CN108830689A - Item recommendation method, device, server and storage medium - Google Patents
Item recommendation method, device, server and storage medium Download PDFInfo
- Publication number
- CN108830689A CN108830689A CN201810607170.4A CN201810607170A CN108830689A CN 108830689 A CN108830689 A CN 108830689A CN 201810607170 A CN201810607170 A CN 201810607170A CN 108830689 A CN108830689 A CN 108830689A
- Authority
- CN
- China
- Prior art keywords
- article
- user
- data
- collection
- historical 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
Classifications
-
- 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
Abstract
The embodiment of the invention discloses a kind of item recommendation method, device, server and storage medium, this method includes：Collection is recalled according to article building original goods associated in the same day behavioral data of user；It is recalled according to historical behavior data and original goods and the article for the including extension original goods is concentrated to recall collection, and determine that the article after extension recalls the characteristic for concentrating each article；According to the user characteristic data that the article after extension is recalled the characteristic for concentrating each article and extracted from the historical behavior data, article to be recommended is determined for user.The embodiment of the present invention is to recommend the main determination basis of article by the current behavior data with user, the article for meeting user's near real-time demand can be recommended user in time, the viscosity of article and user is recommended in enhancing, is improved the conversion efficiency of article and the activity of the user and is chosen efficiency.
Description
Technical field
The present embodiments relate to technical field of data processing more particularly to a kind of item recommendation method, device, servers
And storage medium.
Background technique
With the fast development of Internet technology and electronic commerce affair, more and more users carry out object using network
Product facilitate selection and purchase.
Due to almost covering the article of all kinds in electric business website, how personalized article is provided for user
Recommendation becomes particularly important.Off-line processing system is used in the prior art, is usually come using day or hour as period unit to user
Recommendation article handled and updated.Its implementation is using the historical behavior data of user as foundation, by going through to user
History behavioral data is analyzed, and is updated periodically article to be recommended and is pushed to user.
However, the prior art is lower to the performance requirement of data processing system, method is realized simply, can not will meet user
The article of real-time requirement recommends user, so that the article recommended does not have timeliness.Reduce recommend article practicability and
Recommendability, while reducing the conversion efficiency of article and the activity of the user and choosing efficiency.
Summary of the invention
The embodiment of the invention provides a kind of item recommendation method, device, server and storage mediums, can will meet use
The article of family real-time requirement recommends user in time, improves the conversion efficiency of article and the activity of the user and chooses effect
Rate.
In a first aspect, the embodiment of the invention provides a kind of item recommendation methods, including：
Collection is recalled according to article building original goods associated in the same day behavioral data of user；
It is recalled according to historical behavior data and original goods and the article for the including extension original goods is concentrated to recall collection,
And determine that the article after extension recalls the characteristic for concentrating each article；
It recalls the characteristic for concentrating each article according to the article after extension and is mentioned from the historical behavior data
The user characteristic data taken determines article to be recommended for user.
Second aspect, the embodiment of the invention provides a kind of article recommendation apparatus, including：
Article recalls collection building module, for constructing original goods according to associated article in the same day behavioral data of user
Recall collection；
Article recalls collection expansion module, for recalling the article concentrated and include according to historical behavior data and original goods
It extends the original goods and recalls collection, and determine that the article after extension recalls the characteristic for concentrating each article；
Article determining module to be recommended, for according to extension after article recall concentrate each article characteristic and
The user characteristic data extracted from the historical behavior data determines article to be recommended for user.
The third aspect, the embodiment of the invention provides a kind of servers, including：
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processing
Device realizes item recommendation method described in any embodiment of that present invention.
Fourth aspect, the embodiment of the invention provides a kind of computer readable storage mediums, are stored thereon with computer journey
Sequence realizes item recommendation method described in any embodiment of that present invention when the program is executed by processor.
The embodiment of the present invention is by the real-time acquisition to user behavior, according to same day row caused by user to current time
Original goods, which are constructed, for article associated in data recalls collection；And original goods are recalled according to updated historical behavior data
The each article concentrated is extended, and the article after determining extension recalls the article characteristics data for concentrating each article, thus really
Determined that user is current or today interested to article and article characteristic；Finally according to updated prediction model, object
Product characteristic and the user characteristic data extracted from updated historical behavior data, determine article to be recommended.This hair
Bright embodiment is to recommend the main determination basis of article by the current behavior data with user, can will meet user's near real-time
The article of demand recommends user in time, and the viscosity of article and user is recommended in enhancing, improve article conversion efficiency and
The activity of the user and choose efficiency.
Detailed description of the invention
Fig. 1 is a kind of flow chart for item recommendation method that the embodiment of the present invention one provides；
Fig. 2 is a kind of flow chart of item recommendation method provided by Embodiment 2 of the present invention；
Fig. 3 is a kind of structural schematic diagram for article recommendation apparatus that the embodiment of the present invention three provides；
Fig. 4 is a kind of structural schematic diagram for server that the embodiment of the present invention four provides.
Specific embodiment
The embodiment of the present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this
Locate described specific embodiment and is used only for explaining the embodiment of the present invention, rather than limitation of the invention.It further needs exist for
Bright, only parts related to embodiments of the present invention are shown for ease of description, in attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is a kind of flow chart for item recommendation method that the embodiment of the present invention one provides, and the present embodiment is applicable to use
The case where family is by internet hunt and selection article, this method can be executed by a kind of article recommendation apparatus.This method is specific
Include the following steps：
Step 110 recalls collection according to article building original goods associated in the same day behavioral data of user.
In the specific embodiment of the invention, in real time extract the real-time behavior of user from User action log,
Parsing and cleaning, extract user information, Item Information and contextual information associated by the real-time behavior.Wherein, Yong Huxin
Breath may include the personal user informations such as age and the gender of user；Item Information may include category and the date of manufacture of article
And other items attribute information；Contextual information refers to the mould in time, place and associated webpage that the real-time behavior occurs
The behaviors satellite information such as block.And real-time behavior and all associated information are stored in the storage mediums such as database in real time
In, i.e., only the behavioral data at the current time on the same day is stored, analyze user demand for subsequent near real-time, is user
It determines and article is recommended to provide foundation.
The present embodiment periodically can extract, clean and store use using the even more small time interval of second rank
The real-time behavior at family, such as extraction etc. of the primary behavior in real time of progress per second operate, which can be user and pass through mutually
The behaviors such as browsing, collection, releasing collection, addition shopping cart, rejecting shopping cart, sharing, purchase and reservation that networking carries out.This
Embodiment at set time intervals from the storage medium for being stored with real-time behavioral data obtain the same day in cut-off to it is current when
It carves behavioral data caused by user and is used as same day behavioral data, it is to be understood that same day behavioral data refers to cut-off to working as
All behaviors occurred on the day of the preceding moment, therefore over time, same day behavioral data is also to vary constantly.
Since the real-time behavior of user reflects the current specific and strong interest demand of user, time of the application to set
The same day behavioral data of interval acquiring user can sufficiently analyze the current actual demand of user or hobby tendency.It is worth note
Meaning, the present embodiment can periodically obtain the same day behavioral data of user using the time interval of minute rank, into
And in the case where guaranteeing that system operates normally, analyze near real-time the actual demand of user.
Since the behavior of user intuitively reflects the demand or hobby tendency of user.For example, user is for same article
The browsing of long period and the collection behavior of user user can be indicated to the serious hope of associated article, demand or attention；With
The releasing collection at family or the behavior for rejecting shopping cart can indicate that user is temporarily without demand or indifferent to associated article.Therefore originally
Embodiment can analyze each behavior in same day behavioral data, thus according to each behavior in same day behavioral data attribute with
And article associated by each behavior, it will indicate that user's current demand or interested article are configured to original goods and recall collection.
Illustratively, it is assumed that the extraction for carrying out primary behavior in real time, cleaning and storage operation per second carry out primary per minute
The acquisition of same day behavioral data operates.Know cut-off to currently, user's first has carried out collection behavior to article A, to article B progress
Buying behavior, and reject to article C the behavior of shopping cart.And then the attribute according to each behavior in same day behavioral data
And associated article, article A and the B original goods for constituting user's first can be recalled into collection.
Step 120 recalls according to historical behavior data and original goods and the article for including extension original goods is concentrated to call together
Collection is returned, and determines that the article after extension recalls the characteristic for concentrating each article.
In the specific embodiment of the invention, historical behavior data generated all behavior numbers before referring to user on the day of
According to.The present embodiment can carry out periodically to update with the time interval of day or hour the historical behavior data of user, then history
Behavioral data is generated all behavioral datas before this update cycle, correspondingly, each in same day behavioral data
Behavior is the recent behavioral data within this update cycle.Same day behavioral data is the linking of historical behavior data, the two nothing
Inclusion relation.Same day behavioral data determines that article is recalled and concentrates user required or interested article, historical behavior data
Determine that article recalls the extension and the wherein prediction of each article clicking rate of collection.It is huge due to historical behavior data,
The present embodiment preferably periodically to update the historical behavior data of user using day as time interval, if in historical behavior data not
Comprising user behavior data caused by the previous day, then user behavior data caused by the previous day is added to historical behavior number
In, to update historical behavior data.Carry out more new article clicking rate prediction model simultaneously with updated historical behavior data.Its
In, article clicking rate prediction model is the model predicted the clicking rate of article pre-established, with user's row
For variation and change, therefore synchronized update article clicking rate prediction model is needed after updating historical behavior data.If
Historical behavior data can then call directly historical behavior data and this part of article clicking rate prediction model is offline without updating
Data.
The present embodiment increases time as much as possible to increase the commodity amount that the original goods of each user recall concentration
Commodity are selected, the category of article, the similarity of article and the updated history concentrated and include can be recalled according to original goods
Associated article in behavioral data, to determine that original goods recall the associated article of each article of concentration；It simultaneously can also be according to
Fast-selling article is determined according to updated historical behavior data.To extend original goods according to associated article and/or fast-selling article
Collection is recalled, the article after being extended recalls collection.The characteristic of article can be the history under off-line state according to user
Behavioral data calculates, such as the features such as the scope of application of article.And then it can be according to the related compounds in historical behavior data
Product concentrate the characteristic of each article to determine that the article after extension is recalled.
Illustratively, it is assumed that the update cycle of historical behavior data is 1 day, then before historical behavior data refer to 1 day
All behavioral datas caused by user.Therefore in query history behavioral data each behavior time of origin, if historical behavior number
Do not include user behavior data caused by the previous day in, then user behavior data caused by the previous day is added to history
In behavioral data, to update historical behavior data, and more new article clicking rate prediction model.Correspondingly, same day behavioral data is
The behavioral data occurred on the day of ending to current time.In the examples described above, it includes article A and object that original goods, which recall collection,
Product B.According to updated historical behavior data, collection is recalled to original goods and is extended.If article A is mobile phone a, and user's first
Historical behavior data in include mobile phone shell, and then can be using mobile phone shell as the associated article of article A.It can also be according to all
The historical behavior data of user, if the mobile phone b different from mobile phone a color is higher by public favorable rating, such as its high conversion rate
In certain threshold value, it is determined that mobile phone b is current fast-selling article.Therefore the article after extending recalls collection and includes mobile phone a, article
B, mobile phone shell and mobile phone b.And the characteristic for concentrating each article is recalled according to the article after the determining extension of historical behavior data
According to.
Step 130 recalls the characteristic for concentrating each article according to the article after extension and from historical behavior data
The user characteristic data of middle extraction determines article to be recommended for user.
In the specific embodiment of the invention, user characteristic data is the historical behavior data under off-line state according to user
It calculates, at least may include hobby feature and demand characteristic of user information and user etc..Article after expansion
It recalls on the article i.e. basis of the article of user's current interest for collecting included, is predicted using updated article clicking rate
Model, according to article characteristics data and user characteristic data, the article after prediction extension recalls the clicking rate for concentrating each article,
And the article after extension is recalled, each article is concentrated to carry out duplicate removal and filtration treatment, to simplify the selection model that can recommend article
It encloses.Finally according to the prediction clicking rate of each article, to treated, article is ranked up, and be will click on the higher article of rate and is determined
For the article to be recommended of user.
It is worth noting that, the article after extension, which recalls concentration, may be simultaneously present similar article, such as category is identical
But a certain attribute is different, but since the fine feature of article certainly exists influence to the selection of user, preferably preferential to carry out
The prediction of article clicking rate, so that influence of the article interested to user to clicking rate is adequately taken into account, it is pre- in clicking rate
The duplicate removal and filtration treatment of article are carried out after survey again.And the sequence processing of the duplicate removal and filtration treatment and article of article is successive suitable
The requirement that sequence is not necessarily to, only after duplicate removal and filtration treatment, the efficiency of sequence can be somewhat higher.
Illustratively, determine that user characteristic data and article recall the feature for concentrating each article according to historical behavior data
Data are input in updated clicking rate prediction model, determine the prediction clicking rate of each article.In the examples described above, due to
Mobile phone a is identical with the category of mobile phone b but color is different, therefore carries out duplicate removal processing, and the clicking rate of comprehensive mobile phone a and mobile phone b
Obtain the prediction clicking rate of mobile phone itself.Mobile phone, article B and the respective prediction of mobile phone shell after final foundation article is simplified are clicked
Rate is ranked up, and the higher article of clicking rate will be predicted as article to be recommended.
In practical applications, since the current behavior data of user are limited, original goods recall the article number of concentration
It measures less, and after recalling collection extension, associated article as much as possible and/or fast-selling article can be obtained, number of articles rises
Amplitude is larger.And then article to be recommended really timing, in order to will be best suitable for user's current demand and polymerization sort out after to the greatest extent may be used
The article of energy multiclass recommends user, reduces repeated article or ineffectivity article to the occupancy for recommending item name volume, needs
It carries out simplifying processing as duplicate removal and filtering after carrying out clicking rate prediction using the information of each article, ensure that and fully consider respectively
While influence of the Item Information to clicking rate, article can be recommended by having simplified, so that user obtains a greater variety of articles and recommends,
Improve the practicability and user experience of each article.
The technical solution of the present embodiment, by the real-time acquisition to user behavior, according to produced by user to current time
Same day behavioral data in associated article building original goods recall collection；And according to updated historical behavior data to original
Each article that article recalls concentration is extended, and the article after determining extension recalls the article characteristics number for concentrating each article
According to, so that it is determined that user is current or today interested to article and article characteristic；Finally according to updated pre-
Model, article characteristics data and the user characteristic data extracted from updated historical behavior data are surveyed, is determined to be recommended
Article.The embodiment of the present invention is to recommend the main determination basis of article by the current behavior data with user, can will be met
The article of user's near real-time demand recommends user in time, and enhancing recommends the viscosity of article and user, improves the effective of article
Conversion ratio and the activity of the user and choose efficiency.
Embodiment two
The present embodiment on the basis of the above embodiment 1, provides a preferred embodiment of item recommendation method,
It being capable of extract real-time real-time behavioral data and the prediction to real-time behavioral data associated article progress clicking rate.Fig. 2 is this
The flow chart for a kind of item recommendation method that inventive embodiments two provide, as shown in Fig. 2, this method includes step in detail below：
Step 201, the real-time behavioral data for extracting and storing user.
In the specific embodiment of the invention, it can periodically be mentioned using the even more small time interval of second rank
Take, parse and clean the real-time behavior of user.The real-time behavior can be browsing, the collection, solution that user is carried out by internet
Except collection, the behaviors such as shopping cart, rejecting shopping cart, sharing, purchase and reservation are added.It extracts associated by the real-time behavior
User information, Item Information and contextual information.Wherein, user information may include the personal users such as age and the gender of user
Information；Item Information may include category and the date of manufacture of article and other items attribute information；Contextual information refers to this in real time
The behaviors satellite informations such as the module in time, place and associated webpage that behavior occurs.And in real time by real-time behavior
And all associated information are stored in the storage mediums such as database.
Step 202 obtains user to working as from the storage medium for being stored with real-time behavioral data at set time intervals
Same day behavioral data caused by the preceding moment；The time interval set is the periodical intervals of minute rank.
In the specific embodiment of the invention, the present embodiment is at set time intervals from being stored with depositing for real-time behavioral data
Cut-off to behavioral data caused by current time is used as same day behavioral data on the day of obtaining user in storage media, it is possible to understand that
It is that same day behavioral data refers to all behaviors occurred on the day of cut-off to current time, therefore over time, the same day
Behavioral data is also to vary constantly.Since the real-time behavior of user reflects the current specific and strong interest of user
Demand, therefore the application obtains the same day behavioral data of user at set time intervals, and it is current can sufficiently to analyze user
Actual demand or hobby tendency.It is worth noting that, the present embodiment can be using the time interval of minute rank come periodically
Acquisition user same day behavioral data, and then guarantee system operate normally in the case where, analyze user's near real-time
Actual demand.
Article associated by step 203, the attribute according to each behavior in same day behavioral data and each behavior, constructs original
Article recalls collection.
In the specific embodiment of the invention, since the behavior of user intuitively reflects the demand or hobby of user,
Each behavior in same day behavioral data can be analyzed, thus attribute and each row according to each behavior in same day behavioral data
For associated article, it will indicate that user's current demand or interested article group become original goods and recall collection.
If not including user behavior data caused by the previous day in step 204, historical behavior data, by the previous day institute
The user behavior data of generation is added in historical behavior data, to update historical behavior data.
In the specific embodiment of the invention, it can come to update going through for user periodically with the time interval of day or hour
History behavioral data, then generated all behavioral datas before historical behavior data are this update cycle, correspondingly, the same day
Each behavior in behavioral data is the recent behavioral data within this update cycle.Same day behavioral data is historical behavior number
According to linking, the two is without inclusion relation.Same day behavioral data determines that article is recalled and concentrates user required or interested object
Product, historical behavior data determine that article recalls the extension and the wherein prediction of each article clicking rate of collection.Due to historical behavior
Data it is huge, therefore the present embodiment preferably comes periodically to update the historical behavior data of user as time interval using day, if
User behavior data caused by the previous day is not included in historical behavior data, then by user behavior data caused by the previous day
It is added in historical behavior data, to update historical behavior data.Carry out more new article simultaneously with updated historical behavior data
Clicking rate prediction model.If historical behavior data can call directly historical behavior data and article clicking rate without updating
This part off-line data of prediction model.
Associated object in step 205, the category of foundation article, the similarity of article and updated historical behavior data
Product determine that original goods recall the associated article of each article of concentration；And/or it is determined according to updated historical behavior data
Fast-selling article；Collection is recalled according to associated article and/or fast-selling article extension original goods, the article after being extended recalls collection.
In the specific embodiment of the invention, in order to increase the commodity amount that the original goods of each user recall concentration, increase
Add candidate commodity as much as possible, can be recalled according to original goods the category for concentrating the article for including, the similarity of article with
And associated article in updated historical behavior data, determine that original goods recall the associated article of each article of concentration.
Illustratively, it is assumed that original goods recall concentration and include brand A mobile phone, then can be associated with out according to the category of article with for the moment
The brand B mobile phone of phase listing is selected as associated article for user.
The present embodiment can also determine fast-selling article according to updated historical behavior data, further be expanded with fast-selling article
It fills article and recalls collection.Heat can be determined according to various dimensions information such as the conversion ratio of article, ascendant trend, favorable comment situation and sales volumes
Sell commodity.Wherein, conversion ratio refers to that article browses to the ratio successfully bought by being clicked by user, can be by nearly 7 days
The number of users that places an order with nearly 7 days in the ratio of touching quantity of user determine the conversion ratio of article；Ascendant trend refers to closely
In the short period of phase in the touching quantity and long term time of user the touching quantity of user ratio, nearly 7 days points can be passed through
The ratio of quantity and nearly 28 days touching quantities is hit to determine the ascendant trend of article.The calculation formula of conversion ratio and ascendant trend
In, the specific number of days that calculates can adjust according to the actual situation.According to the historical behavior data of whole users, above-mentioned calculating is utilized
Formula and to the evaluation of article and the analysis of sales volume situation, that is, can determine hot item.
The present embodiment extends the original goods according to associated article and/or fast-selling article and recalls collection, after being extended
Article recalls collection.Article can be recalled through the above way and the article of limited quantity is concentrated to expand to thousand ranks, for subsequent place
Reason therefrom it is excellent it is middle select excellent and be that user determines several or more than ten recommendation article, reach the recommendation effect in thousand people, thousand face, mention
Rise the clicking rate and buying rate of the article recommended.
Step 206 determines that the article after extension recalls the characteristic for concentrating each article.
In the specific embodiment of the invention, the characteristic of article can be the history row under off-line state according to user
It is calculated for data, such as the features such as the scope of application of article.And then it can be according to the associated article in historical behavior data
The characteristic of each article is concentrated to determine that the article after extension is recalled.
Step 207, foundation article characteristics data and user characteristic data, the article after prediction extension, which is recalled, concentrates each object
The clicking rate of product.
In the specific embodiment of the invention, user characteristic data is the historical behavior data under off-line state according to user
It calculates, at least may include hobby feature and demand characteristic of user information and user etc..Article after expansion
It recalls on the article i.e. basis of the article of user's current interest for collecting included, is predicted using updated article clicking rate
Model, according to article characteristics data and user characteristic data, the article after prediction extension recalls the clicking rate for concentrating each article.
Any model that can predict article clicking rate can be using in this present embodiment in the prior art.
Step 208 recalls the article after extension and each article is concentrated to carry out duplicate removal and filtration treatment.
In the specific embodiment of the invention, the article after extension, which recalls concentration, may be simultaneously present similar article, such as
Category is identical but a certain attribute is different, but since the fine feature of article certainly exists influence to the selection of user, excellent
The prediction for first carrying out article clicking rate, after fully considering article interested to user to the influence of clicking rate, then carries out article
Duplicate removal and filtration treatment, to simplify the range of choice that can recommend article.Duplicate removal, which refers to, removes ware according to Item Information
It removes, such as by same brand same model but different colours article is removed, while article and the removal of reservation can be integrated
Article prediction clicking rate, the clicking rate as the category article.Filtering is generally according to certain business rule, according to user
Whether characteristic and product features data meet business rule, and the non-conforming article for recalling concentration is removed.For example, according to
The gender label of user characteristic data and article characteristics data is filtered article, not by the inconsistent article, that is, user of gender
Applicable article filters out；In another example by being filtered out currently without the article of inventory.
Step 209, according to the prediction clicking rate of each article, to treated, article is ranked up.
In the specific embodiment of the invention, sequence is that the article after extension is recalled to concentration after duplicate removal and filtration treatment
Remaining article is ranked up according to respective prediction clicking rate.Use is preferentially exposed to for the ease of will click on the biggish article of rate
Family can be ranked up according to predicting that clicking rate is descending.
Step 210 determines article to be recommended according to ranking results.
In the specific embodiment of the invention, the article in ranking results is all to predict by clicking rate and simplify that treated
Article, and then direct basis ranking results will click on the article to be recommended that the biggish article of rate is determined as user.
The technical solution of the present embodiment, by the extract real-time to real-time behavior, behavioral data on the day of timing acquisition, thus
Determine that original goods recall collection；And collection is recalled to original goods according to updated historical behavior data and is extended, it determines and expands
Article after exhibition recalls the article characteristics data and user characteristic data for concentrating each article；To utilize updated article
Clicking rate prediction model carries out clicking rate prediction to each article according to article characteristics data and user characteristic data；Finally according to
According to duplicate removal, filtering and sequence, treated that article determines article to be recommended.The embodiment of the present invention passes through with the current behavior of user
Data are to recommend the main determination basis of article, the article for meeting user's near real-time demand can be recommended user in time,
The viscosity of article and user is recommended in enhancing, is improved the conversion efficiency of article and the activity of the user and is chosen efficiency.
Embodiment three
Fig. 3 is a kind of structural schematic diagram for article recommendation apparatus that the embodiment of the present invention three provides, and the present embodiment is applicable
Pass through the case where internet hunt is with article is selected in user, which can realize that article described in any embodiment of that present invention pushes away
Recommend method.The device specifically includes：
Article recalls collection building module 310, for original according to article building associated in the same day behavioral data of user
Article recalls collection；
Article recalls collection expansion module 320, includes for recalling concentration according to historical behavior data and original goods
Article extends the original goods and recalls collection, and determines that the article after extension recalls the characteristic for concentrating each article；
Article determining module 330 to be recommended, for recalling the characteristic for concentrating each article according to the article after extension
And the user characteristic data extracted from the historical behavior data, article to be recommended is determined for user.
Preferably, the article recalls collection building module 310, including：
Real-time behavioral data extraction unit, for extracting and storing the real-time behavioral data of user；
Same day behavioral data acquiring unit, for being situated between at set time intervals from the storage for being stored with real-time behavioral data
Same day behavioral data caused by user to current time is obtained in matter；
Article recalls collection construction unit, for according to each behavior in the same day behavioral data attribute and each behavior institute
Associated article, building original goods recall collection.
Preferably, the time interval set is the periodical intervals of minute rank.
Further, described device further includes：
Off-line data update module 340, if for not including user behavior caused by the previous day in historical behavior data
User behavior data caused by described the previous day is then added in historical behavior data by data, to update historical behavior number
According to.
Preferably, the off-line data update module 340 is also used to according to updated historical behavior data more new article
Clicking rate prediction model.
Preferably, the article recalls collection expansion module 320, including：
Associated article determination unit, for category, the similarity of article and the updated historical behavior according to article
Associated article in data determines that the original goods recall the associated article of each article of concentration；
Fast-selling article determination unit, for determining fast-selling article according to updated historical behavior data；
Article recalls collection expanding element, for described original according to the associated article and/or the fast-selling article extension
Article recalls collection, and the article after being extended recalls collection.
Preferably, the article determining module 330 to be recommended, including：
Article clicking rate predicting unit, for the object according to article characteristics data and user characteristic data, after prediction extension
Product recall the clicking rate for concentrating each article；
Article simplifies processing unit, concentrates each article to carry out at duplicate removal and filtering for recalling to the article after extension
Reason；
Article sequencing unit, for the prediction clicking rate according to each article, to treated, article is ranked up；
Article determination unit to be recommended, for determining article to be recommended according to ranking results.
The technical solution of the present embodiment, by the mutual cooperation between each functional module, realize historical behavior data and
The update of prediction model, the extraction of real-time behavioral data, cleaning and storage, article recall the building and extension, article characteristics of collection
The determination of data, the determination of user characteristic data, the prediction of article clicking rate, duplicate removal, filtering and the sequence of article and wait push away
Recommend the functions such as the determination of article.The embodiment of the present invention by the current behavior data with user be recommend article main determination according to
According to the article for meeting user's near real-time demand capable of recommending in time to user, the viscosity of article and user is recommended in enhancing, is mentioned
The conversion efficiency and the activity of the user of high article and choose efficiency.
Example IV
Fig. 4 is a kind of structural schematic diagram for server that the embodiment of the present invention four provides.As shown in figure 4, the service utensil
Body includes：One or more processors 410, in Fig. 4 by taking a processor 410 as an example；Memory 420, for store one or
Multiple programs, when one or more programs are executed by one or more processors 410, so that one or more processors 410 are real
Item recommendation method described in existing any embodiment of that present invention.Processor 410 and memory 420 can pass through bus or its other party
Formula connects, in Fig. 4 for being connected by bus.
It is executable to can be used for storing software program, computer as a kind of computer readable storage medium for memory 420
Program and module, if the corresponding program instruction of the item recommendation method in the embodiment of the present invention is (for example, behavioral data in real time
Extraction, cleaning and storage and the prediction and sequence of article clicking rate).Processor 410 is stored in memory 420 by operation
Software program, instruction and module realized above-mentioned thereby executing the various function application and data processing of server
Item recommendation method.
Memory 420 can mainly include storing program area and storage data area, wherein storing program area can store operation system
Application program needed for system, at least one function；Storage data area, which can be stored, uses created data etc. according to server.
It can also include nonvolatile memory in addition, memory 420 may include high-speed random access memory, for example, at least one
A disk memory, flush memory device or other non-volatile solid state memory parts.In some instances, memory 420 can be into
One step includes the memory remotely located relative to processor 410, these remote memories can pass through network connection to service
Device.The example of above-mentioned network includes but is not limited to internet, intranet, local area network, mobile radio communication and combinations thereof.
Embodiment five
The embodiment of the present invention five also provides a kind of computer readable storage medium, be stored thereon with computer program (or
For computer executable instructions), for executing a kind of item recommendation method when which is executed by processor, this method includes：
Collection is recalled according to article building original goods associated in the same day behavioral data of user；
It is recalled according to historical behavior data and original goods and the article for the including extension original goods is concentrated to recall collection,
And determine that the article after extension recalls the characteristic for concentrating each article；
It recalls the characteristic for concentrating each article according to the article after extension and is mentioned from the historical behavior data
The user characteristic data taken determines article to be recommended for user.
Certainly, a kind of computer readable storage medium provided by the embodiment of the present invention, computer executable instructions are not
It is limited to method operation as described above, the correlation in item recommendation method provided by any embodiment of the invention can also be performed
Operation.
By the description above with respect to embodiment, it is apparent to those skilled in the art that, the present invention
Embodiment can be realized by software and required common hardware, naturally it is also possible to by hardware realization, but in many cases before
Person is more preferably embodiment.Based on this understanding, the technical solution of the embodiment of the present invention is substantially in other words to existing skill
The part that art contributes can be embodied in the form of software products, which can store in computer
Floppy disk, read-only memory (Read-Only Memory, ROM), random access memory in readable storage medium storing program for executing, such as computer
(Random Access Memory, RAM), flash memory (FLASH), hard disk or CD etc., including some instructions are used so that one
Computer equipment (can be personal computer, server or the network equipment etc.) executes described in each embodiment of the present invention
Method.
It is worth noting that, included each unit and module are only according to function in the embodiment of above-mentioned searcher
Energy logic is divided, but is not limited to the above division, as long as corresponding functions can be realized；In addition, each function
The specific name of energy unit is also only for convenience of distinguishing each other, the protection scope being not intended to restrict the invention.
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being implemented by above embodiments to the present invention
Example is described in further detail, but the embodiment of the present invention is not limited only to above embodiments, is not departing from structure of the present invention
It can also include more other equivalent embodiments in the case where think of, and the scope of the present invention is determined by scope of the appended claims
It is fixed.
Claims (10)
1. a kind of item recommendation method, which is characterized in that including：
Collection is recalled according to article building original goods associated in the same day behavioral data of user；
It is recalled according to historical behavior data and original goods and the article for the including extension original goods is concentrated to recall collection, and really
Article after fixed extension recalls the characteristic for concentrating each article；
The characteristic for concentrating each article is recalled according to the article after extension and is extracted from the historical behavior data
User characteristic data determines article to be recommended for user.
2. the method according to claim 1, wherein associated object in the same day behavioral data according to user
Product building original goods recall collection, including：
Extract and store the real-time behavioral data of user；
It is obtained produced by user to current time from the storage medium for being stored with real-time behavioral data at set time intervals
Same day behavioral data；
According to article associated by the attribute of each behavior in the same day behavioral data and each behavior, constructs original goods and recall
Collection.
3. the method according to claim 1, wherein being called together described according to historical behavior data and original goods
It returns before concentrating the article for the including extension original goods to recall collection, further includes：
If not including user behavior data caused by the previous day in historical behavior data, will be used caused by described the previous day
Family behavioral data is added in historical behavior data, to update historical behavior data.
4. the method according to claim 1, wherein described recall according to historical behavior data and original goods
The each article concentrated extends the original goods and recalls collection, including：
According to associated article in the category of article, the similarity of article and updated historical behavior data, determine described in
Original goods recall the associated article of each article of concentration；And/or
Fast-selling article is determined according to updated historical behavior data；
It extends the original goods according to the associated article and/or the fast-selling article and recalls collection, the article after being extended
Recall collection.
5. the method according to claim 1, wherein the article according to after extension recalls and concentrates each article
Characteristic and the user characteristic data extracted from the historical behavior data, determine article to be recommended for user, wrap
It includes：
According to article characteristics data and user characteristic data, the article after prediction extension recalls the clicking rate for concentrating each article；
Recalling to the article after extension concentrates each article to carry out duplicate removal and filtration treatment；
According to the prediction clicking rate of each article, to treated, article is ranked up；
Article to be recommended is determined according to ranking results.
6. according to the method described in claim 2, it is characterized in that, the time interval set is the periodicity of minute rank
Time interval.
7. a kind of article recommendation apparatus, which is characterized in that including：
Article recalls collection building module, for recalling according to article building original goods associated in the same day behavioral data of user
Collection；
Article recalls collection expansion module, concentrates the article for including extension for recalling according to historical behavior data and original goods
The original goods recall collection, and determine that the article after extension recalls the characteristic for concentrating each article；
Article determining module to be recommended concentrates the characteristic of each article and from institute for recalling according to the article after extension
The user characteristic data extracted in historical behavior data is stated, determines article to be recommended for user.
8. device according to claim 7, which is characterized in that the article recalls collection building module, including：
Real-time behavioral data extraction unit, for extracting and storing the real-time behavioral data of user；
Same day behavioral data acquiring unit, at set time intervals from the storage medium for being stored with real-time behavioral data
Obtain same day behavioral data caused by user to current time；
Article recalls collection construction unit, for according to each behavior in the same day behavioral data attribute and each behavior associated by
Article, building original goods recall collection.
9. a kind of server, which is characterized in that including：
One or more processors；
Memory, for storing one or more programs；
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Now such as item recommendation method described in any one of claims 1 to 6.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Such as item recommendation method described in any one of claims 1 to 6 is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810607170.4A CN108830689A (en) | 2018-06-13 | 2018-06-13 | Item recommendation method, device, server and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810607170.4A CN108830689A (en) | 2018-06-13 | 2018-06-13 | Item recommendation method, device, server and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108830689A true CN108830689A (en) | 2018-11-16 |
Family
ID=64144991
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810607170.4A Pending CN108830689A (en) | 2018-06-13 | 2018-06-13 | Item recommendation method, device, server and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830689A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110602532A (en) * | 2019-09-24 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Entity article recommendation method, device, server and storage medium |
| CN110765310A (en) * | 2019-10-09 | 2020-02-07 | 上海麦克风文化传媒有限公司 | Audio album recommendation method and system based on parameter configuration |
| CN110781341A (en) * | 2019-10-09 | 2020-02-11 | 上海麦克风文化传媒有限公司 | Audio album recommendation method and system fusing multi-strategy recall data sets |
| CN110909200A (en) * | 2019-10-09 | 2020-03-24 | 上海麦克风文化传媒有限公司 | Offline and real-time combined audio album recommendation system recall method and system |
| CN111178970A (en) * | 2019-12-30 | 2020-05-19 | 微梦创科网络科技（中国）有限公司 | Advertisement delivery method and device, electronic equipment and computer readable storage medium |
| CN111324813A (en) * | 2020-02-20 | 2020-06-23 | 深圳前海微众银行股份有限公司 | Recommendation method, device, equipment and computer readable storage medium |
| CN111782927A (en) * | 2019-05-15 | 2020-10-16 | 北京京东尚科信息技术有限公司 | Article recommendation method and device, computer storage medium |
| CN113781147A (en) * | 2020-11-17 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Item recommendation method and device |
| CN114066278A (en) * | 2021-11-22 | 2022-02-18 | 北京百度网讯科技有限公司 | Method, apparatus, medium, and program product for evaluating article recall |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440259A (en) * | 2013-07-31 | 2013-12-11 | 亿赞普(北京)科技有限公司 | Network advertisement push method and device |
| CN103810030A (en) * | 2014-02-20 | 2014-05-21 | 北京奇虎科技有限公司 | Application recommendation method, device and system based on mobile terminal application market |
| CN104951465A (en) * | 2014-03-28 | 2015-09-30 | 腾讯科技（深圳）有限公司 | Application recommendation method and device |
| CN105488689A (en) * | 2015-11-19 | 2016-04-13 | 北京百度网讯科技有限公司 | Real-time recommendation method and device of consumption information |
| CN105975483A (en) * | 2016-04-25 | 2016-09-28 | 北京三快在线科技有限公司 | User preference-based message pushing method and platform |
| CN106056427A (en) * | 2016-05-25 | 2016-10-26 | 中南大学 | Spark-based big data hybrid model mobile recommending method |
| CN106296242A (en) * | 2015-05-22 | 2017-01-04 | 苏宁云商集团股份有限公司 | A kind of generation method of commercial product recommending list in ecommerce and the system of generation |
| CN106407477A (en) * | 2016-11-22 | 2017-02-15 | 深圳市互联在线云计算股份有限公司 | Multidimensional interconnection recommendation method and system |
| CN106874522A (en) * | 2017-03-29 | 2017-06-20 | 珠海习悦信息技术有限公司 | Information recommendation method, device, storage medium and processor |
| CN107038213A (en) * | 2017-02-28 | 2017-08-11 | 华为技术有限公司 | A kind of method and device of video recommendations |
| CN107424043A (en) * | 2017-06-15 | 2017-12-01 | 北京三快在线科技有限公司 | A kind of Products Show method and device, electronic equipment |
| CN107657506A (en) * | 2017-09-06 | 2018-02-02 | 北京五八到家信息技术有限公司 | The intelligent recommendation method and commending system of a kind of service commodity |
-
2018
- 2018-06-13 CN CN201810607170.4A patent/CN108830689A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103440259A (en) * | 2013-07-31 | 2013-12-11 | 亿赞普(北京)科技有限公司 | Network advertisement push method and device |
| CN103810030A (en) * | 2014-02-20 | 2014-05-21 | 北京奇虎科技有限公司 | Application recommendation method, device and system based on mobile terminal application market |
| CN104951465A (en) * | 2014-03-28 | 2015-09-30 | 腾讯科技（深圳）有限公司 | Application recommendation method and device |
| CN106296242A (en) * | 2015-05-22 | 2017-01-04 | 苏宁云商集团股份有限公司 | A kind of generation method of commercial product recommending list in ecommerce and the system of generation |
| CN105488689A (en) * | 2015-11-19 | 2016-04-13 | 北京百度网讯科技有限公司 | Real-time recommendation method and device of consumption information |
| CN105975483A (en) * | 2016-04-25 | 2016-09-28 | 北京三快在线科技有限公司 | User preference-based message pushing method and platform |
| CN106056427A (en) * | 2016-05-25 | 2016-10-26 | 中南大学 | Spark-based big data hybrid model mobile recommending method |
| CN106407477A (en) * | 2016-11-22 | 2017-02-15 | 深圳市互联在线云计算股份有限公司 | Multidimensional interconnection recommendation method and system |
| CN107038213A (en) * | 2017-02-28 | 2017-08-11 | 华为技术有限公司 | A kind of method and device of video recommendations |
| CN106874522A (en) * | 2017-03-29 | 2017-06-20 | 珠海习悦信息技术有限公司 | Information recommendation method, device, storage medium and processor |
| CN107424043A (en) * | 2017-06-15 | 2017-12-01 | 北京三快在线科技有限公司 | A kind of Products Show method and device, electronic equipment |
| CN107657506A (en) * | 2017-09-06 | 2018-02-02 | 北京五八到家信息技术有限公司 | The intelligent recommendation method and commending system of a kind of service commodity |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111782927A (en) * | 2019-05-15 | 2020-10-16 | 北京京东尚科信息技术有限公司 | Article recommendation method and device, computer storage medium |
| CN110602532A (en) * | 2019-09-24 | 2019-12-20 | 腾讯科技（深圳）有限公司 | Entity article recommendation method, device, server and storage medium |
| CN110765310A (en) * | 2019-10-09 | 2020-02-07 | 上海麦克风文化传媒有限公司 | Audio album recommendation method and system based on parameter configuration |
| CN110781341A (en) * | 2019-10-09 | 2020-02-11 | 上海麦克风文化传媒有限公司 | Audio album recommendation method and system fusing multi-strategy recall data sets |
| CN110909200A (en) * | 2019-10-09 | 2020-03-24 | 上海麦克风文化传媒有限公司 | Offline and real-time combined audio album recommendation system recall method and system |
| CN110765310B (en) * | 2019-10-09 | 2024-01-30 | 上海麦克风文化传媒有限公司 | Audio album recommendation method and system based on parameter configuration |
| CN111178970A (en) * | 2019-12-30 | 2020-05-19 | 微梦创科网络科技（中国）有限公司 | Advertisement delivery method and device, electronic equipment and computer readable storage medium |
| CN111324813A (en) * | 2020-02-20 | 2020-06-23 | 深圳前海微众银行股份有限公司 | Recommendation method, device, equipment and computer readable storage medium |
| CN113781147A (en) * | 2020-11-17 | 2021-12-10 | 北京沃东天骏信息技术有限公司 | Item recommendation method and device |
| CN114066278A (en) * | 2021-11-22 | 2022-02-18 | 北京百度网讯科技有限公司 | Method, apparatus, medium, and program product for evaluating article recall |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830689A (en) | Item recommendation method, device, server and storage medium | |
| CN109685631B (en) | Personalized recommendation method based on big data user behavior analysis | |
| CN105740268B (en) | A kind of information-pushing method and device | |
| CN109684538A (en) | A kind of recommended method and recommender system based on individual subscriber feature | |
| CN108416649A (en) | Search result ordering method, device, electronic equipment and storage medium | |
| CN109582876B (en) | Tourist industry user portrait construction method and device and computer equipment | |
| Fournier-Viger et al. | Using partially-ordered sequential rules to generate more accurate sequence prediction | |
| US20150339759A1 (en) | Detecting product attributes associated with product upgrades based on behaviors of users | |
| KR102474747B1 (en) | Apparatus and method for recommending product based on user behavior pattern and for forecasting user preference for recommended product | |
| CN103345695A (en) | Commodity recommendation method and device | |
| CN102682005A (en) | Method and device for determining preference categories | |
| CN104835057A (en) | Method and device for obtaining consumption feature information of network user | |
| CN103118111A (en) | Information push method based on data from a plurality of data interaction centers | |
| CN109242654A (en) | A kind of item recommendation method and system | |
| US20150154685A1 (en) | Automated detection of new item features by analysis of item attribute data | |
| CN105069077A (en) | Search method and device | |
| Dias et al. | Automating the extraction of static content and dynamic behaviour from e-commerce websites | |
| Lacic et al. | Towards a scalable social recommender engine for online marketplaces: The case of apache solr | |
| CN113689259A (en) | Commodity personalized recommendation method and system based on user behaviors | |
| CN115439197A (en) | E-commerce recommendation method and system based on knowledge map deep learning | |
| Feng et al. | Revisit recommender system in the permutation prospective | |
| KR101658714B1 (en) | Method and system for predicting online customer action based on online activity history | |
| CN108984668A (en) | A kind of method, apparatus of data processing, equipment and storage medium | |
| CN115423555A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
| CN110827044A (en) | Method and device for extracting user interest mode |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181116 |