CN108230007A - A kind of recognition methods of user view, device, electronic equipment and storage medium - Google Patents
A kind of recognition methods of user view, device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN108230007A CN108230007A CN201711219349.4A CN201711219349A CN108230007A CN 108230007 A CN108230007 A CN 108230007A CN 201711219349 A CN201711219349 A CN 201711219349A CN 108230007 A CN108230007 A CN 108230007A
- Authority
- CN
- China
- Prior art keywords
- user
- item
- user view
- numerical value
- characteristic item
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- 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
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- 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]
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Marketing (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
An embodiment of the present invention provides a kind of recognition methods of user view, device, electronic equipment and storage medium, which includes：Determining module, for determining multiple user views and multiple characteristic items；Extraction module, for extracting the character numerical value that each characteristic item corresponds to each user view respectively；First acquisition module, for obtaining user's history behavioral data；Training module, for carrying out model training according to character numerical value and user's history behavioral data, to obtain the weighted value of each characteristic item；Second acquisition module, for obtaining current multiple characteristic items, current multiple characteristic items are respectively provided with real-time character numerical value；It is intended to probability distribution computing module, for according to weighted value, summation is weighted to real-time character numerical value, to obtain the current probability distribution of multiple user views, it solves the problems, such as not reflecting that user is instantaneously inclined in the case of no inquiry in the prior art so that be comparable between each user view.
Description
Technical field
The present invention relates to Internet technical fields, anticipate more particularly to a kind of recognition methods of user view, a kind of user
Identification device, a kind of electronic equipment and a kind of storage medium of figure.
Background technology
With reaching its maturity for development of Mobile Internet technology, people can easily at any time, place by movement set
It is standby to access internet, and user is also increasingly personalized for the demand of information, the information requirement of Accurate Prediction user at this very moment,
I.e. active user is intended to, and is to realize a personalized importance.
Comprehensive search under existing O2O (Online To Offline, under line on online offline/line) scene is recommended
System is typically all regard the intention of user as feature by the historical behaviors such as click, place an order and add in learn order models, is passed through
The training of model obtains wanted business recommended to the user, but above-mentioned recommendation method is merely able to reflect user for certain one kind
The history preference of business.In addition, search system also relies on the input inquiry of user mostly other than using user's history behavior, lead to
Overmatching text input by user, obtains corresponding recommendation results.But it cannot but reflect well in the case of no inquiry
Go out the instantaneous tendency of user, for example, intention or tendency of the user when just starting search system.At the same time, multiple systems pair
Repeatedly access is needed in the use of user's history behavioural characteristic, increases the complexity of model calculating, while also increase system
The probability of error.
Invention content
In view of the above problems, it is proposed that the embodiment of the present invention overcomes the above problem or at least partly in order to provide one kind
A kind of recognition methods of user view, a kind of identification device of user view, a kind of electronic equipment and the phase to solve the above problems
A kind of storage medium answered.
To solve the above-mentioned problems, the embodiment of the invention discloses a kind of identification device of user view, including：
Determining module, for determining multiple user views and multiple characteristic items；
Extraction module, for extracting the character numerical value that each characteristic item corresponds to each user view respectively；
First acquisition module, for obtaining user's history behavioral data；
Training module, for carrying out model training according to the character numerical value and the user's history behavioral data, to obtain
Obtain the weighted value of each characteristic item；
Second acquisition module, for obtaining current multiple characteristic items, current multiple characteristic items are respectively provided with reality
When character numerical value；
It is intended to probability distribution computing module, for according to the weighted value, being weighted to the real-time character numerical value
Summation, to obtain the current probability distribution of the multiple user view.
Optionally, the multiple characteristic item includes temporal characteristics item, and the temporal characteristics item corresponds to each user view
Character numerical value be converted into temporal characteristics item hash function；And/or
The multiple characteristic item includes category feature item, and the category feature item corresponds to the characteristic of each user view
Value is converted into category feature item hash function；And/or
The multiple characteristic item includes timeliness-accumulation characteristic item, and the timeliness-accumulation characteristic item is anticipated corresponding to each user
The character numerical value of figure is converted into timeliness-accumulation characteristic item hash function.
Optionally, when the multiple characteristic item includes temporal characteristics item, the extraction module includes：
Submodule is divided, for unit period to be divided into multiple timeslices；
First statistic submodule, for count respectively each user view occurs in each timeslice number and it is whole when
Between each user view occurs in piece the maximum value of number；
First generation submodule, for occurred according to each user view in each timeslice number and whole when
Between each user view occurs in piece the maximum value of number, generate the temporal characteristics item hash function.
Optionally, when the multiple characteristic item includes category feature item, the extraction module includes：
Second statistic submodule, for counting the number that each user view occurs under any classification respectively；
Second generation submodule, the number for user view each under any classification to occur make normalization peace
Sliding processing, generates the category feature item hash function.
Optionally, when the multiple characteristic item includes timeliness-accumulation characteristic item, the extraction module includes：
Acquisition submodule, during for obtaining each user view of active user apart from the generation of current time the last time
Between, generate timeliness parameter；
Third statistic submodule, for counting number that each user view occurs within the preset time cycle respectively
Maximum value and, the number of the user view occurs within the preset time cycle for active user；
Third generates submodule, for the number that is occurred within the preset time cycle according to each user view
The number of the user view, generation cumulative bad ginseng occur within the preset time cycle for maximum value and active user
Number；
4th generation submodule, for using the timeliness parameter and cumulative bad parameter, it is special to generate the timeliness-accumulation
Levy item hash function.
Optionally, the timeliness parameter is nearest apart from current time by each user view to the active user
Primary time of origin takes the logarithm and translates and generate；And/or
The cumulative bad parameter within the preset time cycle to active user by occurring the user view
The ratio of the maximum value of number that number and each user view occur within the preset time cycle take tanh and
Generation.
Optionally, the training module includes：
Training submodule, for according to the user's history behavioral data, to temporal characteristics item hash function, category feature
Item hash function and/or timeliness-accumulation characteristic item hash function carry out model training, to obtain the weight of each characteristic item
Value.
Optionally, it further includes：
Identification module, for identifying that the corresponding user view of most probable value in the probability distribution is anticipated for target user
Figure.
To solve the above-mentioned problems, the embodiment of the invention discloses a kind of recognition methods of user view, including：
Determine multiple user views and multiple characteristic items；
The character numerical value that each characteristic item corresponds to each user view is extracted respectively；
Obtain user's history behavioral data；
Model training is carried out according to the character numerical value and the user's history behavioral data, to obtain each feature
The weighted value of item；
Current multiple characteristic items are obtained, current multiple characteristic items are respectively provided with real-time character numerical value；
According to the weighted value, summation is weighted to the real-time character numerical value, is anticipated with obtaining the multiple user
Scheme current probability distribution.
Optionally, the multiple characteristic item includes temporal characteristics item, and the temporal characteristics item corresponds to each user view
Character numerical value be converted into temporal characteristics item hash function；And/or
The multiple characteristic item includes category feature item, and the category feature item corresponds to the characteristic of each user view
Value is converted into category feature item hash function；And/or
The multiple characteristic item includes timeliness-accumulation characteristic item, and the timeliness-accumulation characteristic item is anticipated corresponding to each user
The character numerical value of figure is converted into timeliness-accumulation characteristic item hash function.
Optionally, it is described to extract each characteristic item respectively and correspond to when the multiple characteristic item includes temporal characteristics item
The step of character numerical value of each user view, includes：
Unit period is divided into multiple timeslices；
Each user view in the number and All Time piece that each user view occurs in each timeslice is counted respectively
The maximum value of the number of generation；
Each user view in the number and All Time piece that are occurred according to user view each in each timeslice
The maximum value of the number of generation generates the temporal characteristics item hash function.
Optionally, it is described to extract each characteristic item respectively and correspond to when the multiple characteristic item includes category feature item
The step of character numerical value of each user view, includes：
The number that each user view occurs under any classification is counted respectively；
The number that each user view under any classification occurs is normalized and smoothing processing, generate the classification
Characteristic item hash function.
Optionally, it is described to extract each characteristic item pair respectively when the multiple characteristic item includes timeliness-accumulation characteristic item
It should include in the step of character numerical value of each user view：
Obtain time of origin of each user view apart from current time the last time of active user, generation timeliness ginseng
Number；
Count respectively the number that each user view occurs within the preset time cycle maximum value and, it is current to use
The number of the user view occurs within the preset time cycle for family；
The maximum value of the number occurred within the preset time cycle according to each user view and active user
The number of the user view occurs within the preset time cycle, generates cumulative bad parameter；
Using the timeliness parameter and cumulative bad parameter, the timeliness-accumulation characteristic item hash function is generated.
Optionally, the timeliness parameter is nearest apart from current time by each user view to the active user
Primary time of origin takes the logarithm and translates and generate；And/or
The cumulative bad parameter within the preset time cycle to active user by occurring the user view
The ratio of the maximum value of number that number and each user view occur within the preset time cycle take tanh and
Generation.
Optionally, it is described that model training is carried out according to the character numerical value and the user's history behavioral data, to obtain
The step of weighted value of each characteristic item, includes：
According to the user's history behavioral data, to temporal characteristics item hash function, category feature item hash function and/or
Timeliness-accumulation characteristic item hash function carries out model training, to obtain the weighted value of each characteristic item.
Optionally, it after the step of the multiple user view of acquisition current probability distribution, further includes：
Identify that the corresponding user view of most probable value in the probability distribution is intended to for target user.
To solve the above-mentioned problems, the embodiment of the invention discloses a kind of electronic equipment, including：
Processor；And
Memory, for storing the executable instruction of the processor；
Wherein, the processor is configured to perform the identification side of above-mentioned user view by performing the executable instruction
The step of method.
To solve the above-mentioned problems, the embodiment of the invention discloses a kind of storage mediums, are stored thereon with computer program,
The program realizes the step of recognition methods of above-mentioned user view when being executed by processor.
Compared with background technology, the embodiment of the present invention includes advantages below：
The embodiment of the present invention by determining multiple user views and multiple characteristic items first, and extracts each feature respectively
Item corresponds to the character numerical value of each user view, then on the basis of user's history behavioral data is obtained, according to above-mentioned spy
It levies numerical value and user's history behavioral data carries out model training, obtain the weighted value of each characteristic item, so as to obtain currently
After multiple characteristic items, summation can be weighted to the current real-time character numerical value of multiple characteristic items according to above-mentioned weighted value,
Obtain the current probability distribution of multiple intentions of user.The embodiment of the present invention passes through the intention to user for different business object
Distribution is predicted that solving can not reflect that user was instantaneously inclined to asks in the case of no inquiry in the prior art in real time
Topic by calculating the probability distribution of user view, preferably reflects intention tendency of the user in request moment, and cause each
It is comparable between user view, it being capable of lateral comparison.
Description of the drawings
Fig. 1 is a kind of step flow chart of the recognition methods of user view of one embodiment of the invention；
Fig. 2 is the step flow chart of the recognition methods of another user view of one embodiment of the invention；
Fig. 3 is a kind of structure diagram of the identification device of user view of one embodiment of the invention.
Specific embodiment
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, it is below in conjunction with the accompanying drawings and specific real
Applying mode, the present invention is described in further detail.
With reference to Fig. 1, a kind of step flow chart of the recognition methods of user view of one embodiment of the invention, tool are shown
Body may include steps of：
Step 101, multiple user views and multiple characteristic items are determined；
It should be noted that the embodiment of the present invention can be applied in client, which can be with third-party clothes
Business device or server cluster are connected, and such as distributed system, can capture the business datum of business object in the network platform, network
Platform can be independent server or server cluster, for carrying out business processing to business object.
There can be different business objects in different business scopes, for example, in the field of communications, business object can
Think communication data；In news media field, business object can be news data；In search field, business object can
Think webpage；In e-commerce field, business object can be commodity or the type of merchandise, etc..
In embodiments of the present invention, user view can be the intention that user is directed to each business object.If for example, business
Object is included to shop food and drink and take-away, then user view can be the meaning that user tends to choose respectively shop food and drink or take-away
Figure.
In embodiments of the present invention, multiple characteristic items can include client condition feature or user behavior feature.For example, when
Between, weather, the click of user, browsing or lower single act etc..The business object and spy that the present embodiment is directed to user view
Sign item is not construed as limiting.
Step 102, the character numerical value that each characteristic item corresponds to each user view is extracted respectively；
In embodiments of the present invention, character numerical value can refer to probability of each user view respectively under each characteristic item
Value.For example, for user view to take out, characteristic item is the time, then it is each small in 24 hours one day can to extract take-away
When probability value.
In the concrete realization, multiple characteristic items can be classified according to respective feature.For example, for only with the time
The characteristic item of close relation can be divided into temporal characteristics item；For in the characteristic item of classification close relation, can only incite somebody to action
It is divided into category feature item；And those are not only needed to consider timeliness, it is also desirable to consider cumulative characteristic item, then may be used
To be divided into timeliness-accumulation characteristic item.
In embodiments of the present invention, each temporal characteristics item can be converted corresponding to the character numerical value of each user view
For temporal characteristics item hash function；The character numerical value that each category feature item corresponds to each user view can be converted into class
Other characteristic item hash function；And the character numerical value that each timeliness-accumulation characteristic item corresponds to each user view can then be turned
It is changed to timeliness-accumulation characteristic item hash function.
Step 103, user's history behavioral data is obtained；
In embodiments of the present invention, user's history behavioral data can refer to the passing operation behavior data of user.For example,
The click for the business object that user is directed to each user view browses or the data for the behaviors such as place an order.
Step 104, model training is carried out according to the character numerical value and the user's history behavioral data, with described in acquisition
The weighted value of each characteristic item；
It in the concrete realization, can be with base after determining that each characteristic item corresponds to the character numerical value of each user view
Model training is carried out in user's history behavioral data, obtains the weighted value of each characteristic item.
It should be noted that the weighted value of each characteristic item being calculated is when carrying out subsequent user view prediction
Constant.
Step 105, current multiple characteristic items are obtained, current multiple characteristic items are respectively provided with real-time characteristic
Value；
Current multiple characteristic items can refer to current possessed complete characteristic.For example, current time, place, day
Gas etc..Current multiple characteristic items can be respectively provided with real-time character numerical value.
Step 106, according to the weighted value, summation is weighted to the real-time character numerical value, it is described more to obtain
The current probability distribution of a user view.
It in embodiments of the present invention, can be right according to the weighted value of each characteristic item in the identification for carrying out user view
The real-time characteristic of current multiple characteristic items is weighted summation, and the result of weighted sum is that each user view is worked as
Preceding probability distribution.
The current probability distribution of each user view can represent tendentious size of the user for each user view.
For example, probability value is bigger, represent that user is stronger for the tendentiousness of the intention.
In embodiments of the present invention, by determining multiple user views and multiple characteristic items first, and extraction is each respectively
Characteristic item corresponds to the character numerical value of each user view, then on the basis of user's history behavioral data is obtained, according to upper
It states character numerical value and user's history behavioral data carries out model training, obtain the weighted value of each characteristic item, so as to work as in acquisition
After preceding multiple characteristic items, the current real-time character numerical value of multiple characteristic items can be weighted according to above-mentioned weighted value
Summation, obtains the current probability distribution of multiple intentions of user.The embodiment of the present invention to user by being directed to different business object
Intention distribution predicted in real time, solving can not reflect that user is instantaneously inclined in the case of no inquiry in the prior art
The problem of, by calculating the probability distribution of user view, intention tendency of the user in request moment is preferably reflected, and cause
It is comparable between each user view, it being capable of lateral comparison.
With reference to Fig. 2, the step flow chart of the recognition methods of another user view of one embodiment of the invention is shown,
Specifically it may include steps of：
Step 201, multiple user views and multiple characteristic items are determined；
In embodiments of the present invention, user view can refer to that user is directed to the intention or tendentiousness of each business object,
By determining the intention of user, corresponding business object can be recommended to user.
In embodiments of the present invention, in order to recommend business object to user, multiple industry to be recommended can be determined first
Business object and the characteristic item for being analyzed above-mentioned multiple business objects, to determine that user is opening some application program
The instantaneous probability for some business object of APP.
In the concrete realization, multiple user views can be an intention set being made of multiple intentions of user
{P1, P2, P3... PN, such as { to shop food and drink, taking out, film, hotel, tourism, KTV, amusement } etc..Meanwhile it can give
One two-valued variable y ∈ { 0,1 }, if user is intended to some, the value of y is 1；If user anticipates without some
Figure, then the value of y is 0.So as to which user can be expressed as the intensity that some is intended to probability when y is 1：P { y=1 }.
When user opens the instantaneous of some APP, the intensity of the intention can be expressed as P { y=1 | open }, wherein open tables
Show that user opens the event of APP.
In embodiments of the present invention, multiple characteristic items can both include objective condition feature, can also include user behavior
Feature.Wherein, objective condition feature can refer to the objective condition such as environmental factor, time factor.For example, time, weather, place
Etc..User behavior feature can refer to user for each business object occurred such as browse, click, place an order subjective row
For.
In the concrete realization, multiple characteristic items can also exist in the form of characteristic item set.For example, { time, day
Gas, place, case history place an order, case history inquiry } etc..
Step 202, the character numerical value that each characteristic item corresponds to each user view, the character numerical value packet are extracted respectively
Include temporal characteristics item hash function, category feature item hash function and/or timeliness-accumulation characteristic item hash function；
In embodiments of the present invention, after objective condition feature and user behavior feature is introduced, user is directed to a certain intention
Intensity P { y=1 | open } can be further represented as：
P y=1 | and open }=∑C ∈ { Context, User }P y=1 | C, open } P C | open } ... (1)
Wherein, it is in objective condition feature and user behavior characteristic set that C ∈ { Context, User }, which can be understood as C,
One.
Therefore, it calculates user and opens APP instantaneously for the probability of some intention, can be converted into and calculate user's opening APP
It is instantaneous, user has the sum of probability of the intention under all characteristic items.
In the concrete realization, an independence assumption can be introduced：If user open APP instantaneously have some feature with
It is independent that user, which has the probability of some intention, under this feature, so as to be calculated with the mode of probability product.In fact,
Some feature that instantaneously has that user opens APP with some intention is not necessarily independent, and still, this independence assumption is from statistics
It is reasonable in meaning, and being capable of simplified model calculating.
In embodiments of the present invention, it is assumed that the probability with some feature is θ when user opens APP, and under this feature
It is x that user, which has the probability of some intention, then above-mentioned formula (1) can be converted into monadic logic regression problem：
P y=1 | and open }=sigmoid (∑siθi*xi), xi∈ [0,1] ... (2)
Therefore, the distribution for calculating user view is to calculate all { P1, P2, P3..., PN, wherein Pi∈ [0,1], ∑
Pi=1.
With reference to above-mentioned derivation, then the problem of solving the probability distribution of user view, which can be converted into, to be solved Softmax and returns
Problem, formula are as follows：
Wherein j ∈ { 1,2 ..., N } represent N number of intention, xiIt is feature vector, includes objective condition feature and user's row
It is characterized.
For above-mentioned formula (4), θ can obtain the study of sample by machine learning algorithm, therefore, pre- when needing
When one user of survey opens the instantaneous intention of APP, it is only necessary to obtain current multiple features, then be brought into and acquired θ's
In formula (4), it is possible to calculate the probability distribution that user is directed to each user view.
In embodiments of the present invention, before above-mentioned solution is carried out, each characteristic item can be extracted respectively first and corresponded to often
The character numerical value of a user view.
In the concrete realization, multiple characteristic items can be classified according to respective feature.For example, temporal characteristics item,
Category feature item and timeliness-accumulation characteristic item.Above-mentioned temporal characteristics item can include the time；Above-mentioned category feature item can include
The features such as weather, place；And the features such as timeliness-accumulation characteristic item can then place an order including case history, case history inquiry.
In embodiments of the present invention, each temporal characteristics item can be converted corresponding to the character numerical value of each user view
For temporal characteristics item hash function；The character numerical value that each category feature item corresponds to each user view can be converted into class
Other characteristic item hash function；And the character numerical value that each timeliness-accumulation characteristic item corresponds to each user view can then be turned
It is changed to timeliness-accumulation characteristic item hash function.
In the concrete realization, for temporal characteristics item, can unit period be divided into multiple timeslices first, often
One timeslice note can be Ti, each timeslice T is then counted respectivelyiIn each user view occur number count (e,
Ti) and All Time piece in each user view maximum value max { count (e, the T of the number that occuri)}；So as to according to upper
State each timeslice TiIn each user view the number count (e, the T that occuri) and All Time piece in each user view hair
Maximum value max { count (e, the T of raw numberi), generated time characteristic item hash function, formula is as follows：
Specifically, a certain user of the different periods in one day can be investigated with 24 hours one day for the unit time cycle
It is intended to the number occurred.It for example, can be with every 5 minutes for a time interval, when being set as 288 by intraday 24 hours
Between piece, then each timeslice be 5 minutes, then by counting the numerical value in nearly three months, obtained hash function value is done flat
, which is then final value.
In practical application, when user is when any moment of one day opens APP, it can pass through this group of hash function
Obtain the character numerical value of the time corresponding user view.
For example, for taking out this user view that places an order, daily 11~13 points can be informed in by hash function is
At the time of this feature is most strong in one day.
For the category features item such as weather, place, corresponding character numerical value can be existed by counting each user view
The number occurred in each classification obtains.
In the concrete realization, number count (e, the C that each user view occurs under any classification can be counted respectivelyi),
Then number count (e, C user view each under any of the above-described classification occurredi) normalize and smoothing processing, generation
Category feature item hash function, wherein, the progress of tanh functions may be used when making smoothing processing.Corresponding formula can represent
It is as follows：
One group of hash function corresponding with category feature item can be obtained using above-mentioned formula, it is hereby achieved that in user
When opening APP, user's intensity that corresponding user view occurs under a certain classification scene, that is, under the category user to certain
The intensity of the intention of a business object.
In addition, for certain form of characteristic item, not only need to consider timeliness, it is also contemplated that cumulative bad.For example, for
User buys the behavior taken out, if user A had unirecord under take-away in one week, and user B had take-away before three weeks
Lower unirecord, then user A and user B may be different to taking out the intensity being intended to, this is the embodiment of timeliness.In addition,
In one cycle, such as in three months, if user A places an order, take-away 30 is single, and user B has only descended 5 lists, then user A and B pairs
The intention power of take-away should also be different, this is the embodiment of accumulation property.Case history places an order special with case history inquiry etc.
Sign item is the characteristic item for needing while considering timeliness and accumulation property.
In embodiments of the present invention, for timeliness-accumulation characteristic item, timeliness can be denoted as x1, accumulation is tired to be denoted as x2,
Then the intensity of corresponding feature can represent：
Y=α * x1+(1-α)*x2, α ∈ [0,1] ... (8)
Wherein, α is timeliness and cumulative weight parameter, and codomain is [0,1], and the size of α can be according to actual needs
Specific to determine, the embodiment of the present invention is not construed as limiting this.
In embodiments of the present invention, each user view of active user can be obtained apart from current time the last time
Time of origin generates timeliness parameter.
In the concrete realization, timeliness parameter can by each user view to above-mentioned active user apart from it is current when
Between the last time of origin take the logarithm and translate and generate.
It is, for example, possible to use time attenuation function, records the corresponding numerical value of a certain user view the last time time of origin,
As x1Value.Wherein time attenuation function can be to do translation using log function pair origins to obtain.
Then the maximum value for the number that each user view occurs within the preset time cycle can be counted respectively, with
And the number of the user view occurs within the above-mentioned preset time cycle for active user.So as to be anticipated according to each user
The maximum value for the number that figure occurs within the preset time cycle and, active user sends out within the above-mentioned preset time cycle
The number of the raw user view, generates cumulative bad parameter.
In the concrete realization, cumulative bad parameter can be anticipated by active user occurring within the preset time cycle user
The ratio of the maximum value of number that the number of figure occurs with each user view within the preset time cycle takes tanh
(tanh) it generates.
For example, can count all users first occurs the number of each user view most within the preset time cycle
Big value, is denoted as max=max { count (u, e, T) }, and establish function as waypoint：
Wherein, u (e) is active user's period interior number that a certain user view occurs at the same time.
Then, normalized is made using tanh function pair u (e) and max, you can obtain x2Value：
After timeliness parameter and cumulative bad parameter is obtained respectively, timeliness parameter and cumulative bad parameter may be used, it is raw
Into timeliness-accumulation characteristic item hash function.
Step 203, user's history behavioral data is obtained；
In embodiments of the present invention, user's history behavioral data can refer to the passing operation behavior data of user.For example,
The click for the business object that user is directed to each user view browses or the data for the behaviors such as place an order.
In embodiments of the present invention, can training sample set be generated according to the user's history behavioral data of extraction.Training sample
This collection can include positive sample collection and negative sample collection.
In the concrete realization, after user opens APP, for multiple business objects that the APP shows, if user clicks
Some business object, then can be using the corresponding classification of the business object as positive sample collection, and other industry that user does not click on
The corresponding classification of business object can then be used as negative sample collection.
Alternatively, if the search key of APP acquiescence displayings is directly searched by user, it may be considered that the search key
It is accurate to recommend, so as to using the corresponding classification of the keyword as positive sample collection., whereas if user does not search for acquiescence displaying
Keyword, then can be used as negative sample collection.Certainly, those skilled in the art can specifically adjust sample set according to actual needs
Data source, the present embodiment is not construed as limiting this.
Step 204, according to the user's history behavioral data, to temporal characteristics item hash function, category feature item Hash
Function and/or timeliness-accumulation characteristic item hash function carry out model training, to obtain the weighted value of each characteristic item；
In embodiments of the present invention, after training sample set is obtained, the temporal characteristics item obtained in step 202 can be breathed out
Uncommon function, category feature item hash function and timeliness-accumulation characteristic item hash function carry out model training, so as to obtain each spy
Levy the weighted value of item.
In the concrete realization, the character numerical value of the characteristic item of extraction can be brought into formula (4), obtains corresponding feature
The value of the weighted value, i.e. θ in formula (4) of item.
Step 205, current multiple characteristic items are obtained, current multiple characteristic items are respectively provided with real-time characteristic
Value；
In embodiments of the present invention, user can be multiple characteristic items and corresponding power for the probability of each user view
Weight values are coefficient as a result, therefore, in the probability distribution for determining multiple user views, can obtain first current multiple
Characteristic item.
Current multiple characteristic items can refer to current possessed complete characteristic.For example, current time, place, day
Gas etc..Current multiple characteristic items can be respectively provided with real-time character numerical value.
In the concrete realization, the current corresponding real-time character numerical value of multiple characteristic items can be carried out according to step 202
It determines, the present embodiment repeats no more this.
Step 206, according to the weighted value, summation is weighted to the real-time character numerical value, it is described more to obtain
The current probability distribution of a user view；
In embodiments of the present invention, current multiple characteristic items are being obtained, and is determining that each characteristic item is corresponding in real time
Character numerical value after, the weighted value that step 204 is calculated may be used, to the current real-time character numerical value of multiple characteristic items
Summation is weighted, obtains the current probability distribution of multiple user views.
In the concrete realization, the current real-time character numerical value of multiple characteristic items can be brought into formula (4) and asked
Solution, directly exports the current probability distribution of multiple user views.
Step 207, identify that the corresponding user view of most probable value in the probability distribution is intended to for target user.
In embodiments of the present invention, user reflects user to each business pair for the probability distribution of each user view
The strong and weak situation of the intention of elephant.Therefore, it after the probability distribution for obtaining multiple user views, can further anticipate from multiple users
Target user's intention is extracted in figure.
For example, the corresponding user view of most probable value that can be extracted in probability distribution is intended to for target user, target
The corresponding business object of user view is the most strong target service object of user view.
In embodiments of the present invention, after determining that target user is intended to, target user can be intended to corresponding target industry
Object recommendation be engaged in user.
Specifically, target service object can be presented in search box, user to be facilitated directly to use target service pair
As scanning for；Target service object can also preferentially be showed in the search results pages of user, to realize of search result
Propertyization sorts.Certainly, those skilled in the art can also select other ways of recommendation, the present embodiment pair according to actual needs
This is not construed as limiting.
In embodiments of the present invention, it by calculating probability distribution of the user to multiple user views, and anticipates from multiple users
Identify that target user is intended in figure, solving can not reflect that user is instantaneously inclined in the case of no inquiry in the prior art
The problem of, by calculating the probability distribution of user view, intention tendency of the user in request moment is preferably reflected, and cause
It is comparable between each user view, it being capable of lateral comparison.Secondly, the embodiment of the present invention is in the recommendation for carrying out business object
When, inquiry that can be independent of user or other operation, instantaneously the intention of user can be predicted a certain, improved
The validity that business object is recommended.
It is situated between in order to make it easy to understand, making one to the recognition methods of the user view of the present invention below with a complete example
It continues.
S1, under O2O scenes, user can be consumed using application APP.In order to user open APP when, soon
The intention of user is predicted on fast ground, so as to preferably recommend consumption type (the i.e. business pair for meeting its intention to user
As), it can determine to be intended to set and characteristic item set first, such as it can be that P ∈ { to shop food and drink, take out, electricity to be intended to set
Shadow, hotel, tourism, KTV, amusement }, characteristic item set can be that { time, weather, place, case history place an order C ∈, and individual goes through
History is inquired }, wherein, the time, weather, place belongs to objective condition feature, and case history places an order belongs to use with case history inquiry
Family behavioural characteristic, totally five dimensional feature.
S2 carries out the extraction of character numerical value according to the determining suitable characteristic model of Feature selection.
For example, for time dimension, temporal characteristics item model may be used, to all under each of the above intent classifier
User weighs lower single event, and calculates the numerical characteristics hash function of different periods of the user group in one day, so as to
There are higher weights at noon with the intention for obtaining food and drink class, and the strong of film is intended to appear in prime time in night, hotel
It is strong to be intended to then after 22 points of night.
For weather and Site characterization item, it is suitble to use classes characteristic item model.Further, due to atrocious weather feelings
Condition will be influenced with the relevant consumer behavior of service for life, therefore, can be by weather typing low temperature, and heavy rain, moderate rain, light rain, greatly
Wind, weight haze, dense fog, thunder and lightning, snowfall, other, and the weather conditions in normal range (NR) are placed in other classification, it is believed that this point
Class influences the consumer behavior of user little；At the same time it can also be divided into { morning, afternoon, night } by one day, it is divided into { work within one week
Make day, day off }.Then, it is combined respectively to classifying above, hash function is calculated using category feature item model, this
Weather characteristics when group hash function can be used for off-line training model and on-line prediction calculate.It is similar with weather characteristics item, also
Further Site characterization item can be classified as place of working, residence, trade company, road, other, and it was divided into { work by one week
Day, day off }, then each classification is combined, obtains the hash function of Site characterization item.
Place an order and case history query characteristics item for case history, timeliness-accumulation characteristic item model may be used, from when
Its numeric score of COMPREHENSIVE CALCULATING in terms of effect property and cumulative bad two.It places an order for case history, each user can be weighed most
Lower one-state near three months under each intent classifier；And timeliness can use function (4+log0.5(t+1))/4,
Its numerical value feature is horizontal axis t levels off to 1 close to y when 0, and with the increase of t, gradually close to horizontal axis, i.e. 0 value, which carves y
It depicts that lower single time is nearer, is intended to the interest to some intention classification, the time is more remote, then interest attenuation.Specific
During calculating, may be used it is recent place an order to update timeliness numerical value, cumulative bad can then use user in three months
The lower monodrome of maximum does piecewise function, so as to making normalizing for each placing an order for classification of intention in three months of each user
The numeric score of change.So it can be seen that it is more to place an order, score is higher.For case history query characteristics item, place an order with aforementioned
It is similar, the data in 4 weeks can be selected to calculate.This feature item relies on query classifier and classifies to each search inquiry,
User can be obtained for each statistics for being intended to Query quantity by calculating, so as to obtain score.It can be seen that inquiry
Number is more, and query time is nearer, then score is higher.
It should be noted that in features above item set, what the objective condition such as time, place, weather characteristic item was investigated
It is the behavior of user group, case history places an order, and then investigate with user behaviors characteristic items such as case history inquiries is user's individual
Behavior.After features described above numerical value is obtained, model training can be proceeded by.
S3 specifically, may be used Softmax regression models and be trained, it is hereby achieved that the θ in formula (4)
Value.In the concrete realization, it is P ∈ { to shop food and drink, taking out, film, hotel, tourism, KTV, amusement } intention to be gathered, feature
For set is C ∈ { time, weather, place, case history place an order, case history inquiry }, when model training is the example
The matrix of a 7*5 can be expressed as, so as to which L-BFGS algorithms is used to acquire θ parameters.
S4, when user open APP it is instantaneous, each characteristic item currently corresponding characteristic information can be obtained, for example, currently
Time, weather etc., so as to which the Feature Extraction Method continued as described above calculates the current corresponding feature of characteristic item
Numerical value inputs trained Softmax regression models, it is hereby achieved that probability distribution of the user for intention set,
The probability distribution embodies strong and weak situation of the user for the intention of each business object.
S5, after the real-time intention distribution of user is obtained, in the application that can recommend in default search word, selection is intended to
The corresponding business object of most probable value is recommended user by the recommendation query under a most strong classification.
It should be noted that for embodiment of the method, in order to be briefly described, therefore it is all expressed as to a series of action group
It closes, but those skilled in the art should know, the embodiment of the present invention is not limited by described sequence of movement, because according to
According to the embodiment of the present invention, certain steps may be used other sequences or be carried out at the same time.Secondly, those skilled in the art also should
Know, embodiment described in this description belongs to preferred embodiment, and the involved action not necessarily present invention is implemented
Necessary to example.
With reference to Fig. 3, a kind of structure diagram of the recommendation apparatus embodiment of business object of the present invention is shown, it specifically can be with
Including following module：
Determining module 301, for determining multiple user views and multiple characteristic items；
Extraction module 302, for extracting the character numerical value that each characteristic item corresponds to each user view respectively；
First acquisition module 303, for obtaining user's history behavioral data；
Training module 304, for carrying out model training according to the character numerical value and the user's history behavioral data, with
Obtain the weighted value of each characteristic item；
Second acquisition module 305, for obtaining current multiple characteristic items, current multiple characteristic items are respectively provided with
Real-time character numerical value；
It is intended to probability distribution computing module 306, for according to the weighted value, adding to the real-time character numerical value
Power summation, to obtain the current probability distribution of the multiple user view.
In embodiments of the present invention, the multiple characteristic item includes temporal characteristics item, and the temporal characteristics item corresponds to every
The character numerical value of a user view is converted into temporal characteristics item hash function；And/or
The multiple characteristic item includes category feature item, and the category feature item corresponds to the characteristic of each user view
Value is converted into category feature item hash function；And/or
The multiple characteristic item includes timeliness-accumulation characteristic item, and the timeliness-accumulation characteristic item is anticipated corresponding to each user
The character numerical value of figure is converted into timeliness-accumulation characteristic item hash function.
In embodiments of the present invention, when the multiple characteristic item includes temporal characteristics item, the extraction module 302 is specific
It can include following submodule：
Submodule is divided, for unit period to be divided into multiple timeslices；
First statistic submodule, for count respectively each user view occurs in each timeslice number and it is whole when
Between each user view occurs in piece the maximum value of number；
First generation submodule, for occurred according to each user view in each timeslice number and whole when
Between each user view occurs in piece the maximum value of number, generate the temporal characteristics item hash function.
In embodiments of the present invention, when the multiple characteristic item includes category feature item, the extraction module 302 is specific
It can include following submodule：
Second statistic submodule, for counting the number that each user view occurs under any classification respectively；
Second generation submodule, the number for user view each under any classification to occur make normalization peace
Sliding processing, generates the category feature item hash function.
In embodiments of the present invention, when the multiple characteristic item includes timeliness-accumulation characteristic item, the extraction module
302 can specifically include following submodule：
Acquisition submodule, during for obtaining each user view of active user apart from the generation of current time the last time
Between, generate timeliness parameter；
Third statistic submodule, for counting number that each user view occurs within the preset time cycle respectively
Maximum value and, the number of the user view occurs within the preset time cycle for active user；
Third generates submodule, for the number that is occurred within the preset time cycle according to each user view
The number of the user view, generation cumulative bad ginseng occur within the preset time cycle for maximum value and active user
Number；
4th generation submodule, for using the timeliness parameter and cumulative bad parameter, it is special to generate the timeliness-accumulation
Levy item hash function.
In embodiments of the present invention, the timeliness parameter is worked as by each user view distance to the active user
The time of origin of preceding the last time time takes the logarithm and translates and generate；And/or
The cumulative bad parameter within the preset time cycle to active user by occurring the user view
The ratio of the maximum value of number that number and each user view occur within the preset time cycle take tanh and
Generation.
In embodiments of the present invention, the training module 304 can specifically include following submodule：
Training submodule, for according to the user's history behavioral data, to temporal characteristics item hash function, category feature
Item hash function and/or timeliness-accumulation characteristic item hash function carry out model training, to obtain the weight of each characteristic item
Value.
In embodiments of the present invention, described device can also include following module：
Identification module, for identifying that the corresponding user view of most probable value in the probability distribution is anticipated for target user
Figure.
For device embodiment, since it is basicly similar to embodiment of the method, so description is fairly simple, it is related
Part illustrates referring to the part of embodiment of the method.
The embodiment of the present invention also provides a kind of electronics, including processor；And
Memory, for storing the executable instruction of above-mentioned processor；
Wherein, which is configured to perform the recognition methods of above-mentioned user view by performing above-mentioned executable instruction
Each process of embodiment, and identical technique effect can be reached, it is repeated to avoid, which is not described herein again.
The embodiment of the present invention also provides a kind of computer readable storage medium, and meter is stored on computer readable storage medium
Calculation machine program, the computer program realize each mistake of the recognition methods embodiment of above-mentioned user view when being executed by processor
Journey, and identical technique effect can be reached, it is repeated to avoid, which is not described herein again.Wherein, computer readable storage medium, such as
Read-only memory (Read-Only Memory, abbreviation ROM), random access memory (Random Access Memory, abbreviation
RAM), magnetic disc or CD etc..
Each embodiment in this specification is described by the way of progressive, the highlights of each of the examples are with
The difference of other embodiment, just to refer each other for identical similar part between each embodiment.
It should be understood by those skilled in the art that, the embodiment of the embodiment of the present invention can be provided as method, apparatus or calculate
Machine program product.Therefore, the embodiment of the present invention can be used complete hardware embodiment, complete software embodiment or combine software and
The form of the embodiment of hardware aspect.Moreover, the embodiment of the present invention can be used one or more wherein include computer can
With in the computer-usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) of program code
The form of the computer program product of implementation.
The embodiment of the present invention be with reference to according to the method for the embodiment of the present invention, terminal device (system) and computer program
The flowchart and/or the block diagram of product describes.It should be understood that flowchart and/or the block diagram can be realized by computer program instructions
In each flow and/or block and flowchart and/or the block diagram in flow and/or box combination.These can be provided
Computer program instructions are set to all-purpose computer, special purpose computer, Embedded Processor or other programmable data processing terminals
Standby processor is to generate a machine so that is held by the processor of computer or other programmable data processing terminal equipments
Capable instruction generation is used to implement in one flow of flow chart or multiple flows and/or one box of block diagram or multiple boxes
The device for the function of specifying.
These computer program instructions, which may also be stored in, can guide computer or other programmable data processing terminal equipments
In the computer-readable memory to work in a specific way so that the instruction being stored in the computer-readable memory generates packet
The manufacture of command device is included, which realizes in one flow of flow chart or multiple flows and/or one side of block diagram
The function of being specified in frame or multiple boxes.
These computer program instructions can be also loaded into computer or other programmable data processing terminal equipments so that
Series of operation steps are performed on computer or other programmable terminal equipments to generate computer implemented processing, thus
The instruction offer performed on computer or other programmable terminal equipments is used to implement in one flow of flow chart or multiple flows
And/or specified in one box of block diagram or multiple boxes function the step of.
Although the preferred embodiment of the embodiment of the present invention has been described, those skilled in the art once know base
This creative concept can then make these embodiments other change and modification.So appended claims are intended to be construed to
Including preferred embodiment and fall into all change and modification of range of embodiment of the invention.
Finally, it is to be noted that, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, term " comprising ", "comprising" or its any other variant meaning
Covering non-exclusive inclusion, so that process, method, article or terminal device including a series of elements are not only wrapped
Those elements are included, but also including other elements that are not explicitly listed or are further included as this process, method, article
Or the element that terminal device is intrinsic.In the absence of more restrictions, it is wanted by what sentence "including a ..." limited
Element, it is not excluded that also there are other identical elements in the process including the element, method, article or terminal device.
Above to a kind of recognition methods of user view provided by the present invention, a kind of identification device of user view, one
Kind of electronic equipment and a kind of storage medium, are described in detail, specific case used herein to the principle of the present invention and
Embodiment is expounded, and the explanation of above example is only intended to facilitate the understanding of the method and its core concept of the invention；
Meanwhile for those of ordinary skill in the art, thought according to the present invention can in specific embodiments and applications
There is change part, in conclusion the content of the present specification should not be construed as limiting the invention.
Claims (14)
1. a kind of identification device of user view, which is characterized in that including：
Determining module, for determining multiple user views and multiple characteristic items；
Extraction module, for extracting the character numerical value that each characteristic item corresponds to each user view respectively；
First acquisition module, for obtaining user's history behavioral data；
Training module, for carrying out model training according to the character numerical value and the user's history behavioral data, to obtain
State the weighted value of each characteristic item；
Second acquisition module, for obtaining current multiple characteristic items, current multiple characteristic items are respectively provided in real time
Character numerical value；
It is intended to probability distribution computing module, for according to the weighted value, summation to be weighted to the real-time character numerical value,
To obtain the current probability distribution of the multiple user view.
2. the apparatus according to claim 1, which is characterized in that
The multiple characteristic item includes temporal characteristics item, and the temporal characteristics item corresponds to the character numerical value quilt of each user view
Be converted to temporal characteristics item hash function；And/or
The multiple characteristic item includes category feature item, and the category feature item corresponds to the character numerical value quilt of each user view
Be converted to category feature item hash function；And/or
The multiple characteristic item includes timeliness-accumulation characteristic item, and the timeliness-accumulation characteristic item corresponds to each user view
Character numerical value is converted into timeliness-accumulation characteristic item hash function.
3. the apparatus of claim 2, which is characterized in that when the multiple characteristic item includes temporal characteristics item, institute
Extraction module is stated to include：
Submodule is divided, for unit period to be divided into multiple timeslices；
First statistic submodule, for counting the number and All Time piece that each user view occurs in each timeslice respectively
In the maximum value of number that occurs of each user view；
First generation submodule, for the number and All Time piece occurred according to user view each in each timeslice
In the maximum value of number that occurs of each user view, generate the temporal characteristics item hash function.
4. the apparatus of claim 2, which is characterized in that when the multiple characteristic item includes category feature item, institute
Extraction module is stated to include：
Second statistic submodule, for counting the number that each user view occurs under any classification respectively；
Second generation submodule, the number for user view each under any classification to occur make normalization and smooth place
Reason, generates the category feature item hash function.
5. the apparatus of claim 2, which is characterized in that when the multiple characteristic item includes timeliness-accumulation characteristic item
When, the extraction module includes：
Acquisition submodule, for obtaining time of origin of each user view apart from current time the last time of active user,
Generate timeliness parameter；
Third statistic submodule, for counting the maximum for the number that each user view occurs within the preset time cycle respectively
Value and, the number of the user view occurs within the preset time cycle for active user；
Third generates submodule, for the maximum of number occurred within the preset time cycle according to each user view
The number of the user view occurs within the preset time cycle for value and active user, generates cumulative bad parameter；
4th generation submodule, for using the timeliness parameter and cumulative bad parameter, generates the timeliness-accumulation characteristic item
Hash function.
6. device according to claim 5, which is characterized in that
The timeliness parameter is by each user view to the active user apart from the generation of current time the last time
Time takes the logarithm and translates and generate；And/or
The cumulative bad parameter within the preset time cycle by occurring active user the number of the user view
The ratio of the maximum value of number occurred within the preset time cycle with each user view takes tanh and generates.
7. the apparatus of claim 2, which is characterized in that the training module includes：
Training submodule, for according to the user's history behavioral data, being breathed out to temporal characteristics item hash function, category feature item
Uncommon function and/or timeliness-accumulation characteristic item hash function carry out model training, to obtain the weighted value of each characteristic item.
8. the apparatus according to claim 1, which is characterized in that further include：
Identification module, for identifying that the corresponding user view of most probable value in the probability distribution is intended to for target user.
9. a kind of recognition methods of user view, which is characterized in that including：
Determine multiple user views and multiple characteristic items；
The character numerical value that each characteristic item corresponds to each user view is extracted respectively；
Obtain user's history behavioral data；
Model training is carried out according to the character numerical value and the user's history behavioral data, to obtain each characteristic item
Weighted value；
Current multiple characteristic items are obtained, current multiple characteristic items are respectively provided with real-time character numerical value；
According to the weighted value, summation is weighted to the real-time character numerical value, is worked as with obtaining the multiple user view
Preceding probability distribution.
10. according to the method described in claim 9, it is characterized in that,
The multiple characteristic item includes temporal characteristics item, and the temporal characteristics item corresponds to the character numerical value quilt of each user view
Be converted to temporal characteristics item hash function；And/or
The multiple characteristic item includes category feature item, and the category feature item corresponds to the character numerical value quilt of each user view
Be converted to category feature item hash function；And/or
The multiple characteristic item includes timeliness-accumulation characteristic item, and the timeliness-accumulation characteristic item corresponds to each user view
Character numerical value is converted into timeliness-accumulation characteristic item hash function.
It is 11. according to the method described in claim 10, it is characterized in that, described according to the character numerical value and the user's history
Behavioral data carries out model training, is included the step of the weighted value of each characteristic item with obtaining：
According to the user's history behavioral data, to temporal characteristics item hash function, category feature item hash function and/or when
Effect-accumulation characteristic item hash function carries out model training, to obtain the weighted value of each characteristic item.
12. according to the method described in claim 9, it is characterized in that, obtain current general of the multiple user view described
After the step of rate is distributed, further include：
Identify that the corresponding user view of most probable value in the probability distribution is intended to for target user.
13. a kind of electronic equipment, which is characterized in that including：
Processor；And
Memory, for storing the executable instruction of the processor；
Wherein, the processor is configured to require 9-12 any one of them by performing the executable instruction come perform claim
The step of recognition methods.
14. a kind of storage medium, is stored thereon with computer program, which is characterized in that the program is realized when being executed by processor
The step of claim 9-12 any one of them recognition methods.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711219349.4A CN108230007B (en) | 2017-11-28 | 2017-11-28 | User intention identification method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711219349.4A CN108230007B (en) | 2017-11-28 | 2017-11-28 | User intention identification method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108230007A true CN108230007A (en) | 2018-06-29 |
| CN108230007B CN108230007B (en) | 2021-09-10 |
Family
ID=62653016
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711219349.4A Active CN108230007B (en) | 2017-11-28 | 2017-11-28 | User intention identification method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108230007B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109087135A (en) * | 2018-07-25 | 2018-12-25 | 百度在线网络技术（北京）有限公司 | The method for digging and device, computer equipment and readable medium that user is intended to |
| CN109299994A (en) * | 2018-07-27 | 2019-02-01 | 北京三快在线科技有限公司 | Recommended method, device, equipment and readable storage medium storing program for executing |
| CN109492104A (en) * | 2018-11-09 | 2019-03-19 | 北京京东尚科信息技术有限公司 | Training method, classification method, system, equipment and the medium of intent classifier model |
| CN109582865A (en) * | 2018-11-19 | 2019-04-05 | 北京奇虎科技有限公司 | A kind of method and device of pushing application program |
| CN109710941A (en) * | 2018-12-29 | 2019-05-03 | 上海点融信息科技有限责任公司 | User's intension recognizing method and device based on artificial intelligence |
| CN110162535A (en) * | 2019-03-26 | 2019-08-23 | 腾讯科技（深圳）有限公司 | For executing personalized searching method, device, equipment and storage medium |
| CN110765346A (en) * | 2018-07-10 | 2020-02-07 | 阿里巴巴集团控股有限公司 | User intention mining method, device and equipment |
| CN111177571A (en) * | 2020-01-13 | 2020-05-19 | 海南大学 | Value-driven dynamic recommendation system with multi-factor dimension space and multi-mesoscale fusion |
| CN111400473A (en) * | 2020-03-18 | 2020-07-10 | 北京三快在线科技有限公司 | Method and device for training intention recognition model, storage medium and electronic equipment |
| WO2020258836A1 (en) * | 2019-06-24 | 2020-12-30 | 创新先进技术有限公司 | Method and apparatus for sensing transaction user in advance |
| CN112711717A (en) * | 2021-03-26 | 2021-04-27 | 北京三快在线科技有限公司 | Room product searching method and device and electronic equipment |
| US11216832B2 (en) | 2019-06-24 | 2022-01-04 | Advanced New Technologies Co., Ltd. | Predicting future user transactions |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105468161A (en) * | 2016-01-21 | 2016-04-06 | 北京百度网讯科技有限公司 | Instruction execution method and device |
| CN105488221A (en) * | 2015-12-25 | 2016-04-13 | 北京奇虎科技有限公司 | Method and system for recommending query terms for conducting searching in search interface |
| US20160140125A1 (en) * | 2014-11-18 | 2016-05-19 | Yahoo! Inc. | Method and system for providing query suggestions based on user feedback |
| CN106779230A (en) * | 2016-12-27 | 2017-05-31 | 北京三快在线科技有限公司 | Order volume predictor method and device, electronic equipment |
| CN107146089A (en) * | 2017-03-29 | 2017-09-08 | 北京三快在线科技有限公司 | The single recognition methods of one kind brush and device, electronic equipment |
-
2017
- 2017-11-28 CN CN201711219349.4A patent/CN108230007B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160140125A1 (en) * | 2014-11-18 | 2016-05-19 | Yahoo! Inc. | Method and system for providing query suggestions based on user feedback |
| CN105488221A (en) * | 2015-12-25 | 2016-04-13 | 北京奇虎科技有限公司 | Method and system for recommending query terms for conducting searching in search interface |
| CN105468161A (en) * | 2016-01-21 | 2016-04-06 | 北京百度网讯科技有限公司 | Instruction execution method and device |
| CN106779230A (en) * | 2016-12-27 | 2017-05-31 | 北京三快在线科技有限公司 | Order volume predictor method and device, electronic equipment |
| CN107146089A (en) * | 2017-03-29 | 2017-09-08 | 北京三快在线科技有限公司 | The single recognition methods of one kind brush and device, electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 孟佩等: "基于Softmax回归模型的协同过滤算法研究与应用", 《计算机技术与发展》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110765346A (en) * | 2018-07-10 | 2020-02-07 | 阿里巴巴集团控股有限公司 | User intention mining method, device and equipment |
| CN110765346B (en) * | 2018-07-10 | 2023-04-18 | 阿里巴巴集团控股有限公司 | User intention mining method, device and equipment |
| CN109087135A (en) * | 2018-07-25 | 2018-12-25 | 百度在线网络技术（北京）有限公司 | The method for digging and device, computer equipment and readable medium that user is intended to |
| US11138381B2 (en) | 2018-07-25 | 2021-10-05 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method, computer device and readable medium for user's intent mining |
| CN109299994A (en) * | 2018-07-27 | 2019-02-01 | 北京三快在线科技有限公司 | Recommended method, device, equipment and readable storage medium storing program for executing |
| CN109492104A (en) * | 2018-11-09 | 2019-03-19 | 北京京东尚科信息技术有限公司 | Training method, classification method, system, equipment and the medium of intent classifier model |
| CN109582865A (en) * | 2018-11-19 | 2019-04-05 | 北京奇虎科技有限公司 | A kind of method and device of pushing application program |
| CN109710941A (en) * | 2018-12-29 | 2019-05-03 | 上海点融信息科技有限责任公司 | User's intension recognizing method and device based on artificial intelligence |
| CN110162535A (en) * | 2019-03-26 | 2019-08-23 | 腾讯科技（深圳）有限公司 | For executing personalized searching method, device, equipment and storage medium |
| CN110162535B (en) * | 2019-03-26 | 2023-11-07 | 腾讯科技（深圳）有限公司 | Search method, apparatus, device and storage medium for performing personalization |
| WO2020258836A1 (en) * | 2019-06-24 | 2020-12-30 | 创新先进技术有限公司 | Method and apparatus for sensing transaction user in advance |
| US11216832B2 (en) | 2019-06-24 | 2022-01-04 | Advanced New Technologies Co., Ltd. | Predicting future user transactions |
| CN111177571A (en) * | 2020-01-13 | 2020-05-19 | 海南大学 | Value-driven dynamic recommendation system with multi-factor dimension space and multi-mesoscale fusion |
| CN111400473A (en) * | 2020-03-18 | 2020-07-10 | 北京三快在线科技有限公司 | Method and device for training intention recognition model, storage medium and electronic equipment |
| CN112711717A (en) * | 2021-03-26 | 2021-04-27 | 北京三快在线科技有限公司 | Room product searching method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108230007B (en) | 2021-09-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108230007A (en) | A kind of recognition methods of user view, device, electronic equipment and storage medium | |
| CN111444428B (en) | Information recommendation method and device based on artificial intelligence, electronic equipment and storage medium | |
| US11397772B2 (en) | Information search method, apparatus, and system | |
| CN111382361B (en) | Information pushing method, device, storage medium and computer equipment | |
| CN111898032B (en) | Information recommendation method and device based on artificial intelligence, electronic equipment and storage medium | |
| KR20200003106A (en) | Information retrieval methods, devices and systems | |
| CN105677780A (en) | Scalable user intent mining method and system thereof | |
| WO2019016614A2 (en) | Method and apparatus for displaying search results | |
| CN112765480B (en) | Information pushing method and device and computer readable storage medium | |
| CN110555753A (en) | recommendation-based ranking control method and device, computer equipment and storage medium | |
| CN106610970A (en) | Collaborative filtering-based content recommendation system and method | |
| CN110647696A (en) | Business object sorting method and device | |
| CN114265979A (en) | Method for determining fusion parameters, information recommendation method and model training method | |
| CN109690581A (en) | User guided system and method | |
| CN113793182A (en) | Commodity object recommendation method and device, equipment, medium and product thereof | |
| CN116049536A (en) | Recommendation method and related device | |
| CN111310035A (en) | Recommendation method and device based on psychological and behavioral characteristics | |
| CN109118282B (en) | Dual-mode mutual inductance intelligent space user portrait management method and terminal | |
| CN111177564B (en) | Product recommendation method and device | |
| CN111445280A (en) | Model generation method, restaurant ranking method, system, device and medium | |
| Bharadhwaj | Layer-wise relevance propagation for explainable recommendations | |
| CN116401450A (en) | Recommendation method and device based on knowledge graph and computer equipment | |
| CN116308640A (en) | Recommendation method and related device | |
| CN111831892A (en) | Information recommendation method, information recommendation device, server and storage medium | |
| CN115795156A (en) | Material recall and neural network training method, device, equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |