CN107193831A - Information recommendation method and device - Google Patents

Information recommendation method and device Download PDF

Info

Publication number
CN107193831A
CN107193831A CN201610145260.7A CN201610145260A CN107193831A CN 107193831 A CN107193831 A CN 107193831A CN 201610145260 A CN201610145260 A CN 201610145260A CN 107193831 A CN107193831 A CN 107193831A
Authority
CN
China
Prior art keywords
user
action
data
user behavior
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
Application number
CN201610145260.7A
Other languages
Chinese (zh)
Inventor
吕韬
刘睿
方小刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Alibaba Group Holding Ltd
Original Assignee
Alibaba Group Holding Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Alibaba Group Holding Ltd filed Critical Alibaba Group Holding Ltd
Priority to CN201610145260.7A priority Critical patent/CN107193831A/en
Publication of CN107193831A publication Critical patent/CN107193831A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements
    • G06Q30/0255—Targeted advertisements based on user history
    • 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 application provides a kind of information recommendation method and device.Method includes：According to the universal data collection specification suitable for each business scenario, the user behavior in network system is gathered in real time, to obtain active user behavioral data；According to the general data processing logic suitable for each business scenario, active user behavioral data is analyzed, to obtain at least one recommending data model；According to business scenario to be recommended, from least one recommending data model, the reference data needed for business scenario to be recommended is obtained；Reference data according to needed for business scenario to be recommended carries out information recommendation to business scenario to be recommended.The application provides a kind of versatility stronger recommendation method, to adapt to miscellaneous service scene.

Description

Information recommendation method and device
【Technical field】
The application is related to Internet technical field, more particularly to a kind of information recommendation method and device.
【Background technology】
With the development of Internet technology, the content (referred to as network object) that user can obtain from website is increasingly It is many.User browse web sites selection network object during, recommendation of websites system is played a very important role, particularly pair In the user without clear and definite demand, it is likely that the network object that recommendation of websites system is recommended can be directly selected.One efficiently Commending system, can not only be user-friendly, lifted website self-value, and can more importantly reduce user overflow It is random the behavior such as to browse, click on, be conducive to mitigating the burden of Website server, save network bandwidth resources.
At present, increasing operation system begins with the commending system of oneself.Existing commending system is both for specific Business scenario, the data flow needed for it is solidification, and the type of service that current operation system is provided is being on the increase, if Directly existing commending system is applied in new business, it is necessary to access data relevant with new business again, it is this to realize cost It is high, so the versatility of existing commending system is poor, it is impossible to directly apply to multiple business scene, it is impossible to meet class of business Growing number of application demand.
【The content of the invention】
The many aspects of the application provide a kind of information recommendation method and device, and to provide, a kind of versatility is stronger to be pushed away Method is recommended, to adapt to miscellaneous service scene.
The one side of the application there is provided a kind of information recommendation method, including：
According to the universal data collection specification suitable for each business scenario, the user behavior in network system is carried out real-time Collection, to obtain active user behavioral data；
According to the general data processing logic suitable for each business scenario, the active user behavioral data is divided Analysis, to obtain at least one recommending data model；
According to business scenario to be recommended, from least one recommending data model, the business to be recommended is obtained Reference data needed for scape；
Reference data according to needed for the business scenario to be recommended carries out information recommendation to the business scenario to be recommended.
In the optional embodiment of the application one, the universal data collection specification according to suitable for each business scenario, User behavior in network system is gathered in real time, to obtain active user behavioral data, including：
The business road where service identification, the network object mark, network object that the user behavior is related to is gathered in real time Footpath, user's mark and user behavior type, to form the active user behavioral data.
In the optional embodiment of the application one, the general data processing logic according to suitable for each business scenario, The active user behavioral data is analyzed, to obtain at least one recommending data model, including：
From two dimensions of user behavior type and object type, the active user behavioral data is parsed, to obtain User behavior data of each object of action under various user behavior types is obtained, wherein, the object of action is user's row The node on service path where the network object being related to for the network object being related to or the user behavior；
According to the information recommendation demand of each business scenario, to use of each object of action under various user behavior types Family behavioral data is analyzed and processed, to obtain at least one recommending data model.
It is described from two dimensions of user behavior type and object type in the optional embodiment of the application one, to described Active user behavioral data is parsed, to obtain user behavior data of each object of action under various user behavior types, Including：
The service identification being related to according to the user behavior, will draw including the active user behavioral data that identical services are identified Enter a packet；
From two dimensions of user behavior type and object type, the active user behavioral data in each packet is entered Row parsing, to obtain user behavior data of each object of action under various user behavior types in each packet.
In the optional embodiment of the application one, the information recommendation demand according to each business scenario, to each row Analyzed and processed for user behavior data of the object under various user behavior types, to obtain at least one recommendation number According to model, including：
According to network object recommended requirements, user preference recommended requirements, network object temperature recommended requirements and network pair As at least one of transfer relationship recommended requirements recommended requirements, to each object of action under various user behavior types User behavior data is analyzed and processed, with obtain the basic recommending data model of network object, user preference recommending data model, User's history behavior recommending data model, network object temperature recommending data model and network object transfer relationship recommending data At least one of model recommending data model.
It is described according to network object recommended requirements in the optional embodiment of the application one, each object of action is existed User behavior data under various user behavior types is analyzed and processed, to obtain the basic recommending data model of network object, Including：
From the dimension of network object, user behavior data of each object of action under various user behavior types is entered Row organization and administration, to obtain the basic data of each network object.
It is described according to user preference recommended requirements in the optional embodiment of the application one, each object of action is existed User behavior data under various user behavior types is analyzed and processed, to obtain user preference recommending data model, including：
From user, three dimensions of network object and user behavior type, to each object of action in various user's rows Counted for the user behavior data under type, obtain each user and occur each type of user behavior to each object of action Number of times；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each behavior The preference-score of object.
It is described according to network object temperature recommended requirements in the optional embodiment of the application one, to each behavior pair As the user behavior data under various user behavior types is analyzed and processed, to obtain network object temperature recommending data mould Type, including：
From user, three dimensions of network object and user behavior type, to each object of action in various user's rows Counted for the user behavior data under type, obtain each user and occur each type of user behavior to each object of action Number of times；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each behavior The preference-score of object；
According to preference-score of each user to each object of action, the hot value of each object of action is calculated.
It is described that each type of user is occurred to each object of action according to each user in the optional embodiment of the application one The number of times of behavior, calculates preference-score of each user to each object of action, including：
To each user, occurs each type of user behavior to each object of action in each time window to the user Number of times is weighted summation, to obtain preference-score of the user to each object of action in each time window；
According to time decay factor, the user is weighted in each time window to the preference-score of each object of action Summation, to obtain preference-score of the user to each object of action.
It is described according to preference-score of each user to each object of action in the optional embodiment of the application one, calculate each The hot value of object of action, including：
To each object of action, each user is added up to the preference-score of the object of action, to obtain the row For the hot value of object.
It is described according to network object transfer relationship recommended requirements in the optional embodiment of the application one, to each row Analyzed and processed, pushed away with obtaining network object transfer relationship for user behavior data of the object under various user behavior types Data model is recommended, including：
The dimension shifted from object, enters to user behavior data of each object of action under various user behavior types Row statistics, to obtain the transfer pair of at least one object of action, the object of action transfer is to representing to use twice before and after same user Transfer relationship between the object of action that family behavior is related to.
It is described according to business scenario to be recommended in the optional embodiment of the application one, from least one recommendation number According in model, the reference data needed for the business scenario to be recommended is obtained, including：
When the business scenario to be recommended is that the goal behavior object searched for according to targeted customer is used to the target Family recommend object of action when, from least one described object of action shift centering, obtain comprising the goal behavior object and from The goal behavior object turns at least one candidate's object of action transfer pair of another object of action；
The reference data according to needed for the business scenario to be recommended enters row information to the business scenario to be recommended Recommend, including：
Each candidate's object of action transfer pair of centering is shifted at least one described candidate's object of action, according to the mesh Mark the number of times that object of action turns to another object of action, calculate candidate's object of action transfer to transition probability；
According to each candidate's object of action shift to transition probability, from least one described candidate's object of action transfer pair Middle selection target object of action transfer pair, row of the centering in addition to the goal behavior object is shifted by the goal behavior object The targeted customer is given for object recommendation.
The another aspect of the application there is provided a kind of information recommending apparatus, including：
Real-time acquisition module, for according to the universal data collection specification suitable for each business scenario, in network system User behavior gathered in real time, to obtain active user behavioral data；
Model computation module, for according to the general data processing logic suitable for each business scenario, to the real-time use Family behavioral data is analyzed, to obtain at least one recommending data model；
Parameter extraction module, for according to business scenario to be recommended, from least one recommending data model, is obtained Reference data needed for the business scenario to be recommended；
Information recommendation module, for the reference data according to needed for the business scenario to be recommended to the business to be recommended Scene carries out information recommendation.
In the optional embodiment of the application one, the real-time acquisition module specifically for：
The business road where service identification, the network object mark, network object that the user behavior is related to is gathered in real time Footpath, user's mark and user behavior type, to form the active user behavioral data.
In the optional embodiment of the application one, the model computation module includes：
Data resolution unit, for from two dimensions of user behavior type and object type, to the active user behavior Data are parsed, to obtain user behavior data of each object of action under various user behavior types, wherein, the behavior On service path where the network object that the network object or the user behavior that object is related to for the user behavior are related to Node；
Model treatment unit, for the information recommendation demand according to each business scenario, to each object of action various User behavior data under user behavior type is analyzed and processed, to obtain at least one recommending data model.
In the optional embodiment of the application one, the data resolution unit specifically for：
The service identification being related to according to the user behavior, will draw including the active user behavioral data that identical services are identified Enter a packet；
From two dimensions of user behavior type and object type, the active user behavioral data in each packet is entered Row parsing, to obtain user behavior data of each object of action under various user behavior types in each packet.
In the optional embodiment of the application one, the model treatment unit specifically for：
According to network object recommended requirements, user preference recommended requirements, network object temperature recommended requirements and network pair As at least one of transfer relationship recommended requirements recommended requirements, to each object of action under various user behavior types User behavior data is analyzed and processed, with obtain the basic recommending data model of network object, user preference recommending data model, User's history behavior recommending data model, network object temperature recommending data model and network object transfer relationship recommending data At least one of model recommending data model.
In the optional embodiment of the application one, the model treatment unit is obtaining the network object basis recommendation number During according to model, specifically for：
From the dimension of network object, user behavior data of each object of action under various user behavior types is entered Row organization and administration, to obtain the basic data of each network object.
In the optional embodiment of the application one, the model treatment unit is obtaining the user preference recommending data mould During type, specifically for：
From user, three dimensions of network object and user behavior type, to each object of action in various user's rows Counted for the user behavior data under type, obtain each user and occur each type of user behavior to each object of action Number of times；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each behavior The preference-score of object.
In the optional embodiment of the application one, the model treatment unit is obtaining the network object temperature recommendation number During according to model, specifically for：
From user, three dimensions of network object and user behavior type, to each object of action in various user's rows Counted for the user behavior data under type, obtain each user and occur each type of user behavior to each object of action Number of times；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each behavior The preference-score of object；
According to preference-score of each user to each object of action, the hot value of each object of action is calculated.
In the optional embodiment of the application one, the model treatment unit specifically for：
To each user, occurs each type of user behavior to each object of action in each time window to the user Number of times is weighted summation, to obtain preference-score of the user to each object of action in each time window；
According to time decay factor, the user is weighted in each time window to the preference-score of each object of action Summation, to obtain preference-score of the user to each object of action.
In the optional embodiment of the application one, the model treatment unit specifically for：
To each object of action, each user is added up to the preference-score of the object of action, to obtain the row For the hot value of object.
In the optional embodiment of the application one, the model treatment unit is pushed away in the acquisition network object transfer relationship When recommending data model, specifically for：
The dimension shifted from object, enters to user behavior data of each object of action under various user behavior types Row statistics, to obtain the transfer pair of at least one object of action；The object of action transfer is to representing to use twice before and after same user Transfer relationship between the object of action that family behavior is related to.
In the optional embodiment of the application one, the parameter extraction module specifically for：
When the business scenario to be recommended is that the goal behavior object searched for according to targeted customer is used to the target Family recommend object of action when, from least one described object of action shift centering, obtain comprising the goal behavior object and from The goal behavior object turns at least one candidate's object of action transfer pair of another object of action；
Described information recommending module specifically for：
Each candidate's object of action transfer pair of centering is shifted at least one described candidate's object of action, according to the mesh Mark the number of times that object of action turns to another object of action, calculate candidate's object of action transfer to transition probability；
According to each candidate's object of action shift to transition probability, from least one described candidate's object of action transfer pair Middle selection target object of action transfer pair, row of the centering in addition to the goal behavior object is shifted by the goal behavior object The targeted customer is given for object recommendation.
In this application, according to the universal data collection specification suitable for each business scenario, in real time in collection network system User behavior, obtain active user behavioral data, according to the general data processing logic suitable for each business scenario, to real-time User behavior data is analyzed, to obtain at least one recommending data model, can basis when needing to carry out information recommendation Business scenario to be recommended obtains reference data from least one recommending data model, is treated according to the reference data of acquisition to described Business scenario is recommended to carry out information recommendation.As can be seen here, the application is based on the universal data collection specification for being adapted to each business scenario Data acquisition is carried out, and recommending data model is provided based on the general data processing logic for being adapted to each business scenario so that can be with According to the difference of business scenario, specific reference data is selected to be carried out to the business scenario to be recommended from recommending data model Information recommendation, therefore go for miscellaneous service scene, versatility is stronger.
【Brief description of the drawings】
, below will be to embodiment or description of the prior art in order to illustrate more clearly of the technical scheme in the embodiment of the present application In required for the accompanying drawing that uses be briefly described, it should be apparent that, drawings in the following description are some realities of the application Example is applied, for those of ordinary skill in the art, on the premise of not paying creative work, can also be according to these accompanying drawings Obtain other accompanying drawings.
The schematic flow sheet for the information recommendation method that Fig. 1 provides for the embodiment of the application one；
Fig. 2 delivers block schematic illustration for the data that another embodiment of the application is provided；
The structural representation for the information recommending apparatus that Fig. 3 provides for the another embodiment of the application；
The structural representation for the information recommending apparatus that Fig. 4 provides for the another embodiment of the application.
【Embodiment】
To make the purpose, technical scheme and advantage of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application In accompanying drawing, the technical scheme in the embodiment of the present application is clearly and completely described, it is clear that described embodiment is Some embodiments of the present application, rather than whole embodiments.Based on the embodiment in the application, those of ordinary skill in the art The every other embodiment obtained under the premise of creative work is not made, belongs to the scope of the application protection.
The schematic flow sheet for the information recommendation method that Fig. 1 provides for the embodiment of the application one.As shown in figure 1, this method bag Include：
101st, according to the universal data collection specification suitable for each business scenario, the user behavior in network system is carried out Collection in real time, to obtain active user behavioral data.
102nd, according to the general data processing logic suitable for each business scenario, active user behavioral data is divided Analysis, to obtain at least one recommending data model.
103rd, according to business scenario to be recommended, from least one recommending data model, business scenario institute to be recommended is obtained The reference data needed.
104th, the reference data according to needed for business scenario to be recommended carries out information recommendation to business scenario to be recommended.
The present embodiment provides a kind of information recommendation method, can be performed by information recommending apparatus, a kind of general to provide The stronger recommendation method of property, to be adapted to miscellaneous service scene.
In actual applications, because the business of large-scale website is various, business is not high with the information commons before business.It is existing There is commending system both for specific transactions scene, its required data flow is solidification, if existing commending system is straight In the new business scene for accessing a Heterogeneous data, its cost of implementation is very high.Using the application in e-commerce field as Example, it is assumed that be used for a commending system for being used for Recommendations to recommend promotion/preferential activity, this needs to access and promote again Pin/related the data of preferential activity, and also need to update the data calculating, its cost of access is no less than for promotion/preferential work It is dynamic to do a set of commending system again.
In view of the above-mentioned problems, the present embodiment provides a kind of general information recommendation method, specifically：
First there is provided a kind of universal data collection specification for being adapted to each business scenario, then according to suitable for each business The universal data collection specification of scape, is gathered in real time to the user behavior in network system, to obtain active user behavior number According to being used as the initial data of information recommendation.Here user behavior is primarily referred to as user and the network object in network system is entered The behavior of row operation.
According to the difference of application scenarios, the network system, the user behavior in network system and network object of the present embodiment Can be different.The present embodiment does not enter to the user behavior and the way of realization of network object in network system, network system Row is limited.
For example, network system can be the network system using each electric business website as core, and accordingly, the net in network system Network object can be that the user behavior in merchandise items, network system can include browsing, collecting plus shopping cart, purchase, payment At least one of and comment.
In another example, network system can be the network system using resource download site as core, accordingly, in network system Network object can for download resource, such as audio and video resources, textual resources, the user behavior in network system can be wrapped Include at least one of preview, download, broadcasting etc..
In addition, the network object of the present embodiment can also be the service object based on internet, the service object can be with It is that carwash service, maintenance service, massage service, cleaning service, cook make house calls, household services, private tutor service, entertainment garment Business, food and drink service, travel service, hotel service, service of cars on hire etc..
Can often occur having user in a user behavior, such as network system in network system on implementing Clicking has user to collect some network object etc. in some network object, or network system, information recommending apparatus is just held Acquisition operations of row, are gathered in real time to the user behavior, to obtain active user behavioral data.Or,
In order to reduce the number of times that information recommending apparatus is gathered in real time, to mitigate the burden of information recommending apparatus, it can set One caching (cache) space, for the user behavior occurred in logical volume transfer method and storage network system.User behavior one user of correspondence Behavioral data.For example, when the spatial cache is fully written, information recommending apparatus performs an acquisition operations, the caching is gathered empty Between in user behavior data.Or, when the user behavior data recorded in the spatial cache reaches certain amount (such as 500 Bar) when, information recommending apparatus performs an acquisition operations, gathers the user behavior data in the spatial cache.Or, can be pre- Intervals (such as 1 minute) are first set, information recommending apparatus can be performed at interval of intervals and once gathered Operation, gathers the user behavior data in the spatial cache.
Due to the information recommendation method that the present embodiment is provided be towards whole network system, preferably the need for adapt to network Each business scenario in system, it is therefore necessary to define a set of to embody the details of user behavior information and network object Data standard, and on the one hand the data standard will can distinguish different business scenarios, on the other hand also be adapted to difference Business scenario.In the present embodiment, the data standard is referred to as universal data collection specification.Based on this, information recommending apparatus When the user behavior in network system is gathered in real time, it can be advised according to the universal data collection for being adapted to each business scenario Model, is gathered in real time to the user behavior in network system.
In view of in each business scenario, user behavior is typically produced by behavior user (referred to as user) to network object Behavior constitute, i.e., user behavior can be related to the information such as user, network object and user behavior type.Based on this, above-mentioned In a kind of example of universal data collection specification, can specify that collection user behavior be related to user mark, network object mark, User behavior type.In addition, business can uniquely be distinguished by its mark, what service identification was typically unified by system provider Distribution and management.In order to distinguish different business scene, it may further specify that collection user behavior is related in universal data collection specification And service identification, to identify the business belonging to the user behavior.
Further, the service path where network object can reflect the hierarchical relationship or ownership of network object.For example, mesh Preceding most of electric business business all can manage commodity using tree-shaped or the institutional framework of directory type.For example, commodity have ownership Industry, classification；The seller of ownership can be divided again simultaneously, it is also possible to divide activity of ownership etc.., can for some commodity To there is multiple service paths, different service paths records different hierarchical structures.Such as the first service path record commodity are returned The trade information of category, the businessman of the ownership of the second service path record commodity, the activity of the 3rd service path record commodity ownership Information.Based on this, above-mentioned universal data collection specification may further specify that the industry where the network object that collection user behavior is related to Business path.
In summary, a kind of universal data collection specification provides that the parameter of collection is as shown in table 1.
Table 1
By taking e-commerce field as an example, table 2 provides a kind of example, for illustrating between user behavior type and its label Corresponding relation.According to the difference of application scenarios, the corresponding relation adaptability between user behavior type and label is set.
Table 2
Behavior type (action_type) label Explanation
1 Click is browsed
2 System is left a message
3 Qq or wechat message
4 Check contact method
5 Collection
6 Place an order
7 Pay
8 Receive
9 Return goods
10 Transaction is closed
11 Valid contract is put on record
12 Add receiving tally
13 Search
14 Like
15 Browse for a long time
16 Vip identity is activated
… …
Based on the example of above-mentioned universal data collection specification, information recommending apparatus can specifically be advised according to universal data collection Model, collection user behavior is related in real time service identification, network object mark, the service path where network object, Yong Hubiao The information such as knowledge and user behavior type, to form active user behavioral data.Here active user behavioral data includes：Industry Business mark, network object mark, service path, user's mark and user behavior type.
After active user behavioral data is obtained, information recommending apparatus can be according to the general number for being adapted to each business scenario According to processing logic, the active user behavioral data collected is analyzed, to obtain at least one recommending data model.At this In embodiment, every kind of recommending data model includes the reference data needed for information recommendation, and different recommending data models include Different reference datas.Because the present embodiment is real-time to what is collected based on the general data processing logic for being adapted to each business scenario User behavior data is handled, therefore is adapted to each business scenario.
, can be according to business scenario, from least one recommending data when some business scenario needs to carry out information recommendation The reference data needed for the business scenario is obtained in model, entering row information to the business scenario according to acquired reference data pushes away Recommend.For ease of description, the business scenario for needing to carry out information recommendation is referred to as business scenario to be recommended by the present embodiment.Industry to be recommended Business scene can be the miscellaneous service scene that network system is supported.
Because the present embodiment is based on the universal data collection specification progress data acquisition for being adapted to each business scenario, and based on suitable The general data processing logic for closing each business scenario provides recommending data model so that can according to the difference of business scenario, from Select specific reference data to carry out information recommendation to the business scenario to be recommended in recommending data model, therefore can be applicable In miscellaneous service scene, versatility is stronger.
How according to the general data processing logic suitable for each business scenario, active user behavioral data is divided Analysis, to obtain the core that at least one recommending data model is technical scheme.To specifically it illustrate below.
In a kind of optional embodiment, the above-mentioned general data processing logic for being adapted to each business scenario can be specifically full The data process method of the information recommendation demand of each business scenario of foot.Information recommending apparatus can be from user behavior type and object Two dimensions of type, are parsed to active user behavioral data, to obtain each object of action under various user behavior types User behavior data, wherein, the net that the network object or user behavior that object of action here is related to for user behavior are related to The node on service path where network object；Then, according to the information recommendation demand of each business scenario, each object of action is existed User behavior data under various user behavior types is analyzed and processed, to obtain at least one recommending data model.
In the above-described embodiment, from two dimensions of user behavior type and object type, to active user behavioral data Parsed, maintenance management can be carried out to user behavior data from two dimensions of user behavior type and object type, can be just Assembled in data according to different dimensions, be conducive to improving the efficiency of follow-up data processing.
Further, it is contemplated that be usually separate between different business scene.So, can be according to user behavior The service identification being related to, will be included in a packet, the data point including the active user behavioral data that identical services are identified Group carries out storage management by independent message queue, so that the isolation between realizing business.Then, from user behavior type and Two dimensions of object type, are parsed to the active user behavioral data in each packet, to obtain each data point User behavior data of each object of action under various user behavior types in group.Subsequently, information recommending apparatus can be to each User behavior data of each object of action under various user behavior types in packet, to obtain each packet correspondence At least one recommending data model.
In an optional embodiment, being adapted to the information recommendation demand of each business scenario includes：Network object recommended requirements, At least one of user preference recommended requirements, network object temperature recommended requirements and network object transfer relationship recommended requirements Recommended requirements.Accordingly, at least one recommending data model includes：Network object basis recommending data model, user preference are pushed away Recommend data model, user's history behavior recommending data model, network object temperature recommending data model and network object transfer At least one of relation recommending data model recommending data model.Based on this, the above-mentioned information recommendation according to each business scenario Demand, is analyzed and processed to user behavior data of each object of action under various user behavior types, to obtain at least one The embodiment for planting recommending data model is specifically included：
According to network object recommended requirements, user preference recommended requirements, network object temperature recommended requirements and network pair As at least one of transfer relationship recommended requirements recommended requirements, to user of each object of action under various user behavior types Behavioral data is analyzed and processed, to obtain the basic recommending data model of network object, user preference recommending data model, user Historical behavior recommending data model, network object temperature recommending data model and network object transfer relationship recommending data model At least one of recommending data model.
Process for obtaining the basic recommending data model of network object according to network object recommended requirements：Information recommendation is filled Tissue can be carried out from the dimension of network object to user behavior data of each object of action under various user behavior types by putting Management, to obtain the basic data of each network object.It means that network object basis recommending data model includes each network pair The basic data of elephant.The basic data of network object includes but is not limited to：Service identification, the network pair of the affiliated business of network object Service path as where mark, network object etc..
On implementing, the service identification and network object of the affiliated business of network object mark can as major key, and Service path and other information where network object can be as the corresponding values of major key, in order to be inquired about.
In addition, the service path where network object may change, it can now be replaced with new service path Old service path.In actual applications, may occur mistake in data acquisition causes service path to change, This stylish service path is actually mistake, if mistake will be occurred by replacing old service path with new service path By mistake, to avoid this mistake, the occurrence number of new service path can be recorded, when the occurrence number of new service path reaches During predetermined threshold value, then with the old service path of new service path replacement, to reduce the probability that service path is updated by mistake.
Process for obtaining user preference recommending data model according to user preference recommended requirements：Information recommending apparatus can With from user, three dimensions of network object and user behavior type, to each object of action under various user behavior types User behavior data is counted, and obtains each user and each object of action occurs the number of times of each type of user behavior；Then, The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to the inclined of each object of action Good score.
Further, it is contemplated that user behavior can over time and gradually to the influence degree of user or network object Weaken, so the present embodiment is theoretical according to this great forgetting curve of Chinese mugwort guest, it is that different user behavior type sets different weights, should Weights are using the time as attenuation parameter, so that more influence of the reasonable contemplation user behavior to user preference.
Based on above-mentioned, to each user, each object of action occur in each time window each type of use to the user The number of times of family behavior is weighted summation, to obtain preference-score of the user to object of action in the time window；According to Time decay factor, is weighted summation, to be somebody's turn to do in each time window to the user to the preference-score of each object of action Preference-score of the user to each object of action.
Specifically, information recommending apparatus can calculate preference-score of each user to each object of action according to formula (1).
In above-mentioned formula (1), f (u, o) represents preference-scores of the user u to object of action o；U represents user；O is represented Object of action, can be the node on the service path where the network object or network object that user behavior is related to；M is represented The sum of time window, each time window, which is represented, specifies duration；For example, the duration that each time window is represented can be but not limited to： 20 minutes；yi(u, o) represents preference-scores of the user u in i-th of time window to object of action o；P (t) represent Chinese mugwort guest it is great this The time decay factor of forgetting curve.
Wherein,P (t)=K1+exp((-t-K2)/k3)。
N represents the quantity of user behavior type, such as clicking on, place an order, pay, collect belong to different types of user's row For；wjRepresent the weight of jth kind user behavior type；Xj(u, o) represents that user u object of action o occurs the use of jth type The number of times of family behavior；T represents time span；K1、K2、K3It is decay factor.
By above-mentioned processing, the preference-score of each user can be obtained.User preference recommending data model includes each user Preference-score.
In addition, during above-mentioned calculating user preference recommending data model, the history row of each user can also be obtained For data so as to form user's history behavior recommending data model.
Process for obtaining network object temperature recommending data model according to network object temperature recommended requirements：Information is pushed away Recommending device can be from user, three dimension of network object and user behavior type, to each object of action in various user behaviors User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each object of action Preference-score；According to preference-score of each user to each object of action, the hot value of each object of action is calculated.
Similarly, information recommending apparatus can calculate each user according to above-mentioned concrete mode and the preference of each object of action is obtained Point, it will not be repeated here.
In a kind of embodiment, the process for calculating the hot value of each object of action is specially：To each behavior pair As being added up to each user to the preference-score of behavior object, to obtain the hot value of behavior object.
Specifically, information recommending apparatus according to formula (2), can calculate the hot value of each object of action.
In above-mentioned formula (2), h (o) represents object of action o hot value, and other parameters can be found in the solution in formula (1) Release or illustrate, will not be repeated here.
What deserves to be explained is, it is above-mentioned that each user is added up to the preference-score of consolidated network object, and by cumulative knot Fruit is only a kind of embodiment as the hot value of the network object, however it is not limited to this.For example, it is also possible to from single dimension Network object is ranked up, for example, is ranked up, is ranked up according to clicking rate according to sales volume, according to the sale of network object Price is ranked up etc., or can also be carried out according to many factors integrated ordered, is then determined according to sequence sequencing The hot value of network object.
Mistake for obtaining network object transfer relationship recommending data model according to network object transfer relationship recommended requirements Journey：The dimension that information recommending apparatus can be shifted from object, to user row of each object of action under various user behavior types Counted for data, to obtain the transfer pair of at least one object of action, object of action transfer here is to representing same user Transfer relationship between the object of action that front and rear user behavior twice is related to.
It is preferred that, the transfer of above-mentioned object of action is to that can be specifically behavior pair that second from the bottom user behavior of user is related to As the transfer relationship between the object of action that is related to last time user behavior.
It is preferred that, the type of the object of action of above-mentioned object of action transfer centering is identical with rank.For example, two behaviors pair As being all commodity or two object of action are all businessmans, or two object of action are all classifications of same levels, etc..
The dimension shifted for above- mentioned information recommendation apparatus from object, to each object of action under various user behavior types User behavior data counted, obtain object of action transfer to operation, it can be on implementing：Information recommendation Device obtains a user behavior data, note from user behavior data of each object of action under various user behavior types Record user's mark and network object mark that the user behavior data is related to；The user is inquired about according to user's mark to be identified User occur the network object mark that user behavior is related to the last time, the two network objects are identified into identified network Object formation object of action transfer pair.
Optionally, obtain object of action transfer to while, can be the transfer of behavior object to plus timestamp, with It is easy to the transfer of behavior object to being updated.For example, can be shifted by object of action to timestamp judge the row It is whether object is shifted to expired, if the time interval for example from timestamp to current time is more than the threshold value of prefixed time interval (such as threshold value can be but not limited to 7 days), it is determined that otherwise the transfer of behavior object determines that behavior object turns to expired Move to not out of date；If object of action is shifted to expired, it can be deleted, on the one hand can save memory space, it is another Aspect can also improve the precision of network object transfer relationship recommending data model, and then improve subsequently based on the network object turn Shifting relation recommending data model carries out precision during information recommendation.
By above-mentioned processing, the basic recommending data model of above-mentioned network object, user preference recommending data mould can be obtained Type, user's history behavior recommending data model, network object temperature recommending data model and network object transfer relationship are recommended At least one of data model recommending data model., can be according to business scenario after these recommending data models are obtained Reference data is extracted from these recommending data models, and then row information is entered to the business scenario to be recommended based on reference data Recommend, this part can be described as data launch process.A kind of schematic diagram of data launch process is as shown in Fig. 2 the present embodiment is provided These recommending data model supports multiple business scenes.
In data launch process, a unique mark (id), and record traffic are distributed in advance for each business scenario Parameter needed for scene and combinations thereof mode；When each business scenario sends parameter request, the mark of the business scenario is carried And the parameter (can for example include user's mark, object identity etc.) such as the parameter needed for it and combinations thereof mode；Model is configured Ginseng is obtained into corresponding recommending data model according to business scenario and its parameter of transmission and combinations thereof mode with assembling module Examine data and assembled according to combination, then returned in business scenario, so that business scenario exports recommendation results.
Illustrate：When user's search commercial articles on electric business platform, electric business platform can be recommended from user's history behavior Commodity similar to the commodity of user's current search in user's commodity browsed recently are obtained in data model, and it is inclined from user The higher commodity of preference-score in these similar commodity are obtained in good recommending data model, then by these similar and preference-scores Higher commercial product recommending is to user.Or, when user's search commercial articles on electric business platform, electric business platform can be from user's history Commodity similar to the commodity of user's current search in user's commodity browsed recently are obtained in behavior recommending data model, so The higher commodity of preference-score in these commodity are obtained from user preference recommending data model afterwards, then are pushed away from network object temperature Recommend and the higher commodity of hot value in these similar commodity obtained in data model, final choice preference-score is higher and hot value compared with High commercial product recommending is to user.Or, when user's search commercial articles on electric business platform, electric business platform can determine that user is current Classification path where search commercial articles, it is assumed for example that the commodity of user's search belong to one-piece dress class now, then according to the classification Path is scanned for into network object temperature recommending data model, obtains hot value n commodity of highest, and such as selection connects clothing Skirt class 50 best commodity of sales volume now, then recommend user.
In a kind of embodiment, business scenario to be recommended is specially the target line searched for according to targeted customer It is object to targeted customer's recommendation network object.In the business scenario, one kind is obtained from least one recommending data model The embodiment of reference data needed for business scenario to be recommended includes：At least one included from network object transfer relationship model Individual object of action shifts centering, obtain comprising the goal behavior object and from goal behavior object turn to another object of action to Few candidate's object of action transfer pair.Here at least one candidate's object of action transfer is needed for being business scenario to be recommended Reference data.
Accordingly, the above-mentioned reference data according to needed for business scenario to be recommended is believed to the business scenario to be recommended Ceasing recommendation is specially：
Each candidate's object of action transfer pair of centering is shifted at least one candidate's object of action, according to candidate's behavior The goal behavior object of object transfer centering turns to the number of times of another object of action, calculate candidate's object of action transfer to turn Move probability；
For example, can according to equation below (3), obtain the transfer of each candidate's object of action to transition probability；
Afterwards, according to each candidate's object of action shift to transition probability, from least one candidate's object of action transfer The transfer pair of centering selection target object of action, object of action of the centering in addition to goal behavior object is shifted by goal behavior object Recommend targeted customer.
In above-mentioned formula (3),Represent by goal behavior object aiTurn to another object of action ajThe candidate row of formation For object shift to transition probability；Represent from goal behavior object aiTurn to another object of action ajTransfer number, it is real It is also number of times of candidate's object of action transfer to appearance on border, such as, it is assumed that from goal behavior object aiTurn to another behavior pair As ajThere are 100 times or 200 inferior, such as user browses after one-piece dress and browses the number of times of jeans and has 100 times or 200 times, Or user browses after Huawei's mobile phone and browses the number of times of iPhone for 50 times；k1、k2Represent adjustable parameter.
What deserves to be explained is, information recommending apparatus according to above-mentioned formula (3) calculate the transfer of candidate object of action to transfer Probability is actually ANALOGY OF BOLTZMANN DISTRIBUTION probability.Why the present embodiment uses ANALOGY OF BOLTZMANN DISTRIBUTION probability, mainly considers The measurement period of real-time system is shorter, and most sample statistics difference is little.And ANALOGY OF BOLTZMANN DISTRIBUTION can be tiny difference Away from being amplified from probability, so as to ensure that result has stronger discrimination.
In the embodiment of the present application, based on the universal data collection specification suitable for each business scenario, network is gathered in real time User behavior in system, to obtain active user behavioral data, is then based on being adapted to the general data processing of each business scenario These active user behavioral datas are analyzed to form recommending data model, recommending data model can use key by logic It is worth (key-value) form, versatility and portability are very strong.The technical scheme that the embodiment of the present application is provided is applied to not of the same trade or business Business scene, solves the defect that existing commending system is present, it is adaptable to the growing number of application scenarios of type of service, is conducive to drop The low development and maintenance cost recommended for Added Business scene information.
In addition, the embodiment of the present application, which is based on active user behavioral data, carries out information recommendation, it is not necessary to collected offline data Process, can be with rapid abutting joint new business, it is not necessary to which the excessive time carries out data accumulation, solve business change frequently and use Sparse the brought cold start-up problem of family behavior, lifts Consumer's Experience.
In product realization, above-mentioned real-time collection and the operation parsed to active user behavioral data can be by one Individual message-oriented middleware realizes, such as notify middlewares.The operation of each recommending data model of intermediate computations can pass through one Real time computation system realizes, such as storm.Above-mentioned data launch process can be configured by model and realized with assembling module, But not limited to this.
It should be noted that for foregoing each method embodiment, in order to be briefly described, therefore it is all expressed as a series of Combination of actions, but those skilled in the art should know, the application is not limited by described sequence of movement because According to the application, some steps can be carried out sequentially or simultaneously using other.Secondly, those skilled in the art should also know Know, embodiment described in this description belongs to preferred embodiment, involved action and module not necessarily the application It is necessary.
In the above-described embodiments, the description to each embodiment all emphasizes particularly on different fields, and does not have the portion being described in detail in some embodiment Point, it may refer to the associated description of other embodiment.
The structural representation for the information recommending apparatus that Fig. 3 provides for the another embodiment of the application.As shown in figure 3, the device Including：Real-time acquisition module 31, model computation module 32, parameter extraction module 33 and information recommendation module 34.
Real-time acquisition module 31, for according to the universal data collection specification suitable for each business scenario, to network system In user behavior gathered in real time, to obtain active user behavioral data.
Model computation module 32, for according to the general data processing logic suitable for each business scenario, to collection in real time The active user behavioral data that module 31 is collected is analyzed, to obtain at least one recommending data model.
Parameter extraction module 33, for according to business scenario to be recommended, at least one obtained from model computation module 32 In recommending data model, the reference data needed for business scenario to be recommended is obtained.
Information recommendation module 34, for the reference number needed for the business scenario to be recommended that is obtained according to parameter extraction module 33 Information recommendation is carried out according to business scenario to be recommended.
In an optional embodiment, real-time acquisition module 31 specifically for：
Collection user behavior is related in real time service identification, network object mark, the service path where network object, use Family is identified and user behavior type, to form active user behavioral data.
In an optional embodiment, as shown in figure 4, one kind of model computation module 32 realizes that structure includes：Data solution Analyse unit 321 and model treatment unit 322.
Data resolution unit 321, for from two dimensions of user behavior type and object type, to active user behavior number According to being parsed, to obtain user behavior data of each object of action under various user behavior types, wherein, object of action is The node on service path where the network object that the network object or user behavior that user behavior is related to are related to.
Model treatment unit 322, for the information recommendation demand according to each business scenario, is used each object of action various User behavior data under the behavior type of family is analyzed and processed, to obtain at least one recommending data model.
Further, data resolution unit 321 specifically for：
The service identification being related to according to user behavior, will be included in one including the active user behavioral data that identical services are identified Individual packet；
From two dimensions of user behavior type and object type, the active user behavioral data in each packet is entered Row parsing, to obtain user behavior data of each object of action under various user behavior types in each packet.
Further, model treatment unit 322 specifically for：
According to network object recommended requirements, user preference recommended requirements, network object temperature recommended requirements and network pair As at least one of transfer relationship recommended requirements recommended requirements, to user of each object of action under various user behavior types Behavioral data is analyzed and processed, to obtain the basic recommending data model of network object, user preference recommending data model, user Historical behavior recommending data model, network object temperature recommending data model and network object transfer relationship recommending data model At least one of recommending data model.
In an optional embodiment, model treatment unit 322 is when obtaining network object basis recommending data model, tool Body is used for：
From the dimension of network object, to user behavior data carry out group of each object of action under various user behavior types Management is knitted, to obtain the basic data of each network object.
In an optional embodiment, model treatment unit 322 is specific to use when obtaining user preference recommending data model In：
From user, three dimensions of network object and user behavior type, to each object of action in various user behavior classes User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each behavior The preference-score of object.
In an optional embodiment, model treatment unit 322 is when obtaining network object temperature recommending data model, tool Body is used for：
From user, three dimensions of network object and user behavior type, to each object of action in various user behavior classes User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each behavior The preference-score of object；
According to preference-score of each user to each object of action, the hot value of each object of action is calculated.
Further, in the process of above-mentioned acquisition user preference recommending data model or network object temperature recommending data model In, model treatment unit 322 each object of action is occurring the number of times of each type of user behavior according to each user, calculates each When user is to the preference-score of each object of action, specifically for：
To each user, each object of action occur in each time window time of each type of user behavior to the user Number is weighted summation, to obtain preference-score of the user to each object of action in each time window；
According to time decay factor, the user is weighted in each time window to the preference-score of each object of action and asked With to obtain preference-score of the user to each object of action.
Specifically, preference-score of each user to each object of action can be calculated according to formula (1).Can on formula (1) Referring to above method embodiment, it will not be repeated here.
Based on above-mentioned according to formula (1), each user is calculated to the preference-score of each object of action, model treatment unit 322 According to preference-score of each user to each object of action, the hot value for calculating each object of action is pushed away with constituting network object temperature When recommending data model, specifically for：
To each object of action, each user is added up to the preference-score of the object of action, to obtain the row For the hot value of object.
Specifically, the hot value of each object of action according to formula (2), can be calculated.Above-mentioned side is can be found on formula (2) Method embodiment, will not be repeated here.
In an optional embodiment, model treatment unit 322 is obtaining network object transfer relationship recommending data model When, specifically for：
The dimension shifted from object, unites to user behavior data of each object of action under various user behavior types Meter, to obtain the transfer pair of at least one object of action.Object of action transfer is to user behavior is related to twice before and after the same user of expression And object of action between transfer relationship.
Based on above-mentioned, parameter extraction module 33 is particularly used in：
When business scenario to be recommended is that the goal behavior object searched for according to targeted customer is recommended to go to targeted customer During for object, centering is shifted from least one object of action, obtains comprising goal behavior object and is turned to from goal behavior object At least one candidate's object of action transfer pair of another object of action.
Accordingly, information recommendation module 34 specifically for：
Each candidate's object of action transfer pair of centering is shifted at least one candidate's object of action, according to goal behavior pair Number of times as turning to another object of action, calculate candidate's object of action transfer to transition probability；
For example, can according to formula (3), calculate the transfer of each candidate's object of action to transition probability；
Afterwards, according to each candidate's object of action shift to transition probability, from least one candidate's object of action transfer The transfer pair of centering selection target object of action, object of action of the centering in addition to goal behavior object is shifted by goal behavior object Recommend targeted customer.Above method embodiment is can be found on formula (3), be will not be repeated here.
The information recommending apparatus that the present embodiment is provided, on the one hand based on the universal data collection specification for being adapted to each business scenario Data acquisition is carried out, and recommending data model is provided based on the general data processing logic for being adapted to each business scenario so that can be with According to the difference of business scenario, specific reference data is selected to enter row information to business scenario to be recommended from recommending data model Recommend, therefore go for miscellaneous service scene, versatility is stronger；On the other hand, carried out using active user behavioral data Information recommendation, it is not necessary to the process of collected offline data, can be with rapid abutting joint new business, it is not necessary to which the excessive time carries out data product It is tired, solve the problems, such as that business change is frequent and sparse the brought cold start-up of user behavior, lift Consumer's Experience.
It is apparent to those skilled in the art that, for convenience and simplicity of description, the system of foregoing description, The specific work process of device and unit, may be referred to the corresponding process in preceding method embodiment, will not be repeated here.
In several embodiments provided herein, it should be understood that disclosed system, apparatus and method can be with Realize by another way.For example, device embodiment described above is only schematical, for example, the unit Divide, only a kind of division of logic function there can be other dividing mode when actually realizing, such as multiple units or component Another system can be combined or be desirably integrated into, or some features can be ignored, or do not perform.It is another, it is shown or The coupling each other discussed or direct-coupling or communication connection can be the indirect couplings of device or unit by some interfaces Close or communicate to connect, can be electrical, machinery or other forms.
The unit illustrated as separating component can be or may not be it is physically separate, it is aobvious as unit The part shown can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple On NE.Some or all of unit therein can be selected to realize the mesh of this embodiment scheme according to the actual needs 's.
In addition, each functional unit in the application each embodiment can be integrated in a processing unit, can also That unit is individually physically present, can also two or more units it is integrated in a unit.Above-mentioned integrated list Member can both be realized in the form of hardware, it would however also be possible to employ hardware adds the form of SFU software functional unit to realize.
The above-mentioned integrated unit realized in the form of SFU software functional unit, can be stored in an embodied on computer readable and deposit In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are to cause a computer Equipment (can be personal computer, server, or network equipment etc.) or processor (processor) perform the application each The part steps of embodiment methods described.And foregoing storage medium includes：USB flash disk, mobile hard disk, read-only storage (Read- Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disc or CD etc. it is various Can be with the medium of store program codes.
Finally it should be noted that：Above example is only to the technical scheme for illustrating the application, rather than its limitations；Although The application is described in detail with reference to the foregoing embodiments, it will be understood by those within the art that：It still may be used To be modified to the technical scheme described in foregoing embodiments, or equivalent substitution is carried out to which part technical characteristic； And these modification or replace, do not make appropriate technical solution essence depart from each embodiment technical scheme of the application spirit and Scope.

Claims (24)

1. a kind of information recommendation method, it is characterised in that including：
According to the universal data collection specification suitable for each business scenario, the user behavior in network system is adopted in real time Collection, to obtain active user behavioral data；
According to the general data processing logic suitable for each business scenario, the active user behavioral data is analyzed, with Obtain at least one recommending data model；
According to business scenario to be recommended, from least one recommending data model, the business scenario institute to be recommended is obtained The reference data needed；
Reference data according to needed for the business scenario to be recommended carries out information recommendation to the business scenario to be recommended.
2. according to the method described in claim 1, it is characterised in that described to be adopted according to the conventional data suitable for each business scenario Collect specification, the user behavior in network system is gathered in real time, to obtain active user behavioral data, including：
The service path where service identification, the network object mark, network object that the user behavior is related to is gathered in real time, used Family is identified and user behavior type, to form the active user behavioral data.
3. according to the method described in claim 1, it is characterised in that at the conventional data according to suitable for each business scenario Logic is managed, the active user behavioral data is analyzed, to obtain at least one recommending data model, including：
From two dimensions of user behavior type and object type, the active user behavioral data is parsed, it is each to obtain User behavior data of the object of action under various user behavior types, wherein, the object of action is that the user behavior is related to And the network object that is related to of network object or the user behavior where service path on node；
According to the information recommendation demand of each business scenario, to user row of each object of action under various user behavior types Analyzed and processed for data, to obtain at least one recommending data model.
4. method according to claim 3, it is characterised in that described from two dimensions of user behavior type and object type Degree, is parsed to the active user behavioral data, to obtain user of each object of action under various user behavior types Behavioral data, including：
The service identification being related to according to the user behavior, will be included in one including the active user behavioral data that identical services are identified Individual packet；
From two dimensions of user behavior type and object type, the active user behavioral data in each packet is solved Analysis, to obtain user behavior data of each object of action under various user behavior types in each packet.
5. method according to claim 3, it is characterised in that the information recommendation demand according to each business scenario, right User behavior data of each object of action under various user behavior types is analyzed and processed, to obtain described at least one Recommending data model is planted, including：
Turned according to network object recommended requirements, user preference recommended requirements, network object temperature recommended requirements and network object At least one of shifting relation recommended requirements recommended requirements, to user of each object of action under various user behavior types Behavioral data is analyzed and processed, to obtain the basic recommending data model of network object, user preference recommending data model, user Historical behavior recommending data model, network object temperature recommending data model and network object transfer relationship recommending data model At least one of recommending data model.
6. method according to claim 5, it is characterised in that described according to network object recommended requirements, to each row Analyzed and processed for user behavior data of the object under various user behavior types, number is recommended to obtain network object basis According to model, including：
From the dimension of network object, to user behavior data carry out group of each object of action under various user behavior types Management is knitted, to obtain the basic data of each network object.
7. method according to claim 5, it is characterised in that described according to user preference recommended requirements, to each row Analyzed and processed for user behavior data of the object under various user behavior types, to obtain user preference recommending data mould Type, including：
From user, three dimensions of network object and user behavior type, to each object of action in various user behavior classes User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each object of action Preference-score.
8. method according to claim 5, it is characterised in that described according to network object temperature recommended requirements, to described User behavior data of each object of action under various user behavior types is analyzed and processed, and is pushed away with obtaining network object temperature Data model is recommended, including：
From user, three dimensions of network object and user behavior type, to each object of action in various user behavior classes User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each object of action Preference-score；
According to preference-score of each user to each object of action, the hot value of each object of action is calculated.
9. the method according to claim 7 or 8, it is characterised in that described that each object of action is occurred often according to each user The number of times of the user behavior of type, calculates preference-score of each user to each object of action, including：
To each user, each object of action occurs in each time window the number of times of each type of user behavior to the user Summation is weighted, to obtain preference-score of the user to each object of action in each time window；
According to time decay factor, the user is weighted in each time window to the preference-score of each object of action and asked With to obtain preference-score of the user to each object of action.
10. method according to claim 9, it is characterised in that described to be obtained according to each user to the preference of each object of action Point, the hot value of each object of action is calculated, including：
To each object of action, each user is added up to the preference-score of the object of action, to obtain the behavior pair The hot value of elephant.
11. the method according to claim any one of 5-8, it is characterised in that described to be pushed away according to network object transfer relationship Demand is recommended, user behavior data of each object of action under various user behavior types is analyzed and processed, to obtain Network object transfer relationship recommending data model, including：
The dimension shifted from object, unites to user behavior data of each object of action under various user behavior types Meter, to obtain the transfer pair of at least one object of action, object of action transfer is to representing before and after same user user's row twice For the transfer relationship between the object of action that is related to.
12. method according to claim 11, it is characterised in that described according to business scenario to be recommended, from it is described at least In a kind of recommending data model, the reference data needed for the business scenario to be recommended is obtained, including：
When the business scenario to be recommended is that the goal behavior object searched for according to targeted customer is pushed away to the targeted customer When recommending object of action, centering is shifted from least one described object of action, is obtained comprising the goal behavior object and from described Goal behavior object turns at least one candidate's object of action transfer pair of another object of action；
The reference data according to needed for the business scenario to be recommended carries out information recommendation to the business scenario to be recommended, Including：
Each candidate's object of action transfer pair of centering is shifted at least one described candidate's object of action, according to the target line The number of times of another object of action is turned to for object, calculate candidate's object of action transfer to transition probability；
According to each candidate's object of action shift to transition probability, from least one described candidate's object of action transfer centering choosing The transfer pair of goal behavior object is selected, the goal behavior object is shifted into behavior pair of the centering in addition to the goal behavior object As recommending the targeted customer.
13. a kind of information recommending apparatus, it is characterised in that including：
Real-time acquisition module, for according to the universal data collection specification suitable for each business scenario, to the use in network system Family behavior is gathered in real time, to obtain active user behavioral data；
Model computation module, for according to the general data processing logic suitable for each business scenario, to the active user row Analyzed for data, to obtain at least one recommending data model；
Parameter extraction module, for according to business scenario to be recommended, from least one recommending data model, obtains described Reference data needed for business scenario to be recommended；
Information recommendation module, for the reference data according to needed for the business scenario to be recommended to the business scenario to be recommended Carry out information recommendation.
14. device according to claim 13, it is characterised in that the real-time acquisition module specifically for：
The service path where service identification, the network object mark, network object that the user behavior is related to is gathered in real time, used Family is identified and user behavior type, to form the active user behavioral data.
15. device according to claim 13, it is characterised in that the model computation module includes：
Data resolution unit, for from two dimensions of user behavior type and object type, to the active user behavioral data Parsed, to obtain user behavior data of each object of action under various user behavior types, wherein, the object of action The section on service path where the network object that the network object or the user behavior being related to for the user behavior are related to Point；
Model treatment unit, for the information recommendation demand according to each business scenario, to each object of action in various users User behavior data under behavior type is analyzed and processed, to obtain at least one recommending data model.
16. device according to claim 15, it is characterised in that the data resolution unit specifically for：
The service identification being related to according to the user behavior, will be included in one including the active user behavioral data that identical services are identified Individual packet；
From two dimensions of user behavior type and object type, the active user behavioral data in each packet is solved Analysis, to obtain user behavior data of each object of action under various user behavior types in each packet.
17. device according to claim 15, it is characterised in that the model treatment unit specifically for：
Turned according to network object recommended requirements, user preference recommended requirements, network object temperature recommended requirements and network object At least one of shifting relation recommended requirements recommended requirements, to user of each object of action under various user behavior types Behavioral data is analyzed and processed, to obtain the basic recommending data model of network object, user preference recommending data model, user Historical behavior recommending data model, network object temperature recommending data model and network object transfer relationship recommending data model At least one of recommending data model.
18. device according to claim 17, it is characterised in that the model treatment unit is obtaining the network object During basic recommending data model, specifically for：
From the dimension of network object, to user behavior data carry out group of each object of action under various user behavior types Management is knitted, to obtain the basic data of each network object.
19. device according to claim 17, it is characterised in that the model treatment unit is obtaining the user preference During recommending data model, specifically for：
From user, three dimensions of network object and user behavior type, to each object of action in various user behavior classes User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each object of action Preference-score.
20. device according to claim 17, it is characterised in that the model treatment unit is obtaining the network object During temperature recommending data model, specifically for：
From user, three dimensions of network object and user behavior type, to each object of action in various user behavior classes User behavior data under type is counted, and obtains each user and each object of action occurs time of each type of user behavior Number；
The number of times of each type of user behavior is occurred to each object of action according to each user, each user is calculated to each object of action Preference-score；
According to preference-score of each user to each object of action, the hot value of each object of action is calculated.
21. the device according to claim 19 or 20, it is characterised in that the model treatment unit specifically for：
To each user, each object of action occurs in each time window the number of times of each type of user behavior to the user Summation is weighted, to obtain preference-score of the user to each object of action in each time window；
According to time decay factor, the user is weighted in each time window to the preference-score of each object of action and asked With to obtain preference-score of the user to each object of action.
22. device according to claim 21, it is characterised in that the model treatment unit specifically for：
To each object of action, each user is added up to the preference-score of the object of action, to obtain the behavior pair The hot value of elephant.
23. the device according to claim any one of 17-20, it is characterised in that the model treatment unit is obtaining institute When stating network object transfer relationship recommending data model, specifically for：
The dimension shifted from object, unites to user behavior data of each object of action under various user behavior types Meter, to obtain the transfer pair of at least one object of action；Object of action transfer is to representing before and after same user user's row twice For the transfer relationship between the object of action that is related to.
24. device according to claim 23, it is characterised in that the parameter extraction module specifically for：
When the business scenario to be recommended is that the goal behavior object searched for according to targeted customer is pushed away to the targeted customer When recommending object of action, centering is shifted from least one described object of action, is obtained comprising the goal behavior object and from described Goal behavior object turns at least one candidate's object of action transfer pair of another object of action；
Described information recommending module specifically for：
Each candidate's object of action transfer pair of centering is shifted at least one described candidate's object of action, according to the target line The number of times of another object of action is turned to for object, calculate candidate's object of action transfer to transition probability；
According to each candidate's object of action shift to transition probability, from least one described candidate's object of action transfer centering choosing The transfer pair of goal behavior object is selected, the goal behavior object is shifted into behavior pair of the centering in addition to the goal behavior object As recommending the targeted customer.
CN201610145260.7A 2016-03-15 2016-03-15 Information recommendation method and device Pending CN107193831A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610145260.7A CN107193831A (en) 2016-03-15 2016-03-15 Information recommendation method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610145260.7A CN107193831A (en) 2016-03-15 2016-03-15 Information recommendation method and device

Publications (1)

Publication Number Publication Date
CN107193831A true CN107193831A (en) 2017-09-22

Family

ID=59871527

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610145260.7A Pending CN107193831A (en) 2016-03-15 2016-03-15 Information recommendation method and device

Country Status (1)

Country Link
CN (1) CN107193831A (en)

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108198007A (en) * 2018-02-08 2018-06-22 王四春 A kind of anti-cross-border electric business commercial affairs big data decision and analysis system divulged a secret
CN108346084A (en) * 2017-12-22 2018-07-31 广东鸿威国际会展集团有限公司 A kind of behavior prediction system and method virtually shown for 3D
CN109041005A (en) * 2018-08-06 2018-12-18 佛山市苔藓云链科技有限公司 A kind of system and method for communication path selection in control communication network
CN110019671A (en) * 2017-11-07 2019-07-16 北京京东尚科信息技术有限公司 A kind of method and system handling real-time messages
CN110020360A (en) * 2017-11-09 2019-07-16 北京京东尚科信息技术有限公司 The method that user behavior characteristics are extracted, system and server
CN110059967A (en) * 2019-04-23 2019-07-26 北京相数科技有限公司 A kind of data processing method and device applied to city decision Analysis
CN110119474A (en) * 2018-05-16 2019-08-13 华为技术有限公司 Recommended models training method, the prediction technique based on recommended models and device
CN110766493A (en) * 2018-07-26 2020-02-07 阿里巴巴集团控股有限公司 Business object providing method, server, electronic device and storage medium
CN111241394A (en) * 2020-01-07 2020-06-05 腾讯科技（深圳）有限公司 Data processing method and device, computer readable storage medium and electronic equipment
CN111241381A (en) * 2018-11-28 2020-06-05 北京奇虎科技有限公司 Information recommendation method and device, electronic equipment and computer-readable storage medium
CN111897861A (en) * 2020-06-30 2020-11-06 苏宁金融科技(南京)有限公司 Content recommendation method and device, computer equipment and storage medium
CN112579033A (en) * 2019-09-27 2021-03-30 阿里巴巴集团控股有限公司 Business object switching method and device, electronic equipment and computer storage medium
CN114528482A (en) * 2022-01-25 2022-05-24 北京三快在线科技有限公司 Method and device for determining recommended object, electronic equipment and storage medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1967533A (en) * 2006-07-17 2007-05-23 北京航空航天大学 Gateway personalized recommendation service method and system introduced yuan recommendation engine
CN101616356A (en) * 2009-07-31 2009-12-30 卓望数码技术（深圳）有限公司 A kind of wireless data service product information method for pushing and system
CN102208086A (en) * 2010-03-31 2011-10-05 北京邮电大学 Field-oriented personalized intelligent recommendation system and implementation method
CN105243143A (en) * 2015-10-14 2016-01-13 湖南大学 Recommendation method and system based on instant voice content detection

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1967533A (en) * 2006-07-17 2007-05-23 北京航空航天大学 Gateway personalized recommendation service method and system introduced yuan recommendation engine
CN101616356A (en) * 2009-07-31 2009-12-30 卓望数码技术（深圳）有限公司 A kind of wireless data service product information method for pushing and system
CN102208086A (en) * 2010-03-31 2011-10-05 北京邮电大学 Field-oriented personalized intelligent recommendation system and implementation method
CN105243143A (en) * 2015-10-14 2016-01-13 湖南大学 Recommendation method and system based on instant voice content detection

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110019671A (en) * 2017-11-07 2019-07-16 北京京东尚科信息技术有限公司 A kind of method and system handling real-time messages
CN110019671B (en) * 2017-11-07 2022-04-12 北京京东尚科信息技术有限公司 Method and system for processing real-time message
CN110020360A (en) * 2017-11-09 2019-07-16 北京京东尚科信息技术有限公司 The method that user behavior characteristics are extracted, system and server
CN108346084A (en) * 2017-12-22 2018-07-31 广东鸿威国际会展集团有限公司 A kind of behavior prediction system and method virtually shown for 3D
CN108198007A (en) * 2018-02-08 2018-06-22 王四春 A kind of anti-cross-border electric business commercial affairs big data decision and analysis system divulged a secret
CN110119474A (en) * 2018-05-16 2019-08-13 华为技术有限公司 Recommended models training method, the prediction technique based on recommended models and device
CN110119474B (en) * 2018-05-16 2022-10-18 华为技术有限公司 Recommendation model training method, prediction method and device based on recommendation model
CN110766493A (en) * 2018-07-26 2020-02-07 阿里巴巴集团控股有限公司 Business object providing method, server, electronic device and storage medium
CN110766493B (en) * 2018-07-26 2023-04-28 阿里巴巴集团控股有限公司 Service object providing method, server, electronic device, and storage medium
CN109041005A (en) * 2018-08-06 2018-12-18 佛山市苔藓云链科技有限公司 A kind of system and method for communication path selection in control communication network
CN111241381A (en) * 2018-11-28 2020-06-05 北京奇虎科技有限公司 Information recommendation method and device, electronic equipment and computer-readable storage medium
CN110059967A (en) * 2019-04-23 2019-07-26 北京相数科技有限公司 A kind of data processing method and device applied to city decision Analysis
CN112579033A (en) * 2019-09-27 2021-03-30 阿里巴巴集团控股有限公司 Business object switching method and device, electronic equipment and computer storage medium
CN111241394A (en) * 2020-01-07 2020-06-05 腾讯科技（深圳）有限公司 Data processing method and device, computer readable storage medium and electronic equipment
CN111241394B (en) * 2020-01-07 2023-09-22 腾讯科技（深圳）有限公司 Data processing method, data processing device, computer readable storage medium and electronic equipment
CN111897861A (en) * 2020-06-30 2020-11-06 苏宁金融科技(南京)有限公司 Content recommendation method and device, computer equipment and storage medium
CN114528482A (en) * 2022-01-25 2022-05-24 北京三快在线科技有限公司 Method and device for determining recommended object, electronic equipment and storage medium

Similar Documents

Publication Publication Date Title
CN107193831A (en) Information recommendation method and device
CN109685631B (en) Personalized recommendation method based on big data user behavior analysis
CN105447186B (en) A kind of user behavior analysis system based on big data platform
KR100786795B1 (en) Internet advertising service system and method thereof
CN107665444A (en) A kind of web advertisement acute effect evaluation method and system based on the online behavior of user
CN106803190A (en) A kind of ad personalization supplying system and method
CN104050187B (en) Search result methods of exhibiting and system
CN108009897A (en) A kind of real-time recommendation method of commodity, system and readable storage medium storing program for executing
CN106227832A (en) The Internet big data technique framework application process in operational analysis in enterprise
CN107077486A (en) Affective Evaluation system and method
CN106447463A (en) Commodity recommendation method based on Markov decision-making process model
CN104835057A (en) Method and device for obtaining consumption feature information of network user
CN106445905B (en) Question and answer data processing, automatic question-answering method and device
CN109447713A (en) A kind of recommended method and device of knowledge based map
CN103345695A (en) Commodity recommendation method and device
CN102254028A (en) Personalized commodity recommending method and system which integrate attributes and structural similarity
CN106469392A (en) Select and recommend to show the method and device of object
CN107103062A (en) A kind of webpage recommending method and system
CN107808300A (en) A kind of advertisement push system and method for pushing based on wireless domain marketing platform
CN106327211A (en) SCRM system based on social media and development method thereof
CN106326318A (en) Search method and device
CN108090807A (en) Information recommendation method and device
CN110232591A (en) A kind of page promotion method, page extension system and server
KR101959808B1 (en) On-line Integrated Management System
CN110807691B (en) Cross-commodity-class commodity recommendation method and device

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: 20170922