Summary of the invention
The main purpose of the disclosure is that the touching for providing a kind of information that can be improved push reaches the information push side of efficiency
Case.
According to the first aspect of the disclosure, a kind of information-pushing method is provided, comprising: obtain the relevant letter of terminal
Breath;Based on the relevant information of the terminal, the current scene of the terminal is identified;And it is based on the current scene, Xiang Suoshu
Terminal sending business information.
Preferably, the relevant information of the terminal includes the information of many aspects obtained in several ways.
Preferably, the relevant information of the terminal may include the terminal status information and/or user are for described
The operation information that terminal executes.
Preferably, the status information may include the time, network insertion type, location information, model, system type,
Any one or more and/or described operation letter in electricity, memory headroom, operational mode and current operation application
Breath may include that the first operation information of the pre-installation application execution that user is directed in the terminal and/or user are directed to the end
The second operation information that third-party application on end executes.
Preferably, the step of obtaining terminal relevant information may include: the terminal correlation for receiving and uploading from the terminal
Information.
Preferably, the step of current scene of the identification terminal may include: to carry out to the relevant information of the terminal
Parsing, to obtain the status information under one or more characteristic dimensions;Based on the status information that parsing obtains, the terminal is determined
Current scene.
Preferably, the status information may include one or more of characteristic dimension: time;Position;The SOT state of termination;
Behavior type;Behavior preference.
Preferably, the step of sending business information to terminal may include: current scene based on the terminal and
The customized information of the active user of the terminal, the determination business information to be pushed;And it is determined to terminal push
Business information.
Preferably, the step of determination business information to be pushed may include: the current scene based on the terminal
With the customized information of the active user of the terminal, the multiple candidate Traffic information for being suitable for pushing to the terminal are determined;Meter
Calculate the priority of each candidate Traffic information in the multiple candidate Traffic information;It, will according to the sequence that priority is descending
The candidate Traffic information of predetermined quantity in the top is determined as the business information pushed to the terminal.
Preferably, this method can also include: to push the business information to the terminal in the judgement current push period
Number whether more than the first predetermined threshold;Determining that currently pushing the number pushed in the period is less than the first predetermined threshold
In the case of, Xiang Suoshu terminal sending business information.
Preferably, this method can also include: at the time of judgement the last time pushing the business information to the terminal
Whether the time between current time is more than the second predetermined threshold;In the case where determining that the time is more than the second predetermined threshold,
To the terminal sending business information.
Preferably, it is described to the step of terminal sending business information can also include: using in advance it is trained pre-
Model is surveyed, predicts operating result of the user to the business information pushed based on the current scene;Operating result based on prediction,
Determine whether the terminal sending business information.
Preferably, the operating result may include: to click result and/or conversion results;Or it clicks probability and/or turns
Change probability.
According to the second aspect of the disclosure, a kind of model training method is additionally provided, comprising: the one or more instructions of construction
Practice sample, wherein the training sample is characterized in the scene identified to the relevant information of terminal, the trained sample
This label is operating result of the user to the business information pushed based on the scene;Use one or more trained samples
This training prediction model, wherein the prediction model is for predicting user to the business information pushed based on the current scene
Operating result.
Preferably, the scene includes the status information under one or more characteristic dimensions.
Preferably, the status information may include one or more of characteristic dimension: time;Position;The SOT state of termination;
Behavior type;Behavior preference.
Preferably, the operating result may include: to click result and/or conversion results;Or it clicks probability and/or turns
Change probability.
In terms of according to the third of the disclosure, a kind of information push-delivery apparatus is additionally provided, comprising: information acquisition device is used
In the relevant information of acquisition terminal;Scene Recognition device identifies working as the terminal for being based on the relevant information of the terminal
Preceding scene;And information push-delivery apparatus, for being based on the current scene, Xiang Suoshu terminal sending business information.
Preferably, the relevant information of the terminal includes the information of many aspects obtained in several ways.
Preferably, the relevant information of the terminal includes that the status information of the terminal and/or user are directed to the terminal
The operation information of execution.
Preferably, the status information include the time, network insertion type, location information, model, system type, electricity,
Memory headroom, operational mode and the current any one or more and/or described operation information run in application include
User is for the first operation information of the pre-installation application execution in the terminal and/or user for the third in the terminal
Second operation information of square application execution.
Preferably, information acquisition device can receive the relevant information of terminal uploaded from terminal.
Preferably, the scene Recognition device may include: parsing module, for carrying out to the relevant information of the terminal
Parsing, to obtain the status information under one or more characteristic dimensions;First determining module, the state for being obtained based on parsing
Information determines the current scene of the terminal.
Preferably, the status information may include one or more of characteristic dimension: time;Position;The SOT state of termination;
Behavior type;Behavior preference.
Preferably, the information push-delivery apparatus may include: the second determining module, for working as front court based on the terminal
The customized information of the active user of scape and the terminal, the determination business information to be pushed;And pushing module, it is used for institute
It states terminal and pushes identified business information.
Preferably, second determining module may include: candidate information determining module, for working as based on the terminal
The customized information of the active user of preceding scene and the terminal determines the multiple candidate Traffics letter for being suitable for pushing to the terminal
Breath;Score value computing module, for calculating the priority of each candidate Traffic information in the multiple candidate Traffic information;Sequence is true
The candidate Traffic information of predetermined quantity in the top is determined as by cover half block for the sequence descending according to priority
The business information pushed to the terminal.
Preferably, which can also include: the first judgment means, for judging currently to push in the period to the terminal
The number of the business information is pushed whether more than the first predetermined threshold, determines the current push period in first judgment means
In the case that the number of interior push is less than the first predetermined threshold, the information push-delivery apparatus is believed to the terminal transmission service
Breath.
Preferably, which can also include: the second judgment means, the last to terminal push institute for judging
Whether time at the time of stating business information between current time is sentenced more than the second predetermined threshold in second judgment means
In the case where fixing time more than the second predetermined threshold, the information push-delivery apparatus is to the terminal sending business information.
Preferably, which can also include: prediction meanss, for using preparatory trained prediction model, prediction to be used
Operating result of the family to the business information pushed based on the current scene, operation knot of the information push-delivery apparatus based on prediction
Fruit, it is determined whether Xiang Suoshu terminal sending business information.
Preferably, the operating result may include: to click result and/or conversion results;Or it clicks probability and/or turns
Change probability.
According to the 4th of the disclosure the aspect, a kind of model training apparatus is additionally provided, comprising: constructing apparatus is used for structure
Make one or more training samples, wherein what the training sample was characterized in identifying the relevant information of terminal
Scene, the label of the training sample are operating result of the user to the business information pushed based on the scene;Training device,
For using one or more training sample training prediction models, wherein the prediction model is for predicting user to base
In the operating result of the business information of current scene push.
Preferably, the scene includes the status information under one or more characteristic dimensions.
Preferably, the status information includes one or more of characteristic dimension: time;Position;The SOT state of termination;Behavior
Type;Behavior preference.
Preferably, the operating result may include: to click result and/or conversion results;Or it clicks probability and/or turns
Change probability.
According to the 5th of the disclosure the aspect, a kind of calculating equipment is additionally provided, comprising: processor;And memory,
On be stored with executable code, when the executable code is executed by the processor, the processor is made to execute the disclosure
First aspect or the method addressed of the second aspect.
According to the 6th of the disclosure the aspect, a kind of non-transitory machinable medium is additionally provided, is stored thereon
There is executable code, when the executable code is executed by the processor of electronic equipment, the processor is made to execute the disclosure
First aspect or the method addressed of the second aspect.
Business information using the information push scheme push of the disclosure is to be determined according to current scene, therefore push
Business information can better meet the current individual demand of user.
Specific embodiment
The preferred embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Preferred embodiment, however, it is to be appreciated that may be realized in various forms the disclosure without the embodiment party that should be illustrated here
Formula is limited.On the contrary, these embodiments are provided so that this disclosure will be more thorough and complete, and can be by the disclosure
Range is completely communicated to those skilled in the art.
[term parsing]
The abbreviation of POI, Point of Interest, can translate into " information point ", and each POI may include four aspects
Information, the information such as title, classification, latitude, longitude, neighbouring retail shop, restaurant, hotel.
LBS, location based service, it be by the radio communication network of telecommunications mobile operator (such as GSM net,
CDMA net) or external positioning method (such as GPS) obtain the location information (geographical coordinate or geodetic coordinates) of mobile terminal user,
In the support of GIS-Geographic Information System (foreign language abbreviation: GIS, foreign language full name: Geographic Information System) platform
Under, a kind of value-added service of respective service is provided for user.
Push, the message push of mobile terminal.
The abbreviation of TT, TimeTunnel, a kind of acquisition integrating log, transmission, storage, monitoring log management system
System example.
The abbreviation of SLS, Simple Log Service, simple log services, one kind being directed to log collection, storage, inquiry
With the service system example of analysis.
MetaQ, a message-oriented middleware.
Galaxy, general streaming computing (Stream Computing) platform provide (minute grade, second grade and hundred in real time
Millisecond delay) stream data calculate service.
Blink, a streaming computing frame of extension, optimization based on flink Open Source Framework.
[plan explanation]
The disclosure, which mainly proposes one kind, can make the information of push effectively touch the information push scheme up to user.The disclosure
Core concept be that, by collecting and the relevant information of analysing terminal, the current scene of identification terminal is with the scene identified
The foundation of information push, the business information to match to terminal user's push with current scene.Since the business information of push is
It is determined according to current scene, therefore the business information pushed can better meet the current individual demand of user.
Further, the relevant information of terminal that the disclosure is addressed may include the real-time behavior letter of terminal that terminal generates
Breath, i.e. user are directed to the real-time operation information that terminal executes.The information push scheme of the disclosure is with user for terminal as a result,
The real-time operation information of execution is the triggering factor of information push, therefore the user on the opportunity and terminal that information can be made to push
Use state is consistent, and then the flow distribution that information can be made to push is consistent with the non-human act distribution trend of user, avoids
The problem of flow concentrates on certain several fixed period.
Further, the disclosure information push scheme can be realized by server-side, it can by server-side receive from
The relevant information of terminal that terminal (i.e. client) uploads.It gives the correct time in data when having, shows that terminal is in connected state, therefore,
The triggering factor that is pushed by the relevant information of the terminal uploaded using terminal as information of information push scheme of the disclosure, can be with
The process for avoiding waiting for user's networking, realizes the real-time push of information.
To sum up, scheme is pushed using the information of the disclosure, can be effectively touched on push opportunity and two aspect of push content
Up to user, so as to improve the push experience of user.
The information push scheme that the embodiment of the present disclosure provides can be applied in environment as shown in Figure 1.Fig. 1 is for real
The schematic diagram of the environment of the existing embodiment of the present disclosure.In one embodiment, the terminal 10 in environment can be realized via network 40
With the information transmit-receive of server 20.Server 20 can store the data obtained from terminal 10 to database 30, can also lead to
Access database 30 is crossed come content needed for obtaining terminal 10.Between terminal 10 (for example, between 10_1 and 10_2 or 10_N)
It can communicate with one another via network 40.Network 40 can be it is sensu lato for information transmitting network, may include one or
Multiple communication networks, such as cordless communication network, internet, private domain net, local area network, Metropolitan Area Network (MAN), wide area network or cellular data
Network etc..In one embodiment, network 40 also may include satellite network, thus send the GPS signal of terminal 10 to clothes
Business device 20.It should be noted that will not change example of the invention if adding into Fig. 1 or removing add-on module from Fig. 1 and implement
The underlying concept of example.In addition, though being shown in figure the two-way arrow from database 30 to server 20 for convenience of explanation
Head, but it will be appreciated by those skilled in the art that it is upper, the transmitting-receiving of above-mentioned data is also that can be realized by network 40.
Terminal 10 can be any suitable electronic equipment that can be utilized for network access, such as can be portable electronic
Equipment, including but not limited to smart phone, tablet computer or other portable terminals.Server 20 is then that can pass through network
Any server of the offer interactive service information needed of access.Though it is shown that multiple terminal 10-1 ... N and single in figure
Server 20 and database 30, and can select in the description that follows one of those or part terminal described (for example,
Terminal 10-1), it should be understood by those skilled in the art that above-mentioned 1 ..., N number of terminal, which is intended to indicate that in live network, exists
Multiple terminals, the individual server 20 and database 30 shown be intended to indicate that technical solution of the present invention is related to server and number
According to the operation in library.Terminal and individual server and database to particular number are described in detail to illustrate convenience, and
Non- imply has limitation to the type of terminal and server or position etc..
Fig. 2 shows the schematic flow charts according to the information-pushing method of the disclosure one embodiment.Wherein Fig. 2 shows
Information-pushing method can be realized by the terminal 10 in Fig. 1, can also be realized by server 20, can also be by server 20 and end
10 cooperation of end is realized.Preferably, it can be realized by server 20.
Referring to fig. 2, in step S210, the relevant information of terminal is obtained.
The step of wherein obtaining terminal relevant information can be executed by server, such as can be received by server from end
The relevant information of terminal for holding (i.e. client) to upload.The terminal that the disclosure is addressed preferably refers to mobile phone, IPAD, smartwatch
Equal portable terminals (such as mobile terminal).
The relevant information of acquired terminal may include the information of many aspects obtained in several ways, such as can be with
It is many levels, the information of multiple fields of the terminal obtained in different ways.For example, terminal shape can be obtained in different ways
State information, user for the behavioural information of system-level app, user for the different levels such as the behavioural information of third party app across
Realm information obtains the relevant information of terminal.
The relevant information of the terminal that the disclosure is addressed may include that the status information of terminal and user execute for terminal
Operation information.Status information can include but is not limited to the time, network insertion type (WIFI, 2G, 3G, 4G etc.), location information,
Model (type of such as terminal), system type, electricity, memory headroom (can be in current residual memory headroom or current occupancy
Deposit space), operational mode (such as standby mode, battery saving mode, saves flow rate mode at offline mode) and current operation application
(may include that front stage operation is applied and running background application) etc..
The operation information that user executes for terminal may include user for the of the pre-installation application execution in terminal
The second operation information that one operation information and user execute for the third-party application in terminal.
The third-party application that the disclosure is addressed refers in particular to the application independently installed by terminal user with pre-installation application opposition.
Pre-installation application may include system-level application and the application of other pre-installations.Wherein, system-level application can include but is not limited to
Application market, calendar, information, contact person, clock, recorder etc., other pre-installation applications can be by producer and/or dealer
The application pre-installed at the terminal.
As an alternative embodiment of the disclosure, the obtaining step of the relevant information of terminal can be executed by server,
The relevant information of terminal can be obtained from terminal by server.It, can be by installing monitoring at the terminal in specific implementation
Action message (i.e. the relevant information of terminal) on software supervision collection terminal, collected information can be uploaded to server,
So that information of the server to collection parses.
In step S220, it is based on the relevant information of terminal, the current scene of identification terminal.
After obtaining the relevant information of terminal, the relevant information of terminal can be parsed, it is one or more to obtain
Status information under characteristic dimension.Such as time, position, the SOT state of termination, behavior type, Behavior preference can be obtained by parsing
Etc. status information under various features dimension.Wherein, the SOT state of termination can include but is not limited to electricity, memory headroom, operation mould
One of multiple terminals states such as formula, current operation application are a variety of.Behavior type and Behavior preference can be by users
The operation information executed for terminal parses to obtain, for example, can determine behavior when user executes search operation at the terminal
Type is search behavior, repeatedly opens tourism in user in application, can determine that Behavior preference is tourism favor.
As one embodiment of the disclosure, in the feelings for obtaining the relevant information of a plurality of terminal from multiple terminals by server
Under condition, user id (i.e. the terminal i d), User Status, behavior type, Behavior preference, terminal shape of every information can be parsed out
State, time (timestamp), position etc..
Status information under the one or more characteristic dimensions obtained according to parsing, can further identify working as terminal
Preceding scene.For example, according to the current location information parsed to the relevant information of terminal, the commercial circle as locating for active user,
The location informations such as shopping center or POI can identify current LBS scene;It is searched according to user what the desktop of terminal executed
Suo Hangwei, can identify current scene for search scene, further, according to execute search behavior used in search term,
It can further identify the search need scene that can show the current search need of user;According to user to third-party application
Service condition, can identify the Behavior preference scene of user, such as user repeatedly opens tourism application in a short time,
It can then determine that current scene is tourism favor scene;According to timestamp and location information, when capturing user in the quitting time
When going back home middle, it can identify that current scene is amusement and leisure scene.According to the relevant information of terminal, other can also be determined
Several scenes, details are not described herein again.
In step S230, it is based on current scene, to terminal sending business information.
The current scene identified can characterize terminal user in current individual demand to a certain extent, therefore can
To determine the business information for being suitable for pushing to terminal user based on current scene, so that can to the business information that terminal pushes
The current Personalized demand for bettering meet user improves the push experience of user.
For example, it is raw the periphery based on LBS can be pushed to terminal in the case where identifying current scene is LBS scene
Information on services living;It, can be to search term phase terminal push and extracted in the case where identifying current scene to search for scene
The commercialization content of pass;In the case where identifying current scene is tourism favor scene, can targetedly bottom to terminal
Active user pushes the information of tourism or the business promotion information of tourism;It is returned to identifying that current scene is come off duty for user
When in family, relevant entertainment information can be pushed with user.
So far Fig. 2 is combined to be described with regard to the basic implementation process of the information push scheme of the disclosure.Below to the disclosure
The other aspects that are related to of technical solution be described respectively.
[determination of push content]
In the disclosure, based on identified current scene, the customized information of the active user of terminal can be combined (such as
Historical behavior, real-time behavior, for a long time with short-term interest etc.) the determination business information to be pushed (i.e. push content).
As an alternative embodiment of the disclosure, can the customized information based on current scene and active user, really
The fixed multiple candidate Traffic information for being suitable for pushing to terminal, wherein the quantity of candidate Traffic information can be set according to demand, and one
As tens between several hundred.For multiple candidate Traffic information, the priority of each candidate Traffic information can be calculated, according to
The candidate Traffic information of predetermined quantity in the top is determined as the business pushed to terminal by the descending sequence of priority
Information.
The priority of candidate Traffic information can be specifically calculated in several ways.It is, for example, possible to use machine learning moulds
Type, many factors such as portrait, interest, the behavioural characteristic of synthetic user carry out fining marking to each candidate Traffic information,
To determine its priority.Wherein, machine learning model can be for estimating user under current scene to the point of business information
The prediction model for hitting probability and/or transition probability can be what training in advance obtained.Machine learning model prediction can be used
For user to the click probability and/or transition probability of candidate Traffic information under current scene, it is preferential which can be used as it
The measurement factor of grade.And the realization process given a mark according to information such as the portrait of user, interest, behavioural characteristics to business information
Known to those skilled in the art, details are not described herein again.
[appropriateness push]
In order to avoid overdriven generates harassing and wrecking to user, for each terminal, can set in the single push period
The push times upper limit, or setting push interval.For example, the push period can be one day, can set daily as same terminal
In addition the maximum number of times (such as 8 times) of push can also set the adjacent time interval pushed twice executed for same terminal,
As the adjacent time interval pushed twice can be set as not less than 1 hour.
In the case where setting the push times upper limit for user, to before terminal sending business information, it can be determined that current
Push the period in terminal sending business information number whether be more than setting maximum number of times (for the ease of distinguish, claim herein
For the first predetermined threshold), in the case where determining that number is less than the first predetermined threshold, just to terminal sending business information.
In the threshold value for setting the adjacent time interval pushed twice for user, (for the ease of distinguishing, referred to herein as second is pre-
Determine threshold value) in the case where, to before terminal sending business information, it can be determined that it is the last to terminal sending business information when
Carve the time between current time whether more than the second predetermined threshold, in the case where determining more than the second predetermined threshold,
To terminal sending business information.
It preferably, can be after the relevant information of terminal (i.e. step S210) for getting terminal, to the current push period
Interior push times or adjacent push interval judged, in the case where meeting pushing condition under, just execute subsequent step
S220、S230。
[push prediction]
In order to improve the probability that the business information of push is clicked and/or converts, the push experience of user is improved.To with
Before the sending business information of family, it is current to being based on to can use trained prediction model (machine learning model) prediction user in advance
The operating result of the business information of scene push, the operating result based on prediction determine whether push away to terminal for current scene
Send business information.
The training process of prediction model is as follows.
(1) training sample is constructed.
Training sample is characterized in the scene identified to the relevant information of terminal, and the label of training sample is to use
Operating result of the family to the business information pushed based on scene.
Scene may include the status information under one or more characteristic dimensions, that is to say, that the feature of training sample can
To be the status information under one or more characteristic dimensions.For example, can include but is not limited to time, position, the SOT state of termination, row
For the status information under the characteristic dimensions such as type, Behavior preference.Wherein, about the relevant information of terminal, the SOT state of termination, behavior class
Type, Behavior preference may refer to associated description above, and details are not described herein again.
In the disclosure, it can be multiple, therefore each instruction to the business information that terminal user pushes based on Same Scene
The label (i.e. operating result) for practicing sample can be user respectively to each industry in the multiple business information pushed based on scene
The statistical result for the operating result that information of being engaged in executes, and the statistical result can be the final click result obtained by statistics
And/or conversion results, it is also possible to the click probability obtained by statistics and/or transition probability.
That is, operating result, which can be, clicks result and/or conversion results, it is also possible to click probability and/or turns
Change probability.Wherein, click result refer to user to the click of the business information pushed based on scene as a result, include " click " and
" not clicking " two kinds of situations.Conversion results refer to based on scene push business information be converted as a result, include " conversion " and
" unconverted " two kinds of situations, the conversion that the disclosure is addressed refer to that the business information of push is easily accepted by a user, such as with the industry of push
Be engaged in information be Account Registration information for, if user's registration account, then it is assumed that the information is easily accepted by a user, business information conversion
Success, if the unregistered account of user, then it is assumed that the information is not easily accepted by a user, business information conversion failure.Clicking probability is
Refer to user to the click probability of the business information pushed based on scene, transition probability refers to user to the information pushed based on scene
Transition probability.
(2) using training sample training prediction model
Training process is well known in the art, and details are not described herein again.The prediction model that final training obtains can be used for pre-
Survey operating result of the user to the business information pushed based on current scene.Wherein, operating result can be click result and/or
Conversion results are also possible to click probability and/or transition probability.
That is, the prediction model that final training obtains can be used for predicting user to the industry pushed based on a certain scene
Whether business information clicks or whether the business information is converted, and can be used for predicting for the industry pushed based on a certain scene
The probability that business information is clicked or converted.
Prediction model use process is as follows.
In prediction model for predicting user to the click probability of the business information pushed based on current scene and/or conversion
In the case where probability, it can use prediction model and the current scene determined based on step S220 predicted, predict user couple
The click probability and/or transition probability of business information based on current scene push.In the click probability of prediction and/or conversion
It, can not be to terminal sending business information in the lower situation of probability.
In prediction model for predicting user to the click result of the business information pushed based on current scene and/or conversion
As a result in the case where, it can use prediction model and the current scene determined based on step S220 predicted, predict user couple
The click result and/or conversion results of business information based on current scene push.In the click result of prediction and/or conversion
It as a result, can not be to terminal sending business information in the case where being no.
[application examples]
Fig. 3 shows a kind of implementation flow chart of the information push scheme of the disclosure.
It can be with real-time collecting customer traffic in step S310 referring to Fig. 3.
Specifically the real-time stream of multiple levels, multiple fields can be collected based on operating system level, such as can wrap
Include terminal's status information, user is directed to the behavior of third-party application for the behavior of system-level application and user.
The data flow that the present embodiment is addressed refers to that the user data constantly acquired over time, real-time stream are
Refer to the data flow acquired in real time.The real-time stream of collection is equivalent to the relevant information of the terminal addressed above and passes through in other words
Real-time stream is parsed, the relevant information of available terminal.
As an example, terminal's status information may include the network insertion type (WIFI, mobile network) of terminal, positioning letter
Breath, mobile phone model, system type, electricity, memory headroom etc.;User may include user for the behavior of system-level application and answer
Map etc. is used in main screen search, user with the movement in shop, user information stream browsing, user;User answers for third party
Behavior includes peace loading, unloading, starting, closing of third-party application etc..
Above-mentioned multiple cross-domain categorical datas can be buried by client a little or background service is according to the system arranged in advance
One data format is input to Log Administration System in real time, such as can be input in TT, SLS, MetaQ log channel.
In step S320, stream data parsing.
It can use streaming computing platform and dissection process carried out to multiple real-time streams that step S310 is collected.Such as it can
To use the real-time streamings computing platform such as galaxy, blink to realize.Process flow includes the following steps.
(1) galaxy access 1) in the channel SLS, MetaQ data, unified conversion being output to TT.
(2) blink accesses the data of TT, impressions and release time the last time of user is inquired, if user
Impressions reach the upper limit or the last release time and apart from current time be less than certain threshold value, then filter out this and ask
It asks.
(3) for the request not being filtered, user id, status information, the behavior type, time of every data are parsed
The contents such as stamp are used for subsequent module.
In step S330, scene analysis and decision.
User state information, the behavior type etc. that can be exported according to previous step parse the scene and up and down of user
Literary information is recalled different individualized contents according to different scenes for subsequent module and is pushed.
In step S340, content personalization is recalled.
It can be according to determining current scene, history and real-time behavior, for a long time and short-term interest, for different user and not
Relevant content is recalled with scene, which recalls for coarseness, and the set sizes recalled generally arrive between several hundred tens.Such as
Shown in Fig. 3, the content recalled can be a plurality of types of contents such as music, video, information, advertisement (i.e. business information).
In step S350, content fining sequence.
For the content recalled, machine learning model can be applied, the features such as portrait, interest, the behavior of synthetic user are right
Each content carries out fining marking, selects N number of content of highest scoring.
As shown in figure 3, can draw a portrait with reference to user during executing step S330, step S340, step S350
With user interest.
In step S360, content push.
Situation can be kept according to the connection of client and server-side, by N number of content push of highest scoring to terminal.
As it can be seen that the real-time scene based on terminal that the disclosure proposes carries out information push (including push, short message, Instant Messenger
News tool etc.) scheme can effectively identify the special scenes demand of user compared with traditional push scheme.Such as it is logical
Cross POI and relevant inquiring information identify user Shopping Behaviors scene, push the relevant favor information of periphery businessman;It is logical
The position zinformation and temporal information for crossing user identify the trip scene of user, push relevant music and press service.
This message based on scene pushes during actual operation, can effectively improve the degree of intelligence of content distribution
And user, to the acceptance level of content, reaching in efficiency and effect in touching has very big promotion.In practical commercialization push, based on this
Clicking rate/conversion ratio of the content (business information) of disclosed information push scheme push relative to traditional scheme improve 80% with
On, it is obviously improved in distribution efficiency and user experience.
The information push scheme of the disclosure can be used as information recommendation system and provide for the types of applications of installation at the terminal
Information recommendation service can also be used as information recommendation Module-embedding in installing specific application at the terminal.That is, this
The open service product that may be embodied as pushing based on push and short message, also may be embodied as the shifting there are push and short message service
Moved end system (such as Android, ios, millet, Huawei) is also embedded within the instant communication softwares such as wechat, QQ and exists more
In the mobile APP of product line.
[information push-delivery apparatus]
Fig. 4 shows the structural block diagram of the information push-delivery apparatus according to one embodiment of the disclosure.Wherein, each function in device
Energy module can be realized by the combination of the hardware of the realization principle of the invention, software or hardware and software.Those skilled in the art
It is understood that Fig. 4 described function module can combine or be divided into submodule, to realize foregoing invention
Principle.Therefore, description herein can support to functions described herein module it is any it is possible combination or division,
Or further restriction.The operation that the functional module and each functional module that only can have below with regard to device can execute
It is described briefly, above description may refer to for the detail section being directed to, which is not described herein again.
Referring to fig. 4, information push-delivery apparatus 400 includes that information acquisition device 410, scene Recognition device 420 and information push away
Send device 430.
Information acquisition device 410 for obtaining the relevant information of terminal, preferably information acquisition device 410 can receive from
The relevant information of terminal that terminal uploads.The relevant information of terminal may include the letter of many aspects obtained in several ways
Breath.For example, the relevant information of terminal may include the status information of terminal and/or user is directed to the operation information that terminal executes.
Status information includes time, network insertion type, location information, model, system type, electricity, memory headroom, fortune
It is any one or more in row mode and current operation application.
Operation information includes that user is directed to for the first operation information of the pre-installation application execution in terminal and/or user
The second operation information that third-party application in terminal executes.
Scene Recognition device 420 is used to be based on the relevant information of terminal, the current scene of identification terminal.
As shown in figure 4, scene Recognition device 420 can optionally include 421 He of parsing module shown in dotted line frame in figure
First determining module 423.
Parsing module 421 is for parsing the relevant information of terminal, to obtain under one or more characteristic dimensions
Status information.As an example, characteristic dimension can include but is not limited to one or more of: the time, position, the SOT state of termination,
Behavior type and Behavior preference.First determining module 423 is used for the status information obtained based on parsing, determines the current of terminal
Scene.
Information push-delivery apparatus 430 is used to be based on current scene, to terminal sending business information.
As shown in figure 4, information push-delivery apparatus 430 can optionally include the second determining module shown in dotted line frame in figure
431 and pushing module 433.
Customized information of second determining module 431 for the active user of current scene and terminal based on terminal, really
Surely the business information to be pushed.Pushing module 433 is used to push identified business information to terminal.
As shown in figure 4, the second determining module 431, which can optionally include candidate information shown in dotted line frame in figure, determines mould
Block 4311, score value computing module 4313 and sequence determining module 4315.
Personalized letter of the candidate information determining module 4311 for the active user of current scene and terminal based on terminal
Breath determines the multiple candidate Traffic information for being suitable for pushing to terminal.Score value computing module 4313 is for calculating multiple candidate Traffics
The priority of each candidate Traffic information in information.The determining module 4315 that sorts is used for the sequence descending according to priority,
The candidate Traffic information of predetermined quantity in the top is determined as to the business information pushed to terminal.
As shown in figure 4, information push-delivery apparatus 400 can also optionally include the first judgment means shown in dotted line frame in figure
440。
First judgment means 440 be used for judge currently push the period in terminal sending business information number whether be more than
First predetermined threshold determines that currently pushing the number pushed in the period is less than the first predetermined threshold in the first judgment means 440
In the case where, 430 ability of information push-delivery apparatus is to terminal sending business information.
As shown in figure 4, information push-delivery apparatus 400 can also optionally include the second judgment means shown in dotted line frame in figure
450。
Second judgment means 450 for judge it is the last to terminal sending business information at the time of and current time between
Time whether more than the second predetermined threshold, in the case where the second judgment means 450 determine that the time is more than the second predetermined threshold,
430 ability of information push-delivery apparatus is to terminal sending business information.
As shown in figure 4, information push-delivery apparatus 400 can also optionally include prediction meanss 460 shown in dotted line frame in figure.
Prediction meanss 460 are used for using preparatory trained prediction model, and prediction user based on current scene to being pushed
The operating result of business information, operating result of the information push-delivery apparatus 430 based on prediction, it is determined whether believe to terminal transmission service
Breath.Wherein operating result may include click result and/or conversion results, or may include clicking probability and/or conversion generally
Rate.
[model training apparatus]
Fig. 5 shows the structural block diagram of the model training apparatus according to one embodiment of the disclosure.Wherein, each function in device
Energy module can be realized by the combination of the hardware of the realization principle of the invention, software or hardware and software.Those skilled in the art
It is understood that Fig. 5 described function module can combine or be divided into submodule, to realize foregoing invention
Principle.Therefore, description herein can support to functions described herein module it is any it is possible combination or division,
Or further restriction.The operation that the functional module and each functional module that only can have below with regard to device can execute
It is described briefly, above description may refer to for the detail section being directed to, which is not described herein again.
As shown in figure 5, model training apparatus 500 includes constructing apparatus 510 and training device 520.
Constructing apparatus 510 is for constructing one or more training samples, wherein training sample is characterized in terminal correlation
The scene that is identified of information, the label of training sample is operation knot of the user to the business information pushed based on scene
Fruit,
Scene includes the status information under one or more characteristic dimensions.Status information includes one or more of feature
Dimension: time, position, the SOT state of termination, behavior type, Behavior preference.
Operating result may include click result and/or conversion results, or may include clicking probability and/or conversion generally
Rate.
Training device 520 is used for using one or more training sample training prediction models, wherein prediction model is for pre-
Survey operating result of the user to the business information pushed based on current scene.
[calculating equipment]
A kind of information-pushing method or model training method that can be used for executing the disclosure is additionally provided according to the disclosure
Calculating equipment.
Fig. 6 can be used for execute the disclosure information-pushing method or model training method calculating equipment it is schematic
Block diagram.
As shown in fig. 6, the calculating equipment 600 may include processor 610 and memory 620.It is stored on memory 620
Executable code.When processor 610 executes the executable code, so that processor 610 executes information as described above push
Method or model training method.
Above by reference to attached drawing be described in detail information according to the present invention push and model training method, device and
Calculate equipment.
In addition, being also implemented as a kind of computer program or computer program product, the meter according to the method for the present invention
Calculation machine program or computer program product include the calculating for executing the above steps limited in the above method of the invention
Machine program code instruction.
Alternatively, the present invention can also be embodied as a kind of (or the computer-readable storage of non-transitory machinable medium
Medium or machine readable storage medium), it is stored thereon with executable code (or computer program or computer instruction code),
When the executable code (or computer program or computer instruction code) by electronic equipment (or calculate equipment, server
Deng) processor execute when, so that the processor is executed each step according to the above method of the present invention.
Those skilled in the art will also understand is that, various illustrative logical blocks, mould in conjunction with described in disclosure herein
Block, circuit and algorithm steps may be implemented as the combination of electronic hardware, computer software or both.
The flow chart and block diagram in the drawings show the possibility of the system and method for multiple embodiments according to the present invention realities
Existing architecture, function and operation.In this regard, each box in flowchart or block diagram can represent module, a journey
A part of sequence section or code, a part of the module, section or code include one or more for realizing defined
The executable instruction of logic function.It should also be noted that in some implementations as replacements, the function of being marked in box can also
To be occurred with being different from the sequence marked in attached drawing.For example, two continuous boxes can actually be basically executed in parallel,
They can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram and/or stream
The combination of each box in journey figure and the box in block diagram and or flow chart, can the functions or operations as defined in executing
Dedicated hardware based system realize, or can realize using a combination of dedicated hardware and computer instructions.
Various embodiments of the present invention are described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or improvement to the technology in market for best explaining each embodiment, or make the art
Other those of ordinary skill can understand each embodiment disclosed herein.