017-07-14 | 东软集团股份有限公司 | Personalized recommendation method and device |
| CN107403390A (en) * | 2017-08-02 | 2017-11-28 | 桂林电子科技大学 | A kind of friend recommendation method for merging Bayesian inference and the upper random walk of figure |
| CN107889082A (en) * | 2017-11-01 | 2018-04-06 | 南京邮电大学 | A kind of D2D method for discovering equipment using social networks between user |
| CN108920624A (en) * | 2018-06-29 | 2018-11-30 | 西安电子科技大学 | Recommended method based on evolution multi-objective Algorithm extraction system key user |
| CN109308654A (en) * | 2018-11-20 | 2019-02-05 | 辽宁师范大学 | Collaborative filtering recommending method based on article energy dissipation and user preference |
| CN109523344A (en) * | 2018-10-16 | 2019-03-26 | 深圳壹账通智能科技有限公司 | Product information recommended method, device, computer equipment and storage medium |
| CN109919737A (en) * | 2019-03-20 | 2019-06-21 | 中电科大数据研究院有限公司 | A kind of recommended method and system of production and sales commodity |
| CN111369306A (en) * | 2020-06-01 | 2020-07-03 | 北京搜狐新媒体信息技术有限公司 | Product recommendation method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030019939A1 (en) * | 2001-07-27 | 2003-01-30 | Sellen Abigail Jane | Data acquisition and processing system and method |
| CN102254028A (en) * | 2011-07-22 | 2011-11-23 | 青岛理工大学 | Personalized commodity recommending method and system which integrate attributes and structural similarity |
-
2015
- 2015-04-03 CN CN201510157504.9A patent/CN104778237A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030019939A1 (en) * | 2001-07-27 | 2003-01-30 | Sellen Abigail Jane | Data acquisition and processing system and method |
| CN102254028A (en) * | 2011-07-22 | 2011-11-23 | 青岛理工大学 | Personalized commodity recommending method and system which integrate attributes and structural similarity |
Non-Patent Citations (1)
| Title |
|---|
| EI ZENG、AN ZENG、HAO LIU、MING-SHENG SHANG、TAO ZHOU: "uncovering the information core in recommender systems", 《SCIENTIFIC REPORTS》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426392A (en) * | 2015-10-28 | 2016-03-23 | 浪潮软件集团有限公司 | Collaborative filtering recommendation method and system |
| CN105426392B (en) * | 2015-10-28 | 2019-03-26 | 浪潮软件集团有限公司 | Collaborative filtering recommendation method and system |
| CN106815217A (en) * | 2015-11-30 | 2017-06-09 | 北京云莱坞文化传媒有限公司 | Story recommends method and story recommendation apparatus |
| CN106952111A (en) * | 2017-02-27 | 2017-07-14 | 东软集团股份有限公司 | Personalized recommendation method and device |
| CN106952111B (en) * | 2017-02-27 | 2021-04-02 | 东软集团股份有限公司 | Personalized recommendation method and device |
| CN107403390B (en) * | 2017-08-02 | 2020-06-02 | 桂林电子科技大学 | Friend recommendation method integrating Bayesian reasoning and random walk on graph |
| CN107403390A (en) * | 2017-08-02 | 2017-11-28 | 桂林电子科技大学 | A kind of friend recommendation method for merging Bayesian inference and the upper random walk of figure |
| CN107889082A (en) * | 2017-11-01 | 2018-04-06 | 南京邮电大学 | A kind of D2D method for discovering equipment using social networks between user |
| CN107889082B (en) * | 2017-11-01 | 2020-04-14 | 南京邮电大学 | D2D device discovery method utilizing social relationship among users |
| CN108920624A (en) * | 2018-06-29 | 2018-11-30 | 西安电子科技大学 | Recommended method based on evolution multi-objective Algorithm extraction system key user |
| CN109523344A (en) * | 2018-10-16 | 2019-03-26 | 深圳壹账通智能科技有限公司 | Product information recommended method, device, computer equipment and storage medium |
| CN109308654A (en) * | 2018-11-20 | 2019-02-05 | 辽宁师范大学 | Collaborative filtering recommending method based on article energy dissipation and user preference |
| CN109919737A (en) * | 2019-03-20 | 2019-06-21 | 中电科大数据研究院有限公司 | A kind of recommended method and system of production and sales commodity |
| CN111369306A (en) * | 2020-06-01 | 2020-07-03 | 北京搜狐新媒体信息技术有限公司 | Product recommendation method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104778237A (en) | Individual recommending method and system based on key users | |
| CN102629360B (en) | A kind of effective dynamic commodity recommend method and commercial product recommending system | |
| CN103186539B (en) | A kind of method and system determining user group, information inquiry and recommendation | |
| CN107220365B (en) | Accurate recommendation system and method based on collaborative filtering and association rule parallel processing | |
| CN103544216B (en) | The information recommendation method and system of a kind of combination picture material and keyword | |
| CN100504866C (en) | Integrative searching result sequencing system and method | |
| CN111444394B (en) | Method, system and equipment for obtaining relation expression between entities and advertisement recall system | |
| CN111444395B (en) | Method, system and equipment for obtaining relation expression between entities and advertisement recall system | |
| CN108205766A (en) | Information-pushing method, apparatus and system | |
| CN106471491A (en) | A kind of collaborative filtering recommending method of time-varying | |
| CN103886048B (en) | Cluster-based increment digital book recommendation method | |
| CN107180093A (en) | Information search method and device and ageing inquiry word recognition method and device | |
| CN106708844A (en) | User group partitioning method and device | |
| CN103886090A (en) | Content recommendation method and device based on user favorites | |
| CN103136683A (en) | Method and device for calculating product reference price and method and system for searching products | |
| CN103164804A (en) | Personalized method and personalized device of information push | |
| CN110175895B (en) | Article recommendation method and device | |
| CN109409928A (en) | A kind of material recommended method, device, storage medium, terminal | |
| CN109460519B (en) | Browsing object recommendation method and device, storage medium and server | |
| CN105205188A (en) | Method and device for recommending purchase material suppliers | |
| CN108876537A (en) | A kind of mixed recommendation method for on-line mall system | |
| CN109840833A (en) | Bayes's collaborative filtering recommending method | |
| CN113837842A (en) | Commodity recommendation method and equipment based on user behavior data | |
| CN109977299A (en) | A kind of proposed algorithm of convergence project temperature and expert's coefficient | |
| Wu et al. | Discovery of associated consumer demands: Construction of a co-demanded product network with community detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150715 |
|
| RJ01 | Rejection of invention patent application after publication |