CN104992352A - Individualized resource retrieval method - Google Patents
Individualized resource retrieval method Download PDFInfo
- Publication number
- CN104992352A CN104992352A CN201510402669.8A CN201510402669A CN104992352A CN 104992352 A CN104992352 A CN 104992352A CN 201510402669 A CN201510402669 A CN 201510402669A CN 104992352 A CN104992352 A CN 104992352A
- Authority
- CN
- China
- Prior art keywords
- user
- data
- commodity
- real
- behavior
- 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
Abstract
The invention discloses an individualized resource retrieval method comprising the steps that step A, data collection is performed: the data include user real-time behavior data which are stored in a database and cache and user characteristic data and commodity characteristic data which are stored in distributed file systems; step B, engine calculation is performed: a recommendation engine calculates and acquires initial recommendation results according to the user behavior data collected in real time: step C, filtering processing is performed: filtering processing is performed on the acquired initial recommendation results by applying a collaborative filtering pseudo code algorithm; and step D, ranking explanation is performed: the initial recommendation results after filtering processing are merged and ranked according to certain weight or priority, and then recommendation explanation ranking the top is acquired and a final recommendation result is formed. According to the individualized resource retrieval method, the user behaviors can be accurately predicted, and the visual fields of users can also be expanded and the users are assisted to discover things in which the users are probably interested in while difficult to discover.
Description
Technical field
The present invention relates to a kind of personalized resource retrieval method, belong to Computer Applied Technology field.
Background technology
In electronic commerce times, businessman provides a large amount of commodity by shopping website, and client cannot just understand all commodity by screen at a glance, also directly cannot check the quality of commodity.So client needs a kind of electronic business transaction assistant, commodity that can be possible interested or satisfied according to the hobby recommend customers of client oneself.Personalized recommendation is Characteristic of Interest according to user and buying behavior, recommends the interested information of user and commodity to user.Along with the continuous expansion of ecommerce scale, commodity number and kind increase fast, and customer need spends a large amount of time just can find the commodity oneself wanting to buy.Thisly browse a large amount of irrelevant information and product process can make the consumer be submerged in problem of information overload constantly run off undoubtedly.In order to address these problems, personalized recommendation system arises at the historic moment.Personalized recommendation system is based upon mass data to excavate a kind of Advanced Business intelligent platform on basis, provides completely personalized decision support and information service to help e-commerce website for its customer purchase.
The maximum advantage of personalized recommendation is, it can collect user characteristics data and according to user characteristics, as interest preference, for user initiatively makes personalized recommendation.And the recommendation that system provides can real-time update, namely when the commodity storehouse in system or user characteristics storehouse change, the recommendation sequence provided can change automatically.This just substantially increases simplicity and the validity of e-commerce initiative, also improves the service level of enterprise simultaneously.All things considered, the effect of a successful personalized recommendation system is mainly manifested in following three aspects:
The first, the viewer of e-commerce website is changed into buyer: the visitor of e-commerce system does not often have desire to purchase in navigation process, and personalized recommendation system can recommend their interested commodity to user, thus facilitates purchasing process.
The second, the cross-selling ability of e-commerce website is improved: personalized recommendation system provides other valuable commercial product recommendings to user in user's purchasing process, user can buy from the recommendation list that system provides oneself really need but in purchasing process unexpected commodity, thus effectively improve the cross-selling of e-commerce system.
Three, client is improved to the loyalty of e-commerce website: compared with traditional business model, e-commerce system makes user have increasing selection, it is extremely convenient that user changes businessman, only needs click 1 twice mouse just can redirect between different e-commerce systems.Personalized recommendation system analyzes the buying habit of user, provides valuable commercial product recommending according to user's request to user.If the recommendation of commending system is of high quality, so user can produce this commending system and rely on.Therefore, personalized recommendation system can not only provide personalized recommendation service for user, and can set up relation steady in a long-term with user, thus effectively retains client, improves the loyalty of client, prevents customer churn.
Personalized recommendation system has good development and application prospect.At present, nearly all electronic business system, as Amazon, eBay etc., all in various degree employ various forms of commending system.Various Web site of providing personalized service also needs the support energetically of commending system.Under the competitive environment be growing more intense, personalized recommendation system effectively can retain client, improves the service ability of e-commerce system.Successful commending system can bring huge benefit.
Summary of the invention
The present invention for the demand that prior art exists, provides a kind of personalized resource retrieval method, meets actual operation requirements just.
For solving the problem, the technical solution used in the present invention is as follows:
A kind of personalized resource retrieval method, comprises the following steps:
Steps A Data Collection: comprise the data needing real time access be stored in database and buffer memory, and be stored in the large-scale access data in non real-time in distributed file system; The described real-time behavioral data needing the data of real time access to comprise user, described large-scale access data in non real-time comprises user characteristic data and product features data;
Step B engine calculates: recommended engine calculates according to the user behavior data of real-time collecting and obtains initial recommendation result;
Step C filtration treatment: applicating cooperation filters false code algorithm and carries out filtration treatment to the initial recommendation result obtained;
Step D rank is explained: the initial recommendation result through filtration treatment is merged according to certain weight or priority, sorted, and the recommendation then obtaining sequence prostatitis is explained and forms final recommendation results.
As technique scheme concrete preferably, step B engine calculates and comprises the following steps:
Step B1 structure based data carry out dimension-reduction treatment: by svd, Bayesian Clustering, probability latent semantic analysis and implicit Di Li Cray allocation process, the real-time behavioral data of user is utilized to set up a two-dimentional form about behavior and commodity, carry out matrix computations, the data after dimension-reduction treatment are as the input data of subsequent step;
Step B2 sets up the bigraph (bipartite graph) based on user and commodity: set up user's set A and commodity set B, and do not occur simultaneously between A and B, the input data that the node in A is provided by step B1 with B interior joint form the limit be connected, and forms bigraph (bipartite graph);
Bigraph (bipartite graph) is carried out the conversion of both direction by step B3, is mapped as hypergraph:
Form user hypergraph: producing a super limit because of the behavior to same commodity by each user is connected, and forms user's hypergraph, discloses the feature of user behavior pattern, improve the performance of commending system;
Form commodity hypergraphs: producing a super limit because of the behavior of same user by each commodity carrying out real-time behavior is connected, and forms commodity hypergraph.
Compared with prior art, implementation result of the present invention is as follows in the present invention:
The personalized resource retrieval method of one of the present invention, can not only the behavior of Accurate Prediction user, and can the visual field of extending user, and they may be interested to help user to find those, but so do not hold detectable thing; Meanwhile, a kind of personalized resource retrieval method of the present invention can also help businessman by those by the good buyer's guide that is buried in long-tail to may to they interested users.
Accompanying drawing explanation
Fig. 1 is the overall flow schematic diagram of a kind of personalized resource retrieval method of the present invention;
Fig. 2 is the schematic flow sheet that in the specific embodiment of the invention, structure based data carry out dimension-reduction treatment;
Fig. 3 is the schematic diagram based on the bigraph (bipartite graph) of user and commodity in the specific embodiment of the invention;
Fig. 4 is the schematic diagram of user's hypergraph in the specific embodiment of the invention;
Fig. 5 is the schematic diagram of commodity hypergraph in the specific embodiment of the invention.
Embodiment
Below in conjunction with specific embodiments content of the present invention is described.
As shown in Figure 1, be the personalized resource retrieval method structural representation of one of the present invention.The personalized resource retrieval method of one of the present invention, comprises the following steps:
One, Data Collection:
Comprise the data needing real time access be stored in database and buffer memory, and be stored in the large-scale access data in non real-time in distributed file system; The described real-time behavioral data needing the data of real time access to comprise user, described large-scale access data in non real-time comprises user characteristic data and product features data.
Personalized recommendation algorithm depends on user behavior data, and all there is various user behavior data in any one website.So how accessing these data is exactly the matter of utmost importance that commending system needs to solve.Following table 1 illustrates the typical user's behavioral data on an imaginary e-commerce website.As shown in table 1, from producing the user perspective of behavior, some behavior only has registered user to produce, and some behavior to be all users can produce.From scale, browse webpage, searching record scale all very large because all users of this behavior can produce, and average each user can produce these behaviors a lot.Buy, collection behavior scale is medium, because only have registered user could produce this behavior, but buying behavior is again the main behavior of electric business website, so they relative to comment larger, but it is much smaller relative to scale web page browsing behavior, finally remaining behavior is that the sub-fraction talent in registered user has, so scale can not be very large.From the angle of real time access, behaviors such as buying, collect, comment on, mark, share all needs real time access, as long as because user has had these behaviors, interface just needs embody, after such as user have purchased commodity, the individual of user buys in list the commodity that just should show user immediately and buy.And some behavior, behavior and the search behavior of such as browsing webpage do not need real time access.
Typical behaviour in table 1, platform
According to the scale of earlier data with the need of real time access, different behavioral datas will be stored in different media.In general, need the data of real time access to be stored in database and buffer memory, and large-scale access data is in non real-time stored in distributed file system.
Data can real time access extremely important in commending system because can the real-time of commending system depend on the new behavior taking user in real time.Only have the new behavior taking a large number of users fast, commending system can adapt to the current demand of user in real time, carries out real-time recommendation to user.
Two, engine calculates:
Recommended engine calculates according to the user behavior data of real-time collecting and obtains initial recommendation result.
The feature kind of user is very many, the recommendation task of commending system also has a variety of, if will the various characteristic sum tasks of handle in a system all consider as a whole, so system will be very complicated, and be difficult to the weight being configured different characteristic and task by configuration file easily.Therefore, commending system needs to be made up of multiple recommended engine, and each recommended engine is responsible for a category feature and a kind of task, and the result of recommended engine just merges according to certain weight or priority by the task of commending system, then sequence returns.
Particularly, comprise the following steps:
One), structure based data carry out dimension-reduction treatment: by svd, Bayesian Clustering, probability latent semantic analysis and implicit Di Li Cray allocation process, the real-time behavioral data of user is utilized to set up a two-dimentional form about behavior and commodity, carry out matrix computations, the data after dimension-reduction treatment are as the input data of subsequent step; Particularly, as shown in hereinafter Fig. 2.
Two) bigraph (bipartite graph) based on user and commodity, is set up: as shown in hereinafter Fig. 3, set up user's set A and commodity set B, do not occur simultaneously between A and B, the input data that the node in A is provided by step B1 with B interior joint form the limit be connected, and form bigraph (bipartite graph).
Three), bigraph (bipartite graph) is carried out the conversion of both direction, is mapped as hypergraph:
As shown in hereinafter Fig. 4, form user hypergraph: producing a super limit because of the behavior to same commodity by each user is connected, and forms user's hypergraph, discloses the feature of user behavior pattern, improve the performance of commending system; The research of social networks can being carried out, based on Ba Laiduo law, improving degree of dependence for bringing 20% user of 80% income.
As shown in hereinafter Fig. 5, form commodity hypergraphs: producing a super limit because of the behavior of same user by each commodity carrying out real-time behavior is connected, and forms commodity hypergraph; Can long tail effect research be carried out, carry out the marketing of much-sought-after item and unexpected winner commodity.
Three, filtration treatment:
Applicating cooperation filters false code algorithm and carries out filtration treatment to the initial recommendation result obtained.Collaborative filtering false code algorithm can be particularly:
input:k,data[n];
If E is all limits;
If T is the set on selected limit;
Initialize N=(V,{e});
Initialize T=[];
While (E<> []) & & (| T|<>n-1) // find cluster
{
E=E-(u,v);
u,v→M;
N=N-u-v;
If (in T, (u, v) forms ring), output M;
}
Four, rank is explained:
Initial recommendation result through filtration treatment is merged according to certain weight or priority, sorted, and the recommendation then obtaining sequence prostatitis is explained and forms final recommendation results.Comprise in recommendation results and recommend to explain, on the one hand can adding users to the degree of belief of recommendation results, user can be made to make feedback to recommendation results on the other hand, so that optimize recommendation results.
Above content is detailed description made for the present invention in conjunction with specific embodiments, can not assert that the present invention specifically implements to be only limitted to these explanations.For those skilled in the art, without departing from the inventive concept of the premise, some simple deduction or replace can also be made, all should be considered as belonging to the scope of protection of the invention.
Claims (2)
1. a personalized resource retrieval method, is characterized in that, comprises the following steps:
Steps A Data Collection: comprise the data needing real time access be stored in database and buffer memory, and be stored in the large-scale access data in non real-time in distributed file system; The described real-time behavioral data needing the data of real time access to comprise user, described large-scale access data in non real-time comprises user characteristic data and product features data;
Step B engine calculates: recommended engine calculates according to the user behavior data of real-time collecting and obtains initial recommendation result;
Step C filtration treatment: applicating cooperation filters false code algorithm and carries out filtration treatment to the initial recommendation result obtained;
Step D rank is explained: the initial recommendation result through filtration treatment is merged according to certain weight or priority, sorted, and the recommendation then obtaining sequence prostatitis is explained and forms final recommendation results.
2. a kind of personalized resource retrieval method as claimed in claim 1, is characterized in that, step B engine calculates and comprises the following steps:
Step B1 structure based data carry out dimension-reduction treatment: by svd, Bayesian Clustering, probability latent semantic analysis and implicit Di Li Cray allocation process, the real-time behavioral data of user is utilized to set up a two-dimentional form about behavior and commodity, carry out matrix computations, the data after dimension-reduction treatment are as the input data of subsequent step;
Step B2 sets up the bigraph (bipartite graph) based on user and commodity: set up user's set A and commodity set B, and do not occur simultaneously between A and B, the input data that the node in A is provided by step B1 with B interior joint form the limit be connected, and forms bigraph (bipartite graph);
Bigraph (bipartite graph) is carried out the conversion of both direction by step B3, is mapped as hypergraph:
Form user hypergraph: producing a super limit because of the behavior to same commodity by each user is connected, and forms user's hypergraph, discloses the feature of user behavior pattern, improve the performance of commending system;
Form commodity hypergraphs: producing a super limit because of the behavior of same user by each commodity carrying out real-time behavior is connected, and forms commodity hypergraph.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510402669.8A CN104992352A (en) | 2015-07-10 | 2015-07-10 | Individualized resource retrieval method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510402669.8A CN104992352A (en) | 2015-07-10 | 2015-07-10 | Individualized resource retrieval method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN104992352A true CN104992352A (en) | 2015-10-21 |
Family
ID=54304160
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510402669.8A Pending CN104992352A (en) | 2015-07-10 | 2015-07-10 | Individualized resource retrieval method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104992352A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105488711A (en) * | 2015-11-30 | 2016-04-13 | 北京北纬通信科技股份有限公司 | Multi-Agent based network direct sales system |
CN106021302A (en) * | 2016-05-04 | 2016-10-12 | 北京思特奇信息技术股份有限公司 | Intelligent recommendation technique based wireless music recommendation method and system |
CN107578326A (en) * | 2017-10-23 | 2018-01-12 | 青岛优米信息技术有限公司 | One kind recommends method and system |
CN109271491A (en) * | 2018-11-02 | 2019-01-25 | 合肥工业大学 | Cloud service recommendation method based on non-structured text information |
CN109934689A (en) * | 2019-03-22 | 2019-06-25 | 拉扎斯网络科技(上海)有限公司 | Target object ranking means of interpretation, device, electronic equipment and readable storage medium storing program for executing |
CN110020921A (en) * | 2019-04-09 | 2019-07-16 | 浩鲸云计算科技股份有限公司 | A kind of AI recommended engine is energized commodity marketing method |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104298702A (en) * | 2014-07-18 | 2015-01-21 | 合肥工业大学 | Method and system for electronic reading material recommendation on basis of social network information |
-
2015
- 2015-07-10 CN CN201510402669.8A patent/CN104992352A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104298702A (en) * | 2014-07-18 | 2015-01-21 | 合肥工业大学 | Method and system for electronic reading material recommendation on basis of social network information |
Non-Patent Citations (2)
Title |
---|
周子亮: "结合非负矩阵分解的推荐算法及框架研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
廖贵明: "个性化推荐引擎系统研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105488711A (en) * | 2015-11-30 | 2016-04-13 | 北京北纬通信科技股份有限公司 | Multi-Agent based network direct sales system |
CN106021302A (en) * | 2016-05-04 | 2016-10-12 | 北京思特奇信息技术股份有限公司 | Intelligent recommendation technique based wireless music recommendation method and system |
CN107578326A (en) * | 2017-10-23 | 2018-01-12 | 青岛优米信息技术有限公司 | One kind recommends method and system |
CN109271491A (en) * | 2018-11-02 | 2019-01-25 | 合肥工业大学 | Cloud service recommendation method based on non-structured text information |
CN109271491B (en) * | 2018-11-02 | 2021-09-28 | 合肥工业大学 | Cloud service recommendation method based on unstructured text information |
CN109934689A (en) * | 2019-03-22 | 2019-06-25 | 拉扎斯网络科技(上海)有限公司 | Target object ranking means of interpretation, device, electronic equipment and readable storage medium storing program for executing |
CN110020921A (en) * | 2019-04-09 | 2019-07-16 | 浩鲸云计算科技股份有限公司 | A kind of AI recommended engine is energized commodity marketing method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10095771B1 (en) | Clustering and recommending items based upon keyword analysis | |
JP6356744B2 (en) | Method and system for displaying cross-website information | |
CN102789462B (en) | A kind of item recommendation method and system | |
CN104992352A (en) | Individualized resource retrieval method | |
White et al. | Predicting user interests from contextual information | |
CN103019550B (en) | The real-time exhibiting method of content association and system | |
CN102054003B (en) | Methods and systems for recommending network information and creating network resource index | |
CN103886487A (en) | Individualized recommendation method and system based on distributed B2B platform | |
CN106295832A (en) | Product information method for pushing and device | |
CN102509233A (en) | User online action information-based recommendation method | |
CN102479366A (en) | Commodity recommending method and system | |
CN103886090A (en) | Content recommendation method and device based on user favorites | |
CN111159341B (en) | Information recommendation method and device based on user investment and financial management preference | |
CN103345698A (en) | Personalized recommendation method based on cloud processing mode and applied in e-business environment | |
CN103699603A (en) | Information recommendation method and system based on user behaviors | |
US20160196579A1 (en) | Dynamic deep links based on user activity of a particular user | |
CN108073667A (en) | Method for generating user browsing attributes, and non-transitory computer readable medium | |
CN105894310A (en) | Personalized recommendation method | |
Dias et al. | Automating the extraction of static content and dynamic behaviour from e-commerce websites | |
CN102650991A (en) | Commodity recommending method and system both based on user preference | |
Lian | The construction of personalized Web page recommendation system in e-commerce | |
Wei et al. | A study of customer behaviour through web mining | |
CN103870448A (en) | Method and method for recommending data | |
Chevalier et al. | Information retrieval and folksonomies together for recommender systems | |
CN110827044A (en) | Method and device for extracting user interest mode |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20151021 |
|
WD01 | Invention patent application deemed withdrawn after publication |